Privacy Protection Cross-Domain Experiment Group Division and Monitoring

By using a secure multi-party computation system and a privacy-preserving identifier (PEID), the problems of user privacy leakage and unstable cross-domain experiment results caused by cookies are solved, achieving stable cross-domain experiment results and user privacy protection.

CN115943375BActive Publication Date: 2026-11-17GOOGLE LLC
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202280004199.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2021-06-03
Filing Date
2022-04-05
Publication Date
2026-11-17
Estimated Expiration
2042-04-05

AI Technical Summary

Technical Problem

Existing monitoring mechanisms such as cookies are prone to user privacy leaks, and after abandoning third-party cookies, cross-domain experiment results are unstable, leading to inaccurate results and wasted resources.

Method used

A secure multi-party computation (MPC) system is adopted to ensure user privacy and provide stable cross-domain experimental results through cross-network domain collaboration and the use of privacy-preserving identifiers (PEIDs) to identify experimental and control groups.

Benefits of technology

It enables the provision of stable cross-domain experimental results while protecting user privacy, reducing resource waste, and improving the accuracy of experimental results and user trust.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115943375B_ABST
    Figure CN115943375B_ABST
Patent Text Reader

Abstract

Methods, systems, and apparatus are described, including computer programs encoded on a computer storage medium for privacy-preserving cross-domain experiment monitoring. One method includes: receiving a request for digital content by a first server of an MPC system, the digital content including a first secret share of an application instance identifier identifying an application instance associated with the device. The first server collaborates with a second server of the secure MPC system to perform a privacy-preserving selection process to select a winning digital component from a set of digital components. Each digital component has a corresponding unique experiment identifier and a unique control identifier. A first secret share representing the winning digital component is generated. A response is generated, and the response includes the first secret share of the selection result, and data indicating whether the application for each digital component is in the experiment group or the control group.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This specification covers data processing, encryption, and privacy-preserving web activity analysis. Background Technology

[0002] Persistent identifiers, which uniquely identify users across different experiments, domains, and requests, allow for experiments using user groups that are consistently exposed to processing or retained as part of control. However, existing monitoring mechanisms, such as cookies, are often used as persistent identifiers and can be linked to individual users and information about that user. Such vulnerabilities can make users feel that they are too easily identified and that their information is too easily leaked. Furthermore, cookies can provide more information than users would like to share and are associated with specific content providers, owners, or distributors. Summary of the Invention

[0003] Enhanced online user privacy has led many browser developers to change how they handle user data. For example, some browsers no longer support certain types of cookies, but deprecating third-party cookies can lead to less consistent experimental results. The described system and method allow for real-time experimentation by serving users with different forms of content across different domains and locations on the network, even when cookies are unavailable. The described identifier provides a privacy-preserving mechanism to consistently identify whether an application is part of an experimental group or a control group within an experiment by allowing the system to conduct the experiment, thus providing accurate experimental results across different experiments and domains, and enabling cross-domain experimental group segmentation and monitoring.

[0004] Generally, an innovative aspect of the subject matter described in this specification can be an embodiment of a method for monitoring privacy-preserving cross-domain experiments, the method comprising: receiving a request for digital content from an application on a user device by a first server of a secure multi-party computation (MPC) system, wherein the request includes a first secret share identifying an application instance associated with the device or an application instance identifier of the device itself; in response to receiving the request and in cooperation with a second server of the secure MPC system, performing a privacy-preserving selection process to select a first winning digital component from a set of digital components, wherein each digital component has a corresponding (i) unique experiment identifier and (ii) unique control identifier, the privacy-preserving selection process comprising, for each digital component in the set of digital components, in cooperation with the second server and based on a set of eligibility rules, determining a first secret share of candidate parameters indicating whether the digital component is a qualified candidate for selection in the privacy-preserving selection process, and in cooperation with the second server and based on the first secret share of the application instance identifier and the corresponding first secret share of the candidate parameter value, determining a secret share of experimental parameters indicating whether the application is in the experimental group of the digital components. The method includes: generating a first secret share representing the selection result of a first winning digital component based on a first secret share of candidate parameters for each digital component and a selection value for each digital component; generating a response to a digital content request, the response including: the first secret share of the selection result, and, for each of one or more digital components including the first winning digital component, data indicating whether the application is in the experimental group or the control group of the digital component; and transmitting the response to the digital content request from a first server of a secure MPC system to the application on the user device.

[0005] In some implementations, data indicating whether an application is in the experimental group or the control group for digital components is represented by a Bloom filter. In some implementations, performing a privacy-preserving selection process to select the first winning digital component includes, for each digital component in the set of digital components, adding a corresponding (i) unique experimental identifier if the application is in the experimental group, or (ii) a unique control identifier if the application is in the control group.

[0006] In some implementations, determining whether an application is part of a control group or an experimental group for each digital component includes determining whether the application is part of a control group or an experimental group for each experiment in the set of experiments associated with the digital component.

[0007] In some implementations, the set of eligibility rules includes two or more eligibility rules, where one or more parameters of a first eligibility rule are determined based on an experimental design based on a plurality of mutually orthogonal Latin matrices of a specific size, and the first eligibility rule specifies a first range of values ​​for the one or more parameters. In some implementations, the method includes updating one or more parameters of the first eligibility rule based on the experimental design to specify a second range of values ​​for the one or more parameters, and performing a counterfactual selection process, in collaboration with a first server of the secure MPC system and a second server of the secure MPC system, to select a second winning selection criterion from a set of selection criteria by applying a set of eligibility rules including the updated first eligibility rule.

[0008] In some implementations, the transmission of a response to a request for digital content from a first server of a secure MPC system to an application on a user device includes: the first server of the secure MPC system receiving an encrypted secret share of the response to the request for digital content from a second server of the secure MPC system; the first server of the secure MPC system encrypting a second secret share of the response to the request for digital content to generate an encrypted second secret share of the response; and the first server of the secure MPC system transmitting the encrypted secret share of the response and the encrypted second secret share of the response to the application on the user device.

[0009] In some implementations, the qualification rule set includes one or more qualification rules, which include at least one positive condition, and the selection criterion set is stored at the secure MPC system. In some implementations, the qualification rule set includes at least one negative condition.

[0010] In some implementations, attaching a unique experimental identifier associated with selection criteria to a response to a request for digital content includes adding the unique experimental identifier to a list of unique identifiers included in the response, and attaching a unique control identifier associated with selection criteria to a response to a request for digital content includes adding a unique control identifier to a list of unique identifiers, wherein the list of unique identifiers is represented using a Bloom filter. In some implementations, the method includes: transmitting a first report including a list of unique identifiers from an application to a platform associated with the selected digital content; and uploading a batch of one or more reports including the first report to an aggregated report server from a platform associated with the selected digital component. In some implementations, the method includes transmitting the first report including a list of unique identifiers from an application to an aggregated report server.

[0011] Other embodiments of this aspect include corresponding systems, apparatuses, and computer programs configured to perform actions of methods encoded on computer storage devices.

[0012] The subject matter described in this specification can be implemented in specific embodiments to achieve one or more of the following advantages.

[0013] The selection and presentation of digital content are frequently updated and improved through experimentation, which tests user responses or reception of digital content based on changes in the content itself. User-based experiments are an important technique for assessing how user experience is affected, where the impact of specific features or presentation formats on users receiving the content is measured. Such experiments also allow content researchers, providers, developers, and other entities to quantify the incremental impact of specific features and changes under development and to improve provider effectiveness and target key performance indicators. For example, experiments allow content providers to measure metrics and the impact of the selection and presentation of specific digital content on specific metrics, such as publisher revenue or advertiser reach, as well as audience size and demographics.

[0014] Historically, the selection and implementation of changes to content such as digital components in digital content systems (e.g., generating experimental or incremental changes to digital components) have involved the use of user information (e.g., browsing information, interest group information, etc.) obtained from third-party cookies stored on the client device by a domain different from the domain of the webpage rendered on the client device. In settings such as experiments designed and executed to test the design and presentation of digital content, or other digital experiments, these third-party cookies allow for the consistent assignment of specific users to a control group or processing group exposed to processing (such as different content or content formats). However, browsers block the use of third-party cookies, making it more difficult to consistently assign users to control groups or multiple groups exposed to processing. This means that assigning users to conflicting groups and producing less accurate or inaccurate results can waste computational resources and bandwidth. When users are not consistently assigned to multiple groups, results can be inaccurate because a user can be exposed to processing in one experiment on one publisher's website and exposed to the same experimental control on different publisher's websites. This can confound results, making the results obtained from conducting the experiment unreliable and / or unusable. When control or processing is provided inconsistently, the content may be irrelevant, and new content or experiments will need to be provided. This may require performing experiments on different users or with different content, resulting in additional traffic while invalidating previously provided traffic.

[0015] To overcome this problem, privacy-preserving technologies that use stable cross-domain identifiers for experimental purposes can monitor and analyze network activity while preventing user tracking and leakage of user information between computing systems. This enables important experiments to be conducted while protecting user privacy and data security.

[0016] The described privacy-preserving monitoring mechanism—a privacy-preserving experimental identifier or "PEID"—improves user experience and increases user trust by providing privacy guarantees that this mode of monitoring user data and activity cannot be accessed by any third party, while still providing the benefit of being able to execute stable research methodologies, including experimentation and testing of various content and presentation methods. The described system includes a secure server group performing multi-party computations, with distributed computation across the parties ensuring that no single party can access data in plaintext via cryptographic protocols. This ensures that no participating party can access user data in plaintext without unauthorized collusion among the parties. The described system is capable of supporting stable cross-device partitioning of experimental groups (e.g., control group vs. experimental group) if, for example, applications (such as web browsers) on different devices belonging to the same user can synchronize with each other to use the same PEID. The described system supports cross-domain experimental partitioning on the same device and, in the case of cross-device synchronization of PEIDs, also supports cross-device experimental partitioning.

[0017] Furthermore, these experimental identifiers can be used as alternatives in systems using traditional web activity monitoring methods. For example, PEID can be used in existing systems with minimal adaptation under certain conditions, allowing system designers to reuse existing infrastructure to provide relevant statistics on web activity and implement security features to protect third parties while improving user privacy.

[0018] Details of one or more embodiments of the subject matter described in this specification are set forth in the accompanying drawings and the following description. Other features, aspects, and advantages of the subject matter will become apparent from the description, drawings, and claims. Attached Figure Description

[0019] Figure 1 This is a block diagram of the system that implements cross-domain experimental grouping and monitoring mechanisms to achieve privacy protection.

[0020] Figure 2 This is a data flow diagram of an example process for selecting digital components to provide a display or distribution to client devices.

[0021] Figure 3 This is a swimlane diagram illustrating an example process for issuing and implementing privacy-preserving cross-domain experimental group partitioning and monitoring mechanisms to report experimental data.

[0022] Figure 4 This is a swimlane diagram illustrating an example process for issuing and implementing privacy-preserving cross-domain experimental group partitioning and monitoring mechanisms to support incremental changes in experiments.

[0023] Figure 5This is a flowchart illustrating an example process for issuing and implementing privacy-preserving cross-domain experimental group partitioning and monitoring mechanisms.

[0024] Figure 6 This is a block diagram of an example computer system.

[0025] In the various figures, the same reference numerals and names indicate the same elements. Detailed Implementation

[0026] Overall, this document describes systems and techniques for protecting user data and the privacy of users who are provided with content in experimental settings.

[0027] User-based field experiments are an important technique widely adopted by various companies. These experiments can be used to evaluate the effectiveness and user experience of new features under development and to quantify the incremental effects of specific changes. For example, an experiment can be used to quantify how likely a user is to attend a concert performed by a specific orchestra after being shown, for example, a promotional video from a previous season of a show. Existing systems for conducting user-based cross-domain field experiments may be ineffective when monitoring mechanisms such as third-party cookies are not supported, lacking stable, cross-domain, and definitive user identifiers. The systems and techniques described in this document enable such experiments to be conducted while ensuring data security and user privacy.

[0028] While alternative solutions exist for conducting user-based field experiments, they may not be stable or accurate. For example, request-based experiments might be conducted, where each request for content is randomly or pseudo-randomly assigned to the experimental group. User-based experiments might also be based on first-party cookies. However, in both cases, if a user using a particular browser visits different domains and sends multiple requests for content, the browser could be assigned to the control group once and the exposed group another time. When users are inconsistently assigned to the experimental group, the effect of the variable being tested can spill over between the experimental and control groups, leading to inaccurate and therefore unusable results.

[0029] In some implementations, user-based experiments may be conducted using information such as a user's email address or phone number, which would allow for consistent, cross-domain, and cross-browser tracking. However, users may be concerned about their personally identifiable information being used for any purpose, including assigning users to experimental groups, even if the data is not exposed to any other entity.

[0030] The described system presents a secure multi-party computation (MPC) design that provides information security and privacy. Secure multi-party computation (MPC) is an encryption protocol that prevents data access by distributing computation among multiple parties, ensuring that no single party can access the data of another party. MPC allows multiple parties to collaboratively compute functions on their respective inputs while keeping those inputs private. An MPC system performs computation by splitting data that should be kept private into shares. This data, referred to as secrets, is divided into secret shares, and each of these shares is provided to the participating parties. The parties then collaboratively compute the function on their respective secret shares, thereby ensuring the confidentiality, integrity, and security of the data. An MPC system comprises a cluster of two or more server computers that collaboratively and securely perform various computations and processes.

[0031] For example, the system can distribute experimental design functions, such as assigning specific PEIDs to control or processing groups, and securely transmit PEIDs used for experiments. Compared to performing the process at the requesting client device (which requires transferring securely selected candidate digital components from the MPC cluster to the client device), the MPC cluster provides improved process integrity and user privacy. The level of security provided by the MPC process is that there is no way to expose user data to any entity if only one presumed non-conspirator remains honest and does not deviate from the cryptographic protocol design. The system follows an honest but curious security model, where all parties honestly follow the protocol, and the protocol remains private if any number of conspiring parties only receive their own output.

[0032] By implementing such a system, in which stable identifiers are used only for experimental purposes and transmitted via a secure MPC method, the system can maintain the integrity of experiments and the utility of experimental results, while protecting user privacy, providing content that is more likely to be relevant to users, and delivering content in a way that is more likely to be of interest and / or useful to users.

[0033] The described system enhances the operation of a secure MPC design by providing stable experiment identifiers that support user-based experiments across domains, where PEIDs are synchronized across multiple devices owned by the same user. Applications such as browsers—through which users access specific electronic files and resources—share these stable experiment identifiers with the secure MPC cluster, which can use these identifiers to enable or disable various experiments and experiment features, determine which of various storage selection criteria are eligible or ineligible for the choices presented to users, assign users to different experiment groups, track experiment results, and report those results for use, among other functions. The secure MPC cluster reports enabled experiments to an aggregate reporting server for analysis. In some implementations, the secure MPC cluster performs reporting via a browser, or the secure MPC cluster itself can act as an aggregate reporting server.

[0034] Additionally, to support incremental updates and measurement of changes, if the winning selection criteria and associated digital components are part of an incremental experiment or study, the secure MPC cluster performs a counterfactual selection process during the selection process to determine whether a browser user belongs to the reserved group of the winning selection criteria (i.e., the control group to which they are not provided with processing features). In response to requests for digital components, the secure MPC cluster can request applications, such as browsers, to identify whether they belong to an exposed group or a reserved group (if applicable). The application reports via an aggregated reporting server whether the associated behavior (such as visiting a website, purchasing symphony tickets, issuing an electronic birthday card) occurred within a transition window of all exposed and reserved groups associated with the application.

[0035] Figure 1 This is a block diagram of environment 100, in which MPC cluster 130 performs a secure MPC process to select digital components to distribute to client devices 110 and perform experiments based on different content and content versions. Example environment 100 includes a data communication network 105, such as a local area network (LAN), wide area network (WAN), internet, mobile network, or a combination thereof. Network 105 connects client devices 110, secure MPC cluster 130, publisher 140, website 142, content platforms such as supply-side platform (SSP) 170 and demand-side platform (DSP) 150, and aggregation report server 180. Example environment 100 may include many different client devices 110, secure MPC cluster 130, publisher 140, website 142, DSP 150, SSP 170, and aggregation report server 180.

[0036] Client device 110 is an electronic device capable of communicating via network 105. Example client device 110 includes a personal computer, a mobile communication device, such as a smartphone, and other devices capable of sending and receiving data via network 105. The client device may also include a digital assistant device that accepts audio input via a microphone and outputs audio via a speaker. When the digital assistant detects a “hot word” or “hot phrase” that activates the microphone to accept audio input, the digital assistant can be put into listening mode (e.g., ready to accept audio input). The digital assistant device may also include a camera and / or display to capture images and visually present information. The digital assistant can be implemented in various forms of hardware devices, including wearable devices (e.g., watches or glasses), smartphones, speaker devices, tablet devices, or other hardware devices. The client device may also include a digital media device, such as a streaming device that plugs into a television or other display to stream video to a television, game console, or virtual reality system.

[0037] Client device 110 typically includes application 112, such as a web browser and / or native applications, to facilitate the sending and receiving of data over network 105. Native applications are applications developed for a specific platform or device (e.g., a mobile device with a specific operating system). In some implementations, application 112 can be a program such as an operating system. Publisher 140 can develop and provide, for example, downloadable native applications to client device 110. A web browser can request resource 145 from a web server hosting website 142 of publisher 140, for example, in response to a user of client device 110 entering a resource address for resource 145 in the web browser's address bar or selecting a link referencing the resource address. Similarly, native applications can request application content from a publisher's remote server.

[0038] Some resources, application pages, or other application content may include digital component slots for presenting digital components using resource 145 or application pages. The phrase "digital component" as used throughout this document refers to a discrete unit of digital content or digital information (e.g., a video clip, audio clip, multimedia clip, image, text, or another unit of content). Digital components may be stored electronically as a single file or in a collection of files on a physical storage device, and digital components may take the form of video files, audio files, multimedia files, image files, or text files, and may include advertising information, making advertising a type of digital component. For example, a digital component may be content designed to complement web pages or other resource content presented by application 112. More specifically, a digital component may include digital content related to the resource content (e.g., a digital component may relate to the same topic as the web page content, or to a related topic). Thus, the provision of digital components can complement and generally enhance web page or application content.

[0039] When application 112 loads resources (or application content) that include one or more digital component slots, application 112 can request digital components for each slot. In some implementations, the digital component slots can include code (e.g., a script) that enables application 112 to request digital components from an MPC cluster (such as MPC cluster 130), which selects and provides the digital components to application 112 for presentation to a user of client device 110. As described below, application 112 can request digital components from MPC cluster 130, which can forward the request or parts thereof to other servers, such as SSP 170.

[0040] Some publishers 140 use SSP 170 to manage the process of obtaining digital components for their resources and / or applications' digital component slots. SSP 170 is a technology platform implemented in hardware and / or software that automates the process of obtaining digital components for resources and / or applications. Each publisher 140 is capable of having one or more SSP 170s. Some publishers 140 may use the same SSP 170.

[0041] Digital component provider 160 is capable of creating (or otherwise distributing) digital components that are presented in digital component slots within the resources and applications of publishers. Digital component provider 160 can use DSP 150 to manage the provision of its digital components for presentation in digital component slots. DSP 150 is a technology platform implemented in hardware and / or software that automates the distribution of processes for presenting digital components using resources and / or applications. DSP 150 can interact on behalf of digital component provider 160 with multiple supply-side platforms (SSPs) to provide digital components for presentation through resources and / or applications of multiple different publishers 140. Generally, DSP 150 can receive requests for digital components (e.g., from SSPs), generate (or select) selection criteria for one or more digital components created by one or more digital component providers based on the requests, and provide the SSPs with data related to the digital components (e.g., the digital components themselves) and selection parameters. The selection criteria can indicate the amount that digital component provider 160 is willing to provide for presentation or user interaction with the digital components. Then, the SSP is able to select the digital components to be presented at the client device 110 and provide the client device 110 with data that enables the client device 110 to present the digital components.

[0042] In some cases, it is beneficial for users to receive digital components associated with web pages, application pages, or other electronic resources that the user has previously accessed and / or interacted with. To distribute such digital components to users, when a user accesses a specific resource or performs a specific action at that resource (e.g., interacting with a specific item presented on a web page or adding an item to a virtual shopping cart), the user can be assigned to user groups, such as user interest groups, groups of similar users, or other group types involving similar user data. User groups can be generated by the digital component provider 160. That is, each digital component provider 160 can assign users to their respective user groups when a user accesses an electronic resource provided by the digital component provider 160. User groups can also be created by content platforms, such as those created by the DSP 150 and / or SSP 170.

[0043] To protect user privacy, user group memberships can be maintained at the user's client device 110, for example, by one of the applications 112 or the operating system of client device 110, rather than by the digital component provider, content platform, or other party. In a specific example, a trusted program (e.g., a web browser or operating system) can maintain a list of user group identifiers ("user group list") for users using a web browser or another application (e.g., users logging into the browser, application, or client device 110). The user group list can include a group identifier for each user group, and each user group includes users as members. The digital component provider 160 that creates the user groups can assign user group identifiers to its user groups. The user group identifier used for a user group can be a description of the group (e.g., a gardening group) or a code representing the group (e.g., not a descriptive alphanumeric sequence). The user group list used for a user can be stored in a secure storage device at client device 110 and / or can be encrypted at storage to prevent access to the list by others.

[0044] When application 112 presents resources or application content related to web pages on digital component provider 160 or website 142, the resource may request application 112 to add one or more user group identifiers to the user group list. In response, application 112 may add one or more user group identifiers to the user group list and securely store the user group list.

[0045] MPC cluster 130 can use data including user group membership to select digital components or other content that may be of interest to the user or may otherwise benefit the user / user device. For example, such digital components or other content may include data that improves user experience, enhances the operation of the user device, or otherwise benefits the user or user device. However, the user group identifier of the user's user group list can be provided and used to select digital components to prevent the computing systems MPC1 and MPC2 of MPC cluster 130 from accessing the user's user group identifier in plaintext, thereby protecting user privacy when selecting digital components using user group membership data. Plaintext is text or data (including binary files) without computational tags, special formatting, or code, in a form that can be viewed or used without the need for keys or other decryption devices or processes. MPC cluster 130 also uses a set of candidate selection criteria received from SSP 170 to perform a final selection process to select the winning selection criterion mapped to the digital component. Then, in response to a digital component request, MPC cluster 130 directly provides the client device 110 with the data for mapping the digital component to the winning selection criterion. The data may include parameters such as data identifying the digital component, the location where the digital component is stored and / or the location from which the digital component can be requested, and winning selection criteria. In some implementations, the MPC cluster 130 provides the digital component itself to the client device 110.

[0046] Secure MPC cluster 130 includes two computing systems, MPC1 and MPC2 (e.g., server computers), that perform secure MPC processes to select digital components for distribution to users' client devices based on user group membership and / or other criteria, without explicit access to group membership information (or other sensitive user information). While the example MPC cluster 130 includes two computing systems, more computing systems can be used as long as the MPC cluster 130 includes more than one. For example, MPC cluster 130 can include three, four, or another suitable number of computing systems. Using more computing systems in MPC cluster 130 can provide greater security, but it can also increase the complexity of the MPC process.

[0047] Computing systems MPC1 and MPC2 can be operated by a single entity. In some implementations, MPC1 and MPC2 can be operated by different entities. In this way, each entity may not have plaintext access to a user's group membership. For example, one of the computing systems MPC1 or MPC2 can be operated by a trusted party different from the user, publisher 140, DSP 150, SSP 170, aggregation report server 180, and digital component provider 160. For example, industry groups, government groups, or browser developers can maintain and operate one of the computing systems MPC1 and MPC2. Another computing system can be operated by a different one of these groups, such that different trusted parties operate each computing system MPC1 and MPC2. Preferably, the different parties operating different computing systems MPC1 and MPC2 have no incentive to collude to compromise user privacy. In some implementations, computing systems MPC1 and MPC2 are architecturally separate and are monitored to not communicate with each other outside of performing the secure MPC processes described in this document.

[0048] In addition to the descriptions throughout this document, users may be provided with controls (e.g., user interface elements that users can interact with) that allow them to choose whether and when the systems, programs, or functions described herein may be able to collect user information (e.g., information about the user's social networks, social actions or activities, occupation, user preferences, or the user's current location), and whether to transmit content or communications from a server to the user. Furthermore, some data may be processed in one or more ways before being stored or used to remove personally identifiable information. For example, a user's identity may be processed so that no personally identifiable information about the user can be determined, or the user's geographic location may be generalized (e.g., down to a city, zip code, or state level) if location information is available, so that the user's specific location cannot be determined. Therefore, users can control what information about themselves is collected, how that information is used, and what information is provided to them.

[0049] Figure 2 This is a data flow diagram of an example process 200 for selecting digital components to provide a display or distribution to a client device. The operation of process 200 can be implemented, for example, by an application program 112 on client device 110, computing systems MPC1 and MPC2 of the MPC cluster 130, DSP 150, and SSP 170. The operation of process 200 can also be implemented as instructions stored on one or more computer-readable media that may be non-transitory, and execution of these instructions by one or more data processing devices can cause one or more data processing devices to perform the operation of process 200. Although process 200 and other processes are described below with respect to two computing systems (MPC cluster 130), similar processes can also be performed using an MPC cluster having more than two computing systems.

[0050] This description includes two types of selection criteria—selection criteria conditioned on sensitive user information, such as user group membership or other business-sensitive information, or parameters whose values ​​could allow unscrupulous parties to infer sensitive information, or "conditional selection criteria," and selection criteria not conditioned on sensitive information, or "unconditional selection criteria." To protect user privacy, the conditions used for "conditional selection criteria" are evaluated within the MPC cluster 130, rather than within the SSP 170 or DSP 150, to determine whether the digital components corresponding to the conditional selection criteria are eligible for use in the content selection process.

[0051] This architecture allows MPC cluster 130 to protect user privacy and trade secrets, and to demonstrate its trustworthiness to application providers, such as the provider of application 112. In this example, MPC cluster 130 relies on a secure two-party computation (2PC) architecture, which applies cryptographic techniques to ensure that there is no leakage of confidential user data or trade secrets if at least one of the two computing systems in MPC cluster 130 is honest. If MPC cluster 130 includes more than two computing systems, the current MPC protocol can be extended, or other MPC protocols can be used.

[0052] The MPC cluster 130 runs a secure 2PC protocol to evaluate and apply conditions to select eligible candidates for the selection process, conduct the selection process to select digital components based on winning selection criteria (e.g., where digital components are mapped to winning selection criteria), and receive exposure notifications to update the counters on which those conditions depend. All of these processes can be accomplished using a secret share technique. The protocol is described in detail below.

[0053] Process 200 begins in phase A, where application 112 collaborates with triggering elements from the content distribution server, such as SSP 170, to send requests for digital components to MPC cluster 130. Application 112 can combine multiple requests for digital components into a single composite request to pick up multiple digital components. MPC cluster 130 can then serve each request in the composite request independently or make one or more selection decisions holistically. In this example, the request is for a single digital component and includes requests for digital components selected based on sensitive information or those selected without sensitive information.

[0054] MPC cluster 130 can respond to requests by selecting specific digital components from a set of digital components, each mapped to a corresponding selection criterion. These selection criteria can be selection criteria for digital components previously cached or otherwise stored at MPC cluster 130, and / or selection criteria generated by a platform, such as DSP 150 or SSP170, or Just-In-Time (JIT) selection criteria. JIT selection criteria are generated directly in response to needs, increasing efficiency and reducing waste because selection criteria are generated only when a digital component is needed. For example, JIT selection criteria can be generated when a digital component slot becomes available—indicated by the receipt of a request for a digital component. The cached digital components are those selected by the MPC computer, but the digital components are not necessarily stored in the cache; for example, the digital components or their data can be stored in other types of data storage devices or memories.

[0055] The request includes information used in the digital component selection process, including potentially sensitive information such as identifiers of user groups mapped or otherwise associated with application 112, and non-sensitive information such as contextual signals from application 112 regarding the context in which the digital component will be presented and / or displayed. As described in further detail below, system 110 is designed to improve the protection of user data that may be sensitive or confidential. The request includes a stable experiment identifier (PEID) that can be used to select or remove selection criteria and associated digital components from the selection process. The stable experiment identifier can correspond to an instance of application 112 running on client device 110, which in turn can correspond to a user, such as a user logged into or using application 112. The stable experiment identifier is encrypted using a secret share technique supported by MPC cluster 130 to protect user privacy, ensuring that no single server in MPC cluster 130 is aware of the identifier and, in turn, stores a share of the identifier. Application 112 can send the corresponding share of the stable experiment identifier to each server in MPC cluster 130.

[0056] The triggering element can be, for example, a tag that detects the presence of a digital component slot within an internet location or other electronic resource accessed by application 112. The triggering element can be placed, for example, at an internet location and can notify application 112 of the presence of a digital component slot for which it should request a digital component.

[0057] Process 200 continues to Phase B, in which MPC cluster 130 transmits a digital component request to SSP 170 based on non-sensitive (or less sensitive) information, such as context signals. For example, a digital component request could be based on information such as user activity information within a specific domain. This request is referred to as a "context request," and the response to such a context request can be an unconditional selection criterion because the request is not conditional on receiving sensitive information. In some cases, the response may include criteria against one or more conditions, making the response conditional. Context requests can contain various context signals and user information collected directly by the internet location that triggered the request for the digital component (e.g., a content publisher). For example, context signals can include analytics data, language settings, and other data that help content publishers provide a good user experience. However, the context request provided to SSP 170 does not include sensitive information such as user group identifiers.

[0058] Process 200 continues to stage C, in which SSP 170 forwards the scenario request to one or more DSPs 150. In this particular example, and for simplicity, SSP 170 forwards the scenario request to a single DSP 150. For example, SSP 170 is capable of forwarding the scenario request to DSP 150. In this example, DSP 150 has digital components and selection criteria mapped to the digital components.

[0059] Process 200 continues to stage D, in which one or more DSPs 150 return selection criteria in response to a context request. For example, DSP 150 returns one or more selection criteria mapped to a digital component in response to the context request. DSP 150 is capable of returning any number of selection criteria in response to a context request. In some implementations, DSP 150 is also capable of returning selection criteria in response to a digital component request based on sensitive information such as user group information. These selection criteria are “conditional selection criteria” because they are conditional on sensitive information, and therefore conditional on MPC cluster 130 receiving a request that includes sensitive information matching the sensitive information conditional on the selection criteria. For each selection criterion provided by DSP 150, DSP 150 includes information such as a Time-to-Live (TTL) parameter, i.e., the maximum time span for which MPC cluster 130 can cache the selection criteria. This TTL parameter enables MPC cluster 130 to cache selection criteria received from DSP 150. In some implementations, without a TTL parameter, the MPC cluster 130 does not cache the received selection criteria and instead discards them after the selection criteria have been used in the selection process, for example, in the selection process corresponding to the digital component requests transmitted in phases A, B, and C.

[0060] Process 200 continues to phase E, in which SSP 170 applies content selection criteria. For example, SSP 170 applies rules such as content provider and digital component blocking rules, which prevent certain content providers from qualifying for candidate digital components and selection criteria, or prevent certain digital components from being candidates. In some implementations, MPC cluster 130 is able to apply rules or portions of rules.

[0061] In some implementations, application 112 can maintain a set of blocked identifiers, which includes identifiers of digital components that have been explicitly blocked by the user. For example, suppose a user interacts with a mute element corresponding to a given digital component (or a portion of data delivered to client device 110). In this example, the user's interaction with the mute element is an indication that the user does not want to see that digital component or potentially similar digital components (e.g., from the same or different content source or activity) in the future (e.g., after at least a certain amount of time). In response to the interaction, application 112 can add the identifier for the given digital component to a set of blocked generic identifiers, which can then be used to prevent the delivery or presentation of other digital components with the same identifier as the given digital component. In some implementations, application 112 can maintain a set of blocked identifiers that includes identifiers of digital components that have recently been shown to the user too frequently, for example, more than X times in the recent Y minutes. Application 112 can add the identifier for a given digital component to a set of blocked generic identifiers, which can then be used to prevent the delivery or presentation of other digital components with the same identifier as the given digital component in the near future.

[0062] SSP 170 also applies selection criterion rules to determine, for example, how selection criteria affect the post-release value of a particular content provider. The post-release value can indicate, for example, the amount of digital components provided to the content provider for distribution. SSP 170 then performs a selection process to determine a top unconditional selection criterion that produces the highest post-release value. The unconditional selection criterion is not conditional on sensitive information and therefore does not apply content selection rules such as user group membership, frequency control, content distribution rules, and pacing rules. SSP 170 then forwards the following criteria, which serve as JIT selection criteria, to MPC cluster 130: all selection criteria that can be cached in MPC cluster 130 (selection criteria with TTL values), and all selection criteria whose post-release value is not less than the post-release value of the top unconditional selection criterion.

[0063] Process 200 continues to phase F, in which MPC cluster 130 updates its cache using the received cache-enabled digital components (i.e., those with TTL values) and their corresponding JIT selection criteria. Additionally, MPC cluster 130 applies selection criteria to conditions such as user membership rules, frequency control, pacing rules, and rules preventing specific users from micro-targeting all selection criteria received in phase E, as well as selection criteria for previously cached digital components to select valid candidates for the selection process. Rules can include restrictions and guidelines on the distribution or frequency of digital components, and other factors. Rules include frequency control, mute, budget, and pacing constraints. MPC cluster 130 then runs a final selection process among all valid candidates, selects the winning digital component, and then returns data for the winning digital component to application 112 in response to the digital component request.

[0064] Process 200 continues to phase G, in which the winning digital component is rendered by application 112. Application 112 then provides an exposure notification to MPC cluster 130. This exposure notification includes data that allows MPC cluster 130 to update information related to updating counters that allow MPC cluster 130 to enforce selection rules. In some implementations, application 112 can reduce the amount of network communication and battery / bandwidth consumption for mobile devices by piggybacking on, for example, a future component request A.

[0065] The system implements a secret share algorithm to prevent any single server in the MPC cluster from accessing user information, such as peid. When the system operates under the honest but curious assumption, the secret share algorithm guarantees that no single server in the MPC can see peid in plaintext. More specifically, for each digital component request, browser 112 divides peid into two secret shares, [peid1] and [peid2], where the parameter appearing in parentheses indicates the secret share of the variable within the parentheses. For example, here [peid1] and [peid2] are two secret shares of peid, which are stable experimental identifiers for the user. Regarding the secret shares, subscript 1 indicates the parameter representing the first secret share, and subscript 2 indicates the parameter representing the second secret share. These secret shares can be reconstructed from the original secret. For example, an entity can only reconstruct the original secret peid from the secret shares [peid1] and [peid2] if it has access to both [peid1] and [peid2]. However, the secret shares are encrypted to protect user privacy.

[0066] By relying on schemes such as additive secret shares, the reconstruction process can be performed using simple operations. For example, the reconstruction process using simple additive secret shares can be performed using simple additive operations such as [peid1] + [peid2] = peid. Each of the two secret shares is separately and confidentially stored by one of the two servers in the MPC cluster 130. The sharing of peid between the servers is done through additive sharing. peid can be shared in other ways, such as through two random vectors, where the XOR of the two random vectors gives the peid, or through two separate shares, where one share holds the encrypted peid and the second share holds the decryption key, and other methods. The MPC cluster 130 can be adapted to perform the division of the key into multiple shares based on the specific method of selection.

[0067] To ensure that a secret share used by one server is not intercepted by another server, the secret share can be encrypted. For example, application (which could be a browser) 112 can encrypt each of the secret shares using the public key of the specific MPC server that intends to receive the secret shares, by... ([peid1], The result of 1) is sent to MPC1, and ([peid2], The result of step 2) is sent to MPC2. Here, ( x , y This indicates the use of the server's public key. y exist x The encryption function executed on top. Therefore, ([peid1], 1) indicates the cryptographic function executed on [peid1] using the public key of MPC1.

[0068] Figure 3 and Figure 4 This is a swimlane diagram illustrating example processes 300 and 400 used to issue and implement monitoring mechanisms for privacy-preserving cross-domain experimental group activities performed across different domains. The operation of processes 300 and 400 can be implemented, for example, via client device 110 and / or application 112, MPC cluster 130, DSP 150, and / or aggregation reporting server 180. The operation of processes 300 and 400 can also be implemented as instructions stored on one or more computer-readable media that may be non-transitory, and execution of these instructions by one or more data processing devices can cause one or more data processing devices to perform the operation of processes 300 and 400.

[0069] about Figure 3Process 300 begins with step 1, in which application 112 sends a request for a digital component to MPC cluster 130. For example, step 1 of process 300 can be similar to phase A of process 200, where application 112, in cooperation with a triggering element from a content distribution server, sends a request for a digital component to MPC cluster 130. The request can include, for example, a set of context signals related to the display environment in which the selected digital component will be displayed. The request can also include, for example, a PEID for the user and application 112, enabling the user to be uniquely identified for any experimental purposes. Context signals can include, for example, the resource locator (e.g., a Universal Resource Locator (URL)) by which the digital component will be displayed, the geographic location of the client device 110 on which the digital component will be displayed, and / or the spoken settings by which the digital component will be displayed by application 112. Other context signals can also be used.

[0070] In some implementations, the set of context signals is sent in the form of parameters such as a cache lookup key. The cache lookup key can be a composite message based on the set of context signals. In some implementations, the cache lookup key can be a cryptographic hash (e.g., SHA-256) of the composite message.

[0071] MPC cluster 130 can store a corresponding cache lookup key for each digital component for its storage or storage selection criteria. This cache lookup key indicates the context in which the digital component is eligible for selection. For example, if the selection criteria for a digital component are only eligible for use in a selection process for a given URL, the selection criteria and the digital component can be mapped to a cache lookup key that includes the given URL. These cache lookup keys can also be in the form of compound messages.

[0072] For an incoming request for a digital component, the MPC cluster 130 collects all cache selection criteria associated with the cache lookup key included in the digital component request. That is, the MPC cluster 130 identifies each digital component mapped to a cache lookup key that matches the cache lookup key in the digital component request, along with its corresponding selection criteria. These matching digital components can be considered eligible digital components because they are eligible to be displayed along with the resource for which the requested digital component is being displayed.

[0073] After identifying eligible digital components, computing systems MPC1 and MPC2 are able to determine which of the eligible digital components is a candidate for the selection process. A digital component is a candidate when each of its conditions is met. As mentioned above, these conditions can be based on user group membership, frequency control, content distribution rules, and / or pacing rules.

[0074] Eligible digital components can be filtered, and then, as mentioned above... Figure 2 The selection process described in process 200 is then used by the MPC cluster 130 in step 2 to provide the selection digital component in the response to application 112.

[0075] For each cached digital component, MPC cluster 130 calculates whether application 112 belongs to the experimental group for processing the digital component to analyze its effects, or whether application 112 belongs to the control group for not processing the digital component. Experimental or processing groups receive digital components with specific features, enhancements, interactive elements, colors, shapes, sizes, fonts, sounds, videos, etc., which digital content providers or analytics systems are interested in quantifying when provided to users on a network (such as the Internet). These groups can be different for different digital components, and MPC cluster 130 uses PEID to consistently determine whether a user should be part of the control group or the experimental group, and, based on whether the user / application belongs to the experimental group, whether the digital component is a candidate to be provided to the user and application. Application 112 is assigned to a group acting as an agent for the user.

[0076] To quantify the impact of a specific digital component or changes to a digital component on the user, each digital component cached in the MPC cluster 130 and its corresponding selection criteria can be mapped to, for example, four non-negative numbers, based on the experimental design. , , and One of the related factors is the number of variables. For example, the system can apply the concept of mutually orthogonal Latin squares to ensure that the independent variables are truly independent and without hidden confounding correlations. For example, in statistical designs for experiments, the assignment of experimental treatments such that each treatment occurs exactly once in each row and column makes it possible to use these squares to identify experimental errors from two sources. Non-negative numbers , , and These represent different parameters of the experimental design.

[0077] here, It is a prime number, which represents the number of moduli in each layer of the experiment (where each layer is represented by a Latin square matrix), and the number of rows / columns in the Latin square matrix; It is the layer identifier chosen for running the experiment, where ∈ [0, – 1]; and 0 ≤ ≤ ≤ – 1. In other words, [ , ]Limit[0, The non-empty interval within [-1] specifies the region in which experiments will be conducted. The mod in the specified layer. In this particular example, let c = peid mod ( p 2 ), x = c mod p , y = c div p ,and =( × + ) mod p .

[0078] Conceptually, the MPC cluster 130 determines whether... ∈ [ , This determines whether the cached digital components and their selection criteria are eligible for selection. In other words, the MPC cluster 130 determines whether... ≤ ( × + ) ≤ The cached digital component is then used to determine whether it is a candidate for selection. Otherwise, the cached digital component is not a candidate and therefore ineligible for selection. In some implementations, the digital component must also satisfy all the conditions associated with it in order to be a candidate for selection, as described below.

[0079] For example, MPC cluster 130 can determine whether ∈ [ , In other words, whether or not ≤ ( × + ) ≤ This is used to calculate whether application 112 belongs to the experimental group. MPC cluster 130 can determine whether... ∈ [ , In other words, determining whether ≤ ( × + ) ≤ This is used to determine whether application 112 belongs to the control group.

[0080] In addition, each cached digital component (and its corresponding selection criteria) is associated with one or more globally unique experimental identifiers (eid). experiment Associated. To enhance global uniqueness, the MPC cluster 130 can compute the identifier as related to... ∈ [ , Related eid experiment = HMAC ( _ , _ _ _ Here, HMAC, or hash-based message authentication code, is an example cryptographic hash function that can be used to simultaneously verify data integrity and message authenticity, allowing messages to be uniquely hashed. Here, globally unique means each experimental identifier eid... experiment It is unique across different domains accessible by application 112. For example, the same experimental identifier can be used across different websites and domains accessed by users of application 112 and does not overlap with any other identifier or user, regardless of the domain.

[0081] In addition, each cached digital component (and its corresponding selection criteria) is associated with one or more globally unique experimental identifiers (eid). control Associated. To enhance global uniqueness, MPC cluster 130 can calculate this identifier as related to... ∈ [ , Related eid control = AC( _ , _ _ _ Here, globally unique means that each experimental identifier (eid) is unique. control The control identifier is unique across different domains accessible by application 112. For example, the same control identifier can be used across different websites and domains accessed by users of browser 112 without overlapping with any other identifiers or users, regardless of the domain. Control identifiers and intervals [ , Related to this interval [ , ]and[ , Separate and non-overlapping, and if ∈ [ , If so, the application is in the control group. In other words, MPC cluster 130 determines whether... ≤ ( × + ) ≤ This will determine if the browser is in the control group.

[0082] As an example, the list of experimental identifiers attached to a specific digital component response sent from MPC cluster 130 to client device 110 can contain O(10^2) 4 There are 10 PEIDs, one for each ongoing experiment, where each experiment represents a different digital component. The number of attached experiment identifiers varies for each request for a digital component. In some implementations, the list of PEIDs can be very large, and to concisely represent the list of PEIDs and reduce the required bandwidth, the system can use probabilistic data structures such as Bruker filters or Bloom filters.

[0083] In step 2 of process 300, MPC cluster 130 constructs a Bloom filter with active applicable experiment identifiers. For example, using 11 hash functions, MPC cluster 130 can construct an optimal Bloom filter size of 2 bytes per PEID with a false positive rate (FPR) of 0.045%. Each Bloom filter has a bit array, and to construct the Bloom filter, elements are inserted into the Bloom filter one by one. For each element, a hash function is used to infer the bits in the bit array corresponding to that element. For example, if there are 11 hash functions, the 11 hash functions are used to set 11 bits in the bit array to the value 1 of the element. To prevent potentially compromised applications from extracting or forging confidential information from the PEID list, MPC cluster 130 may rely on cryptographic algorithms.

[0084] MPC cluster 130 uses a Bloom filter because it can accurately represent experimental identifiers activated with a small amount of data and allows communication with minimal bandwidth consumption. MPC cluster 130 can control the FPR of the Bloom filter so that its impact on the statistical results of the experiment is negligible. The construction of the Bloom filter is performed using secret shares to protect user privacy, allowing each server to calculate the position of the identifier within the Bloom filter without knowing whether the value is 0 or 1. The calculation can be done, for example, using the =+ operation, which can be performed within the secret shares. In some implementations, the Bloom filter is constructed as a counting Bloom filter, which allows construction to be done within secret shares and can later be converted to a regular Bloom filter. The construction of the Bloom filter is described in further detail below with respect to steps 2-3 of process 400.

[0085] In step 3 of process 300, MPC cluster 130 is capable of encrypting and / or signing a Bloom filter, which can be divided into multiple secret shares to prevent any party to which the secret shares are meaningful from accessing the Bloom filter parameters and reconstructing the complete Bloom filter. For example, MPC cluster 130 can use public-key encryption to encrypt the Bloom filter.

[0086] In step 4 of process 300, once MPC cluster 130 has determined whether application 112 belongs to the experimental group or the control group, MPC cluster 130 attaches the corresponding experimental identifier to the digital component response returned to application 112. For example, MPC cluster 130 can include a Bloom filter along with the digital component response provided to application 112. Additionally, MPC cluster 130 selects the winning digital component from the candidate set that meets all its conditions, as described below. Figure 5 As described in steps 504-506.

[0087] In step 5 of process 300, after the selection process has been completed, such as the above regarding... Figure 2 After the process selects the winning digital component from the eligible digital components indicated by the Bloom filter, application 112 renders the digital component associated with the winning selection criterion. For example, application 112 is capable of rendering video associated with the winning selection criterion.

[0088] In step 6 of process 300, the rendered digital component, associated with application 112, sends a report to DSP 150 via an aggregated measurement function, such as the aggregated measurement API native to application 112. For example, the aggregated measurement API can be integrated into application 112. In some implementations, the aggregated measurement API includes a server, which can be implemented as a secure MPC cluster. The report can contain a list of PEIDs associated with it. For example, the report can include a previously constructed Bloom filter and measurements associated with the rendering of the digital component, such as the amount of time the digital component is displayed, whether the user's attention is focused on the digital component, whether other applications are running, and whether the digital component is visible, as well as other measurements.

[0089] In step 7 of process 300, after accumulating a threshold number of reports, DSP 10 uploads the reports received in the secret share to the aggregated report server 180. For example, DSP 150 can periodically upload many encrypted reports in the secret share to the aggregated report server 180, making DSP 150 unable to access the plaintext of the reports. DSP 150 can detect whether the threshold number of reports has been accumulated and upload the reports in batches (i.e., secret shares) to the aggregated report server 180.

[0090] In step 8 of process 300, DSP 150 queries aggregated report server 180 to obtain metrics associated with each PEID. For example, DSP 150 can request various analyses, measurements, and other parameters from aggregated report server 180.

[0091] about Figure 4 Procedure 400 is an example procedure for publishing and implementing privacy-preserving cross-domain experimental partitioning and monitoring mechanisms to support incremental changes in the experiment. About Figure 4 The process 400 shows many steps that are related to the above. Figure 3 The steps are similar to those described in process 300.

[0092] In step 1 of process 400, application 112 sends a request for the digital component to MPC cluster 130, as described above regarding... Figure 3 The process described in step 1 of 300 is similar.

[0093] In some cases, content providers, designers, researchers, and other entities wish to quantify the incremental effects attributable to specific features or changes. For example, users are 50% more likely to respond "yes" to watch a virtual performance of a specific orchestra after seeing a digital component, compared to not seeing one. To support this incremental research, the system refers to the above text below regarding... Figure 2 andFigure 3 The baseline process described above introduces a counterfactual selection process.

[0094] Similar to baseline experimental design, the selection criteria for caches are... , , , , , Relatedly, MPC cluster 130 from peid, and Calculate the modulus .

[0095] Here, if ∈ [ , If the value is specified, the user is in the exposure group. In other words, the exposure group is exposed to a digital component that is associated with the cache selection criteria. The exposure group is associated with the experiment identifier eid. exposed Related. To ensure uniqueness, eid exposed Defined as eid exposed = HMAC(owner_domain, domain_specific_exposed_id).

[0096] if ∈ [ , If the user is in a reserved group, they are not exposed to the digital components associated with the cache selection criteria. In other words, processing is done from the group "Reserved". The reserved group is associated with the experimental identifier eid. holdback Related. To ensure uniqueness, eid holdback Defined as eid holdback = HMAC(owner_domain, domain_specific_holdback_id).

[0097] Here, if m is not in either range, then application 112 is neither in the exposed group nor in the reserved group.

[0098] In step 2 of process 400, MPC cluster 130 performs a selection process that applies all selection rules to select at most one of the cached digital components as the winner of the selection process. If the winning digital component is associated with the original experiment eligibility rules, MPC cluster 130 will eid. exposed And associated metadata is inserted into the response to the request for digital components.

[0099] In step 3 of process 400, MPC cluster 130 then executes a counterfactual selection process for all rules of the application selection process, except for removing the experiment eligibility rule from... ∈ [ , Updated to ∈ [ , In addition, MPC cluster 130 can then determine the phantom candidate selection criteria based on the updated eligibility rules, thereby selecting at most one of the cached number components and representing the winner of a counterfactual selection process that is exactly the same as the actual selection process, but for the sake of updating the eligibility rules. If the winning number component is associated with the modified experimental eligibility rules, or in other words, if the winning number component is only eligible to be selected based on the updated eligibility rules, then MPC cluster 130 will eid holdback And associated metadata is inserted into the response to the request for digital components.

[0100] The counterfactual selection process can be executed in parallel with the privacy-preserving selection process. A phantom candidate number component is eligible to win the selection process if its selection criteria meet the original eligibility rules, but those rules have been updated for the counterfactual selection process and therefore have not yet been applied to the selection criteria. The phantom candidate number component can be selected as the winning number component, regardless of the actual candidate selection criteria, which depend on the value of m.

[0101] MPC cluster 130 can perform these two selection processes in parallel, for example, in batch mode, without increasing the number of RPCs between computing systems MPC1 and MPC2. The results of the counterfactual selection process allow MPC cluster 130 to identify digital components that might not qualify under the original eligibility rules.

[0102] The counterfactual selection process can have the same form as the actual selection process, and in order for a digital component to be eligible for use in the selection process, the digital component must meet certain distribution requirements. The counterfactual selection process measures what happens when the eligibility rules are updated by applying all distribution rules other than the original eligibility rules applied in the actual selection process.

[0103] The counterfactual selection process allows a system to determine whether a particular numerical component is eligible for distribution if it is part of a different group in an experiment.

[0104] As described above, example processes 300 and 400 determine the eligibility of digital components for the experimental group, select the winning digital component, and include the experimental group identifier in the response to the digital component request without the many details of the MPC(s) process(s) executed by the MPC cluster 130. The following sections provide additional implementation details that can be used on example processes 300 and 400.

[0105] The computers MPC1 and MPC2 of MPC cluster 130 are capable of executing a secure MPC process using secret shares to determine various parameters for assigning users to experimental groups. For each selection criterion and each digital component request, MPC cluster 130 executes the following function c = peid ( 2 ); = ; = ;and = ( × + ) Each of these. To protect user privacy, a secret share is used to transmit the PEID from the client device 110, and each server in the MPC cluster 130 separately receives a secret share of the PEID. Other input parameters, such as , , , , , This is not sensitive information and therefore can be shared in plaintext between MPC1 and MPC2. In some implementations, other input parameters are also processed in the secret share.

[0106] In the digital component request, MPC1 and MPC2 are in Z. q Z receives the additive secret share of PEID. q Let represent the set of arbitrarily large integers in [0, q-1]. In other words, MPC1 and MPC2 receive [peid1] and [peid2] respectively. Here, q is O(2^35) integers. 128 In the equation ), p is in O(10,000). In other words, q >> p. For example, q can be a large prime number where Z q The additive secret share in q has a value randomly selected from [0, q-1].

[0107] Each of the MPC servers, MPC1 and MPC2, executes a series of functions separately and locally using a corresponding secret share of the data it wants to keep private. For example, MPC1 locally computes [c1] = [peid1].mod (p 2 [x1] = [c1] mod p; [y1] = [c1] div p; and [m1] = ( × [ 1] + [ 1]) p.

[0108] Similarly, MPC2 computes locally [c2] = [peid2] mod (p 2 [x2] = [c2] mod p; [y1] = [c2] div p; and [m2] = ( × [ 2] + [ 2]) p. As mentioned above, the parentheses specify the secret share of the parameter.

[0109] Due to the structure of the secure MPC cluster 130, [x1] and [x2] are Z. p The addition secret shares of x in Z, [y1] and [y2] are Z p The addition secret share of y in Z, and [m1] and [m2] are Z. p The secret share of addition of m in the equation.

[0110] For each cached digital component and its selection criterion sc, MPC1 and MPC2 can then collaboratively perform a secret share comparison algorithm to determine whether a particular experimental identifier is in the exposed group or the reserved group. For example, MPC1 and MPC2 can collaboratively perform a secret share comparison algorithm to evaluate is_in_exposed_group. sc = m r AND m s, and is_in_holdback_group sc = m AND m The comparison algorithm can be performed using the MPC protocol, and the result can be, for example, an additive secret share in Z2, which represents a secret share with an integer value in [0, 2⁻¹] (i.e., a value of 0 or 1). For example, an additive secret share in Z2 means that each additive secret share has a value of 0 and 1. Two such secret shares can represent a secret message of Boolean type. If the two additive secret shares are the same, the corresponding secret message is false. Otherwise, the corresponding secret message is true.

[0111] In some implementations, MPC1 calculates

[0112] [is_in_exposed_group r,1 ] = 1 – ([m1] < r)

[0113] [is_in_exposed_group s,1 ] = 1 – ([m1] > s)

[0114] [is_in_holdback_group r,1 ] = 1 – ([m1] < )

[0115] [is_in_holdback_group s,1 ] = 1 – ([m1] > )

[0116] Similarly, MPC2 calculation

[0117] [is_in_exposed_group r,2 ] = 1 – ([m2] < r)

[0118] [is_in_exposed_group s,2 ] = 1 – ([m2] > s)

[0119] [is_in_holdback_group r,2 ] = 1 – ([m2] < )

[0120] [is_in_holdback_group s,2 ] = 1 – ([m2] > )

[0121] For each cached digital component selected for eligibility evaluation, there are x affirmative conditions, which are represented as a set of secret messages in Z2. For example, these affirmative conditions can be represented as {[condition...}positive,1 ], …[condition positive,x These conditions can be related to user group membership, frequency control, content distribution rules, and pacing rules. Each of these conditions must be true (e.g., represented by the value 1) for the cached digital component to be eligible for the selection process. In other words, all x affirmative conditions must be satisfied for the cached digital component to become a candidate for selection by MPC cluster 130.

[0122] There exist y negation conditions, which are represented as a set of secret messages in Z2. For example, these negation conditions can be represented as {[condition...} negative,1 ],…… [condition negative,y Each of the conditions in condition y must be false (e.g., represented by the value 0) so that the cached digital components become candidates for selection by the MPC cluster 130.

[0123] Examples of affirmative conditions include membership in a specific user group, value and pacing constraints, and privacy protection rules. For example, the distribution of content can be controlled by input from digital component provider 160 via value and pacing control techniques. The digital component provider can specify a total threshold for selection values ​​over a period of time to implement measures to ensure that resources including the movement of digital components are not prematurely exhausted during the activity. For example, digital component provider 160 can specify that the sum of selection values ​​for a particular digital component from digital component provider 160, or for a specific digital component over a period of one week, is limited to a maximum value of 8,000 units. Once the maximum sum of selection values ​​is reached, the specific digital component to which the maximum value is imposed will no longer be eligible for distribution to the selection process of client devices 110 until the end of that period (e.g., one week).

[0124] The digital component provider can also specify a speed or pace at which a threshold for the selection value of a particular digital component from digital component provider 160 is being reached. For example, digital component provider 160 can specify a maximum threshold of 6000 units that must be reached over a one-month period at a pace of less than 250 units per day. Once the daily total selection value is reached, the digital component provided by digital component provider 160, to which the pace is adjusted, will no longer be eligible for use in the selection process distributed to client devices 110 until the end of that period (e.g., a day).

[0125] In addition to enforcing digital component eligibility according to content distribution rules, the privacy-preserving selection process also protects user privacy and improves user experience by protecting user k-anonymity, ensuring that any selected component is offered to at least k other users. By ensuring that a particular digital component is offered to a threshold number of other users, the system prevents a form of user targeting called microtargeting, which narrowly targets users to a degree that some users might find uncomfortable. For example, digital components targeted and distributed to only a few users might feel overly personalized, and users might feel isolated or uncomfortable with the specificity of the digital component. The privacy-preserving selection process ensures that digital components are distributed to a sufficiently broad audience so that users are satisfied with the digital components they receive and the size of the target audience for those components.

[0126] Examples of negative conditions include user preferences regarding reducing or eliminating content or digital components being presented to them. For instance, users often prefer to have the same third-party content displayed discontinuously when navigating across multiple different websites, so continuously distributing the same third-party content to the same user across multiple different websites within a short period of time would result in a waste of computing resources used to distribute content that users don't want to see and may ignore. One way to limit the number of times a particular user is exposed to the same third-party content is to use frequency control techniques that prevent the same user from being exposed to the same third-party content more than a specified number of times within a short period of time, such as any consecutive 5 or 10 minutes, or another appropriate time period.

[0127] The system allows further user input or prevents digital components from being presented to them for a period of time by receiving information from the user to mute the digital components. For example, a user can choose to mute a specific digital component that has been presented to them for a period of five days. In some implementations, the time period can be specified by the client device 110, MPC cluster 130, DSP 150, digital content provider 160, or SSP 170, based on the type of content item and / or the type of feedback provided by the user, as well as other factors.

[0128] In this specific example, MPC1 stores {[condition] for a given digital component. positive,1,1 ], …[condition positive,x,1 ]} and {[condition negative,1,1 ], … [condition negative,y,1 These are secret shares of Boolean values ​​representing whether each condition is met. For example, if the first affirmative condition is met... positive,1 ), then [condition positive,1,1 ] will be the first secret share of the secret Boolean value of one.

[0129] Similarly, MPC2 saves {[condition positive,1,2 ], … [condition positive,x,2 ]} and {[condition negative,1,2 ], … [condition negative,y,2 These are also secret shares of the Boolean value indicating whether each condition is met.

[0130] For cached digital components, in order to use algorithms such as the GMW (Goldreich-Micali-Wigderson) protocol or scrambled circuits, and inadvertent transmission extension (OTe) to compute the candidate parameter is_dc_a_candidate, which indicates whether the cached digital component is a candidate for selection, MPC1 appends a secret share of the set of affirmative conditions to [is_in_exposed_group]. r,1 ] and [is_in_exposed_group s,1 The set of combinations of affirmative conditions is {[condition]}. positive,1,1 ],……[condition positive,x,1 ], [is_in_exposed_group r,1 ], [is_in_exposed_group s,1 ]}.

[0131] Similarly, MPC2 appends the secret share of the affirmative condition set with [is_in_exposed_group] r,2 ] and [is_in_exposed_group s,2 The set of combinations of affirmative conditions is {[condition]}. positive,1,2 ],……[condition positive,x,2 ], [is_in_exposed_group r,2 ], [is_in_exposed_group s,2 ]}.

[0132] The GMW protocol or scrambling circuit is used to determine whether all positive conditions of a digital component are true and whether at least one of the negative conditions of the digital component is false. If so, the digital component is a candidate for the selection process. In other implementations, Boolean operations can be performed on the secret shares by MPC1 and MPC2 to determine whether all positive conditions are true and whether at least one of the negative conditions is false. However, the GMW protocol is more efficient and involves less round-trip computation between MPC1 and MPC2. After completing the GMW protocol using OTe, MPC1 stores a first secret share [is_dc_a_candidate1] for the candidate parameters of the digital component, and MPC2 stores a second secret share [is_dc_a_candidate2] for the candidate parameters of the same digital component. MPC1 and MPC2 are able to determine the secret share of the candidate parameters for each cached digital component and any JIT digital component that can be selected.

[0133] This process can use two comparison tests for each determination of whether to assign user / application 112 to the exposure group, and therefore based on whether... ∈ [ , This is used to activate each experimental identifier as part of the exposure group. The secret share comparison process can be computationally expensive, requiring multiple round trips between servers MPC1 and MPC2 within the MPC cluster 130.

[0134] In some implementations, the protocol used to determine whether to activate a specific experimental identifier can be improved or optimized. In one example, by limiting... Given , , ∈ Z p and It can be proven that if and only if hour, ∈ [ , In this simplification of equations that impose constraints that must be satisfied, The evaluation requires only one comparison test instead of the two required as described above. This reduces the amount of bandwidth consumed between NPC1 and NPC2, and lowers the latency and computational cost in determining whether a user belongs to the experimental group. Since the process can be performed in real-time between requesting a digital component and providing it in response to that request, reducing latency is crucial for minimizing errors at client device 110 and improving the user experience.

[0135] MPC1 and MPC2 can use this improvement to perform secret share comparison test algorithms. Upon completion of the comparison, MPC1 saves... Similarly, MPC2 saves In other words, for each digital component, each MPC server stores a secret share of the exposure parameter is_in_exposed_group, which indicates whether the user is in the digital component's exposure group.

[0136] Therefore, the set of affirmative conditions stored in MPC1 is {[condition positive,1,1 ], ……[condition positive,x,1 Therefore, the set of affirmative conditions stored in MPC2 is {[condition], [is_in_exposed_group1]}. positive,1,2 ], ……[condition positive,x,2 ], [is_in_exposed_group2]}.

[0137] Using these conditions, MPC cluster 130 then continues to execute the GMW algorithm using OTe.

[0138] This improvement halves the number of comparison tests required and halves the number of rows in the truth table built for the GMW protocol, without increasing the number of round-trip operations required between servers in the MPC cluster 130.

[0139] For each digital component request, each digital component (and its corresponding selection criteria) can be associated with the following experimental identifier, and the activated eid if is_in_exposed_group is true. experiment Or, if is_in_holdback_group is true, then the activated eid control These parameters can vary for each digital component based on whether the user who generated the request should be assigned to an experimental group of digital components.

[0140] Returns to application 112 and then returns a variable number of reports to aggregation report server 180. ,like Figure 3 and Figure 4 As described in step 4, the system can utilize Bloom filters to reduce bandwidth and storage costs.

[0141] For example, if a request for a digital component will be activated The expected quantity is For the desired FPR, the MPC cluster 130 can calculate the hash function. The optimal or target number, and for each Bloom filterp Number of bits .

[0142] Next, MPC1 and MPC2 are able to create a size of × The counting Bloom filter R, where the value of each element is within [0, number_of_cached_digital_components]. Here Furthermore, MPC1 and MPC2 can create arrays R1 and R2 respectively. Both arrays are of size [size missing]. × ,exist p Includes × Each array has an additive secret share. Each array is initialized to 0.

[0143] To minimize bandwidth consumption, MPC1 and MPC2 perform the following cryptographic protocol to construct a counting Bloom filter R, and then convert R into a normal Bloom filter, where each element is an additive secret share in Z2.

[0144] For EID experiment For each hash function , Keep it true. For each hash function. MPC1 will array Increment For example, MPC1 performs an operation. .

[0145] Similarly, MPC2 will array Increment For example, MPC2 performs operations. .

[0146] In order to eid control Inserted into the same Bloom filter, for each hash function MPC1 will array Increment For example, MPC1 performs an operation. .

[0147] Similarly, MPC1 will array Increment For example, MPC1 performs an operation. .

[0148] In addition, MPC2 pre-selects a pseudo-random function (PRF), where, Additionally, MPC2 selects random numbers for the digital component response. Here, R2 represents the result Bloom filter of MPC2 in Z2: In some implementations, these steps are performed by MPC1 instead of MPC2.

[0149] Then, MPC2 constructs a scrambled circuit of an AND gate to construct the function. This enables MPC1 to compute the resulting array of its secret shares. , making and It is a secret Boolean message The two addition secret shares in Z2. This scrambled circuit calculation requires one round trip between MPC1 and MPC2.

[0150] In step 4, MPC cluster 130 returns a digital component response to application 112. Depending on the value of m determined above, the response includes information associated with the winning digital component selected during the actual selection process. Or associated with the winning digital component in the counterfactual selection process. Or, it may exclude either of the two experimental identifiers.

[0151] MPC1 and MPC2 will then The random number and the original number are returned to application 112 as part of the digital component response. If MPC1 receives the original digital component request from application 112, and MPC2 needs to return its result to the browser via MPC1, then MPC2 can digitally sign and encrypt its result and send it to the browser via MPC1. The result is returned to the browser to prevent MPC1 from eavesdropping on or forging the result of MPC2. In some implementations, MPC1 may also encrypt and then digitally sign its result.

[0152] In step 5 of process 400, application 112 accumulates the PEID and associated metadata for the exposure group or retention group in local storage. For example, application 112 accumulates experiment identifiers in its cache. The application can be associated with the exposure group of some experiments, or with the retention group of other experiments.

[0153] In step 6 of process 400, the rendered digital component, associated with application 112, sends a report to DSP 150 via aggregated measurement functions, such as the aggregated measurement API native to application 112. This report can contain a list of related parameters. For example, the report can include previously constructed Bloom filters and measurements associated with the rendering of the digital component, such as the amount of time the digital component is displayed, whether the user's attention is focused on the digital component, whether other applications are running, and whether the digital component is visible, as well as other measurements.

[0154] In step 7 of process 400, after accumulating a threshold number of reports, DSP 150 uploads the received reports to the aggregated report server 180 in secret shares. For example, DSP 150 can periodically upload a number of encrypted reports from the secret shares to the aggregated report server 180, preventing DSP 150 from accessing the plaintext of the reports. DSP 150 can detect whether the threshold number of reports has been accumulated and upload the reports to the aggregated report server 180 in batches (i.e., secret shares).

[0155] In step 8 of process 400, DSP 150 queries aggregated report server 180 to obtain metrics associated with each PEID. For example, DSP 150 can request various analyses, measurements, and other parameters from aggregated report server 180.

[0156] Figure 5 This is a flowchart illustrating an example process 500 for issuing and implementing a privacy-preserving experimental monitoring mechanism. The operation of process 500 can be implemented, for example, by application 112, MPC cluster 130, DSP 150, and / or aggregated report server 180. The operation of process 500 can also be implemented as instructions stored on one or more computer-readable media that may be non-transitory, and execution of the instructions by one or more data processing devices can cause one or more data processing devices to perform the operation of process 500.

[0157] Process 500 begins with a first server of a secure multi-party computation (MPC) system receiving a request for digital content from an application on a user device, wherein the request includes a first secret share (502) identifying an application instance instance associated with the device or an application instance identifier of the device itself. For example, MPC cluster 130 is capable of receiving requests for digital content, as referenced above. Figure 2 Step A and Figure 3-4 As described in step 1. The application instance identifier identifies an instance of an application running on a particular device (e.g., a session of the example news application), or identifies the device itself.

[0158] Process 500, in response to the received request and continued by the first server of the secure MPC system, collaborates with the second server of the secure MPC system to perform a privacy-preserving selection process to select a first winning digital component from a set of digital components, wherein each digital component has a corresponding (i) unique experiment identifier and (ii) unique control identifier (504). The set of digital components may include cached digital components having cached lookup keys that match the lookup key included with the request. That is, the set of digital components may be cached digital components whose context signals defined by the requested lookup key match the context signals defined by the cached lookup keys of the digital components. The first MPC system may compare the requested lookup key with the cached lookup keys of the cached digital components.

[0159] The privacy-preserving selection process includes, for each digital component in the set of digital components, collaborating with a second server and based on a set of eligibility rules, determining a first secret share of a candidate parameter that indicates whether a digital component is an eligible candidate for selection in the privacy-preserving selection process (506).

[0160] For example, the selection process includes determining whether a digital component is a candidate for selection based on a set of eligibility rules. As a result of this candidate determination process, each of MPC1 and MPC2 stores a share of the candidate parameter `is_dc_a_candidate` for each digital component in the set of digital components, indicating whether that digital component is a candidate for selection for each component. In this way, the computers in MPC cluster 130 do not store the actual eligibility or candidate list in plaintext. In some implementations, each of the two servers is able to cache candidates in plaintext (not privacy-sensitive).

[0161] In some implementations, portions of the cached candidates may contain sensitive information, and therefore neither server can store the candidates in plaintext. In this case, it is possible to store the sensitive information of the cached digital components in a secret share.

[0162] Related to the above text Figure 2 Similar to process 200, for each condition of a digital component (e.g., a digital component with at least one condition), each computing system MPC1 and MPC2 can store a corresponding secret share of the parameters for each condition of the digital component. In this way, as long as at least one MPC computing system is honest, neither computing system NPC1 nor NPC2 knows the value of the parameters in plaintext. Each digital component can be conditioned with zero or more conditions. For a given digital component selection process, some digital components can have different conditions and / or a different number of conditions than other digital components.

[0163] While some example conditions have been provided above, other conditions can also be used. Generally, computing systems MPC1 and MPC2 can use a secure MPC process to calculate the secret share of the condition parameters. The criteria and techniques used to determine the condition parameters can vary. In some implementations, the secret share of the condition parameters can be received from another computing system, for example, causing computing systems MPC1 and MPC2 not to calculate the condition parameters.

[0164] Computing systems MPC1 and MPC2 can use secret shares of conditional parameters to determine whether a conditional number component is a candidate for the number component selection process. Computing systems MPC1 and MPC2 can use secret shares of (multiple) conditional parameters for each conditional number component... i Calculate candidate parameters is_dc_a_candidate i The secret share. Generally speaking, if a conditional digital component is conditional on each of the conditions described above, then the digital component... i The candidate parameters can be calculated and determined in the following ways: for example, whether all affirmative conditions are true, for example, one in some embodiments or non-zero in others, and whether at least one negative condition is false, for example, zero, as mentioned above. Figure 4 The process described in steps 2 and 3 of 400.

[0165] Because the value of each condition parameter is contained in a secret share, computing systems MPC1 and MPC2 are able to collaborate using round-trip remote procedure calls (RPCs) during a secure MPC process to determine the digital component using the secret share of the condition parameters. i The corresponding secret share of the candidate parameters [ is_dc_a_candidate i,1 ]and[ is_dc_a_candidate i,2 The digital component can be determined using any appropriate secret share algorithm used to determine a logical AND operation. i Secret share of candidate parameters [ is_dc_a_candidate i,1 ]and[ is_dc_a_candidate i,2 For example, as mentioned above, the GMW protocol or scrambled circuit can be used. Computational systems MPC1 and MPC2 are able to determine the secret share of candidate parameters for those conditions using only the secret share of the conditional parameters. At the end of this secure MPC process, computational system MPC1 stores the first secret share of the candidate parameters for the i-th conditional digital component. is_dc_a_candidate i,1 The computational system MPC2 stores the second secret share of the candidate parameters for the i-th conditional digital component. is_dc_a_candidatei,2 ].

[0166] In some implementations, computing systems MPC1 and MPC2 use a scrambled circuit protocol to evaluate the condition of each digital component. In this example, one of computing systems MPC1 or MPC2 is capable of constructing a scrambled circuit. For this example, assume that computing system MPC1 constructs the scrambled circuit. Computing system MPC1 knows its own secret share and also knows that there is only one possible bit pattern that computing system MPC2's secret share must preserve in order for the candidate parameters of the digital component to become true, or one.

[0167] Computing system MPC1 determines the order of digital components based on selection criteria, which can be plaintext values ​​associated with each digital component. Similarly, computing system MPC2 determines the order of digital components based on selection criteria. Selection criteria can also be referred to as selection values. The two orders should be identical because the inputs to the sorting process are the same on both computing systems MPC1 and MPC2. Each computing system MPC1 and MPC2 can determine the order of digital components. Each order can include candidate digital components previously evaluated for candidate eligibility and other digital components. The order can be from the digital component with the highest selection value to the digital component with the lowest selection value. In some implementations, the selection value used for the order can be a value to be provided to the resource publisher 140, and the selected digital components will be displayed through the resource publisher, for example, after any sharing with DSP 150 and / or SS P170, plus any applicable boosts.

[0168] Because the selection values ​​are in plaintext, computation systems MPC1 and MPC2 do not need to perform any round-trip computations to determine the order of the digital components. Instead, each computation system MPC1 and MPC2 can independently order the selection values. If the selection values ​​are stored as secret shares at each computation system MPC1 and MPC2, and each computation system MPC1 and MPC2 has a corresponding secret share for each selection value, then computation systems MPC1 and MPC2 can perform a secure MPC process using round-trip computations to order the selection values. If there is a tie among two or more selection values, computation systems MPC1 and MPC2 can deterministically use additional metadata to break the tie for the digital components corresponding to those selection values.

[0169] The computing systems MPC1 and MPC2 determine a secret share of the cumulative value for each candidate digit component. Conceptually, the cumulative value for a given digit component represents the total number of candidate digit components from the top of the order to the selection value for that given digit component, excluding the given digit component even if it is a candidate. In other words, the cumulative value represents a number of candidate digit components that are more qualified for selection than the given digit component. This concept is illustrated in Table 1 below.

[0170]

[0171] Table 1

[0172] In some implementations, the cumulative value for a given digit component represents the total number of candidate digit components from the top of the sequence to the given digit component, including the given digit component if it is a candidate. In this example, the fourth column will indicate whether the cumulative value is equal to one rather than zero. For the sake of brevity, the remaining discussion will refer to the first example in which the cumulative value for a given digit component represents the total number of candidate digit components from the top of the sequence to the given digit component, excluding the given digit component even if it is a candidate.

[0173] Conceptually, in Table 1, for candidate parameters that are equal to 1 is_dc_a_candidate For each digital component, the cumulative value ( acc The cumulative value increases as it progresses from the top of the sequence to the bottom. As described below, the cumulative value... acc The calculation is performed within the secret share. For example, due to the candidate parameters for the highest selection value. is_dc_a_candidate The cumulative value is equal to zero, therefore it is the cumulative value for the digital component with the highest selected value. acc It is zero. This is due to the candidate parameters for the second highest digital component. is_dc_a_candidate The selected value is equal to one, but none of the choices for the second highest number component has a candidate parameter equal to one. is_dc_a_candidate Therefore, the cumulative value for the second highest digital component acc Also zero. Moving down sequentially, based on the candidate parameter with the second highest selection value of one. is_dc_a_candidate Candidate parameters for the third highest choice digital component is_dc_a_candidate cumulative value acc Increment to a value of one. This is due to the candidate parameters for the third highest-valued component. is_dc_a_candidate The cumulative value is zero, therefore the cumulative value for the fourth digital component is zero. acc It is not incrementing and has the same value of zero as the third highest number component.

[0174] Using Table 1, computational systems MPC1 and MPC2 will select candidate parameters corresponding to their overall values. is_dc_a_candidate It has a value of 1 and a cumulative valueacc A digital component with a selection value of zero is used for distribution to client device 110, as indicated in the fourth column of Table 1. This represents the candidate parameter corresponding to it. is_dc_a_candidate The numerical component with the highest sorting selection value of one. Due to candidate parameters. is_dc_a_candidate In the secret share of computing systems MPC1 and MPC2, to maintain user privacy and ensure that user data is not leaked, computing systems MPC1 and MPC2 determine the cumulative value of each digital component. acc The secret share, and uses round-trip calculations to determine which digital component's cumulative value. acc Equal to zero, and candidate parameters is_dc_a_candidate It equals one.

[0175] The computing systems MPC1 and MPC2 are capable of independently determining the cumulative value of each digital component. acc The secret share is determined without requiring any round-trip computation depending on the secret share algorithm in some implementations. For example, by traversing all digital components in descending order and considering the candidate parameters of the digital components along the way. is_dc_a_candidate Summation, the calculation system MPC1 can perform summation for each digital component. i Determine the cumulative value acc The first share acc i,1 As described in Table 1 above. Similarly, by traversing all digital components in descending order and considering the candidate parameters of the digital components along the way. is_dc_a_candidate Summation, the calculation system MPC1 can perform summation for each digital component. i Determine the cumulative value acc The second share acc i,2 ].

[0176] The computing systems MPC1 and MPC2 determine a secret share of the result for each digital component, indicating whether the cumulative value has a specified value. The specified value can be zero, as shown in columns 3 and 4 of Table 1. As mentioned above, a cumulative value of zero and an overall candidate parameter... is_dc_a_candidate A number component with a value of 1 is the number component with the highest selection value among the candidate number components.

[0177] The computing systems MPC1 and MPC2 are capable of performing multi-round computations, such as multiple RPCs, as part of a secure MPC process to compute on each digital component. i Equal operation in terms of secret share acc i == 0. This equality operation is used to determine the digital component. i cumulative value acc iDoes it have a value of zero? At the end of this process, the computing system MPC1 calculates for each digital component. i With results acc i == 0 is a secret share, and the computation system MPC2 has a result for each digital component. acc i Another secret share of == 0.

[0178] The computing systems MPC1 and MPC2 are designed for each digital component. i Determine the parameters of the winner is_dc_the_winner i The secret share. The computing systems MPC1 and MPC2 are capable of operating on a per-digital basis. i cumulative value acc i == 0 secret share and for each digital component i Candidate parameters is_dc_a_candidate i The secret share determines the winner's parameters. is_dc_the_winner i For each digital component i Winner parameters is_dc_the_winner i It can be a Boolean value, which indicates the digital component. i Whether it is the winner of the selection process, for example, whether it selects a digital component in response to a digital component request. i Distributed to client devices 110.

[0179] In some implementations, computing systems MPC1 and MPC2 are capable of executing a secret share multiplication protocol to compute winner parameters regarding the secret share for each selected value. This can include an RPC between computing systems MPC1 and MPC2 to multiply the two secret shares. At the end of this MPC process, computing system MPC1 has a value represented as [ is_dc_the_winner i,1 ] = [ is_dc_a_candidate i,1 ] x ([ acc i,1 The result of ] == 1) is_dc_the_winner i A secret share. Similarly, the computing system MPC2 has a share represented as [ is_dc_the_ winner i,2 ] = [ is_dc_a_candidate sv, 2 ] x ([ acc i, 2 The result of ] == 0) is_dc_the_winneri Another secret share. Note that for all number components, at most one number component has a winner parameter equal to one. is_dc_the_winner i This digital component is the one selected for distribution to client device 110. All other digital components will be equal to zero. In some implementations, computing systems MPC1 and MPC2 are capable of executing a secret share logic operation protocol (e.g., GMW or scrambled circuitry) to compute a winner parameter in terms of secret share for each selected value. .

[0180] Process 500 continues, wherein privacy protection further includes, for each digital component in the digital component set, collaborating with a second server and based on a first secret share of the application identifier and a corresponding first secret share of the experimental value, determining a secret share of the experimental parameters indicating whether the application is in the experimental group used for the application (508). For example, MPC1 of MPC cluster 130 can collaborate with MPC2 and, based on the first secret share of the application identifier and as described above... Figure 2 to Figure 4 The described secret share of the experimental value, determining whether the application 112 is part of the secret share of the experimental parameters used for digital components, such as eid. holdback or eid exposed .

[0181] Process 500 continues, generating a first secret share (510) representing the selection result of the first winning digital component based on the first secret share of the candidate parameters for each digital component and the selection value for each digital component. For example, MPC cluster 130 is capable of generating a first secret share representing the selection result of the first winning digital component based on the first secret share of the candidate parameters for each digital component and the selection criteria for each digital component. The selection result can identify or include the first winning digital component.

[0182] Process 500 continues, generating a response to the digital content request, which includes a first secret share of the selection result, and for each of one or more digital components including the first winning digital component, generating data indicating whether the application is in the experimental group or the control group of the digital component (512). For example, MPC cluster 130 generates a response to the digital component request, as described above regarding... Figure 2-4 And as described in step 502 of process 500. The response includes data such as eid. holdback or eid exposed This indicates whether application 112 is in the experimental group for digital components or in the control group for digital components.

[0183] Process 500 continues, with the first server of the secure MPC system transmitting a response to the request for digital content to the application on the user device (514). For example, MPC1 in MPC cluster 130 transmits a response to the request for digital components to application 112. In some implementations, MPC1 is able to transmit its own first secret share and second, encrypted secret share as a response to the request for digital components from MPC2 to application 112.

[0184] As described above, the MPC cluster 130 can use scrambled circuitry for the secure MPC process to determine whether a digital component is a candidate for the selection process, for example, to determine a secret share of the candidate parameters for the digital component. Since each digital component can have many conditions, for example, more than 50 in some cases, there will be many gates in the scrambled circuitry for each digital component. Reducing the number of gates will reduce the bandwidth cost of executing the secure MPC protocol and significantly reduce latency when selecting digital components.

[0185] There is a large amount of redundant computation and logical AND operations between the same pair of inputs or intermediate Boolean variables, for example, for the same number components or for groups of number components. A greedy algorithm can be used to avoid redundant computation of subexpressions, and thus reduce the number of AND operations.

[0186] In the above process 500, assume candidate parameters is_dc_a_candidate In Z P In this context, the value is either zero or one. MPC cluster 130 iterates through all cached digital components from highest to lowest, starting with the digital component with the highest selected value (e.g., the part attributed to the publisher). For each cached digital component i, MPC cluster 130 calculates a cumulative value. i Then calculate Z P In acc i == 0. The result is converted back to Z2. From Z2 to Z p The conversion and return operations increase protocol complexity. Z using garbled circuitry... p Equality test in acc i == 0 can involve log2(p) gates, which is a significant performance and bandwidth cost.

[0187] MPC cluster 130 is able to maintain the candidate parameter is_dc_a_candidate in Z2. i Furthermore, in Z2, Boolean circuits are used to calculate the cumulative value at a much lower performance and bandwidth cost. acc i ,Then acc i == 0. Therefore, the accumulated value will be... acc i The value is limited to two possible values: zero if no candidate digit component has a higher selection value than the current selection value of the current digit component i, or one if at least one candidate digit component has a higher selection value than the current selection value of the current digit component. Figure 5 In process 500, the cumulative value can be greater than one because the cumulative value increments for each candidate number component.

[0188] With the cumulative value acc With the updated semantics, MPC cluster 130 is able to compute the winner parameters. is_dc_the_ winner i The cumulative value is calculated by iterating through all cached numeric components from highest to lowest selected value, as described below. The cumulative value `acc` for the numeric component with the highest selected value is always zero. No calculation is required.

[0189] Cumulative value for the digital component with the second highest selection value acc The candidate parameter is always equal to the numeric component with the highest selected value. is_dc_a_candidate The value of is not calculated.

[0190] Assume that MPC cluster 130 has already calculated the cumulative value for the digital component with the highest selection value of the nth rank. acc The cumulative value for the digital component with the (n+1)th highest selection value. acc Can be calculated as acc n+1 = acc n OR is_dc_ a_candidate n The MPC cluster 130 can further calculate the winner parameters. is_dc_a_winner n+1 = is_dc_a_ candidate n+1 AND NOT acc n+1 .

[0191] The MPC cluster 130 can have two gates, where each cached digital component has four lines, for example, two input lines ( acc n and is_dc_a_candidate n ) and two output lines ( acc n+1 and is_dc_a_winner n+1With these enhancements, each cached digital component of the MPC cluster 130 requires only two gates instead of dozens. This cryptographic process can be performed as part of one of the steps described above in process 500, such as step 506. For example, cryptographic evaluation can be performed as part of the candidate eligibility determination described above with respect to step 506, where all affirmative conditions must be true and at least one negative condition must be false. In some implementations, conditions such as membership in a particular experimental group must be true.

[0192] Figure 6 This is a block diagram of an example computer system 600 capable of performing the operations described above. System 600 includes a processor 610, memory 620, storage device 630, and input / output device 640. Each of components 610, 620, 630, and 640 can be interconnected, for example, using a system bus 650. Processor 610 is capable of processing instructions for execution within system 600. In some embodiments, processor 610 is a single-threaded processor. In another embodiment, processor 610 is a multi-threaded processor. Processor 610 is capable of processing instructions stored in memory 620 or storage device 630.

[0193] Memory 620 stores information within system 600. In one embodiment, memory 620 is a computer-readable medium. In some embodiments, memory 620 is a volatile memory cell. In another embodiment, memory 620 is a non-volatile memory cell.

[0194] Storage device 630 provides mass storage for system 600. In some embodiments, storage device 630 is a computer-readable medium. In various embodiments, storage device 630 may include, for example, a hard disk drive, an optical disk drive, a storage device shared by multiple computing devices (e.g., cloud storage devices) over a network, or some other mass storage device.

[0195] Input / output device 640 provides input / output operations for system 600. In some embodiments, input / output device 640 may include one or more network interface devices, such as Ethernet cards, serial communication devices, such as RS-232 ports, and / or wireless interface devices, such as 802.11 cards. In another embodiment, input / output device may include a driver configured to receive input data and send output data to external device 660, such as a keyboard, printer, and display device. However, other embodiments, such as mobile computing devices, mobile communication devices, set-top box television client devices, etc., may also be used.

[0196] Although already Figure 6An example processing system is described herein, but the implementation of the subject matter and functional operations described herein can be implemented in other types of digital electronic circuits, or in computer software, firmware, or hardware, including the structures disclosed herein and their structural equivalents, or in a combination of one or more of them.

[0197] The embodiments of the subject matter and operation described in this specification can be implemented in digital electronic circuits, or in computer software, firmware, or hardware, including the structures disclosed in this specification and their structural equivalents, or in a combination of one or more of them. Embodiments of the subject matter described in this specification can be implemented as one or more computer programs, i.e., one or more modules of computer program instructions encoded on a computer storage medium (or medium) for execution by a data processing apparatus or for controlling the operation of a data processing apparatus. Alternatively or additionally, program instructions can be encoded on artificially generated propagating signals, such as machine-generated electrical, optical, or electromagnetic signals generated to encode information for transmission to a suitable receiver device for execution by the data processing apparatus. The computer storage medium can be or be included in a computer-readable storage device, a computer-readable storage substrate, a random or serial access memory array or device, or a combination of one or more of these. Furthermore, while the computer storage medium is not a propagating signal, it can be a source or destination of computer program instructions encoded in an artificially generated propagating signal. The computer storage medium can also be or be included in one or more separate physical components or media (e.g., multiple CDs, disks, or other storage devices).

[0198] The operations described in this specification can be implemented as operations performed by a data processing device on data stored on one or more computer-readable storage devices or received from other sources.

[0199] The term "data processing apparatus" encompasses a variety of devices, apparatuses, and machines used for processing data, including, for example, programmable processors, computers, system-on-a-chip, or a combination of the aforementioned devices. Apparatus can include special-purpose logic circuitry, such as FPGAs (Field-Programmable Gate Arrays) or ASICs (Application-Specific Integrated Circuits). In addition to hardware, apparatus can also include code that creates an execution environment for the computer program in question, such as code constituting processor firmware, protocol stacks, database management systems, operating systems, cross-platform runtime environments, virtual machines, or combinations thereof. Apparatus and execution environments can implement a variety of different computing model infrastructures, such as web services, distributed computing, and grid computing infrastructures.

[0200] Computer programs (also known as programs, software, software applications, scripts, or code) can be written in any programming language, including compiled or interpreted languages, declarative or procedural languages, and can be deployed in any form, including as standalone programs or modules, components, subroutines, objects, or other units suitable for use in a computing environment. Computer programs can, but do not necessarily, correspond to files in a file system. Programs can be stored as part of a file containing other programs or data (e.g., one or more scripts stored in a markup language document), in a single file dedicated to the program in question, or in multiple coordinating files (e.g., a file storing one or more modules, subroutines, or code sections). Computer programs can be deployed to execute on a single computer or on multiple computers located at a site or distributed across multiple sites interconnected by a communication network.

[0201] The processes and logic flows described in this specification can be executed by one or more programmable processors that execute one or more computer programs to perform actions by manipulating input data and generating outputs. The processes and logic flows can also be executed by dedicated logic circuits, and the devices can also be implemented as dedicated logic circuits, such as FPGAs (Field-Programmable Gate Arrays) or ASICs (Application-Specific Integrated Circuits).

[0202] As an example, processors suitable for executing computer programs include both general-purpose and special-purpose microprocessors. Typically, a processor receives instructions and data from read-only memory or random access memory, or both. The basic components of a computer are a processor for performing actions according to instructions and one or more memory devices for storing instructions and data. Typically, a computer will also include one or more mass storage devices for storing data, or operatively coupled thereto, to receive data from or transfer data to or to both, such as magnetic disks, magneto-optical disks, or optical disks. However, a computer does not necessarily have to have such devices. Furthermore, a computer can be embedded in another device, such as a mobile phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a global positioning system (GPS) receiver, or a portable storage device (e.g., a universal serial bus (USB) flash drive), to name just a few. Devices suitable for storing computer program instructions and data include all forms of non-volatile memory, media, and memory devices, such as semiconductor memory devices, e.g., EPROM, EEPROM, and flash memory devices; magnetic disks, such as hard disks or removable disks; magneto-optical disks; and CD-ROM and DVD-ROM discs. The processor and memory can be supplemented or incorporated by dedicated logic circuitry.

[0203] To provide interaction with the user, embodiments of the subject matter described in this specification can be implemented on a computer with a display device, such as a CRT (cathode ray tube) or LCD (liquid crystal display) monitor, to display information to the user, and a keyboard and pointing device, such as a mouse and trackball, through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback, such as visual, auditory, or tactile feedback; and input from the user can be received in any form, including acoustic, speech, or tactile input. Additionally, the computer can interact with the user by sending documents to and receiving documents from the device used by the user; for example, by sending web pages to a web browser on the user's client device in response to a request received from a web browser.

[0204] Embodiments of the subject matter described in this specification can be implemented in a computing system that includes backend components, such as a data server; or includes middleware components, such as an application server; or includes frontend components, such as a client device having a graphical user interface or web browser that a user can interact with through embodiments of the subject matter described in this specification; or any combination of one or more such backend, middleware, or frontend components. The components of the system can be interconnected via any form or medium of digital data communication, such as a communication network. Examples of communication networks include local area networks (“LANs”) and wide area networks (“WANs”), interconnected networks (e.g., the Internet) and peer-to-peer networks (e.g., ad hoc peer-to-peer networks).

[0205] A computing system can include clients and servers. Clients and servers are typically geographically separated and usually interact via a communication network. The client-server relationship arises through computer programs running on respective computers and the existence of a client-server relationship between them. In some embodiments, the server transmits data (e.g., HTML pages) to the client device (e.g., for the purpose of displaying data to a user interacting with the client device and receiving user input). Data generated at the client device (e.g., the result of user interaction) can be received from the client device at the server.

[0206] While this specification contains numerous specific details of implementation, these should not be construed as limiting the scope of any invention or the scope that may be claimed, but rather as descriptions of specific features of particular embodiments of a particular invention. Certain features described in this specification in the context of separate embodiments can also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment can also be implemented separately or in any suitable sub-combination in multiple embodiments. Furthermore, although features may be described above as functioning in certain combinations, or even as initially claimed, in some cases one or more features from the claimed combination can be removed from the combination, and the claimed combination may involve sub-combinations or variations of sub-combinations.

[0207] Similarly, although operations are depicted in a specific order in the accompanying drawings, this should not be construed as requiring such operations to be performed in the specific order shown or in a sequential order, or requiring all shown operations to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. Furthermore, the separation of various system components in the above embodiments should not be construed as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.

[0208] Therefore, specific embodiments of the subject matter have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recited in the claims can be performed in a different order and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific order shown, or a sequential order, to achieve the desired result. In some embodiments, multitasking and parallel processing may be advantageous.

[0209] The following is a non-exhaustive list of the various aspects of this disclosure.

[0210] Aspect 1. A method for monitoring privacy-preserving cross-domain experiments, comprising: receiving a request for digital content from an application on a user device by a first server of a secure multi-party computation (MPC) system, wherein the request includes a first secret share of an application identifier identifying the application; in response to receiving the request and by the first server of the secure MPC system: cooperating with a second server of the secure MPC system to perform a privacy-preserving selection process to select a first winning digital component from a set of digital components, wherein each digital component has a corresponding (i) unique experiment identifier and (ii) unique control identifier, the privacy-preserving selection process comprising: for each digital component in the set of digital components: cooperating with the second server and determining a first secret share of candidate parameters representing whether the digital component is a candidate for selection in the privacy-preserving selection process based on a set of eligibility rules; And, in cooperation with the second server, a secret share of an experimental parameter indicating whether the application is in the experimental group of the digital component is determined based on the first secret share of the application identifier and the corresponding first secret share of the candidate parameter value; a first secret share representing the selection result of the first selected digital component is generated based on the first secret share of the candidate parameter of each digital component and the selection value of each digital component; and a response to the request for digital content is generated, the response including: the first secret share of the selection result, and, for each of one or more digital components including the first selected digital component, data indicating whether the application is in the experimental group of the digital component or the control group of the digital component; and the response to the request for digital content is transmitted by the first server of the secure MPC system to the application on the user device.

[0211] Aspect 2. According to the method of aspect 1, wherein the data indicating whether the application is in the experimental group of the digital component or in the control group of the digital component is represented by a Bloom filter.

[0212] Aspect 3. According to the method of Aspect 2, wherein the privacy-preserving selection process for selecting the first winning number component includes: For each digital component in the set of digital components: Add corresponding (i) unique experimental identifiers and (ii) unique control identifiers to the Bloom filter.

[0213] Aspect 4. The method according to any one of Aspects 1 to 3, wherein determining whether the application is part of a control group or an experimental group with respect to the first winning number component includes: For each experiment in the set of experiments associated with the digital component, determine whether the application is part of a control group or an experimental group with respect to the first winning digital component.

[0214] Aspect 5. The method according to any one of Aspects 1 to 4, wherein the set of eligibility rules comprises two or more eligibility rules. In this context, one or more parameters of the first eligibility rule are determined based on an experimental design, which is based on several mutually orthogonal Latin matrices of a specific size, and The first eligibility rule specifies a first range of values ​​for the one or more parameters.

[0215] Aspect 6. According to the method of aspect 5, it further includes: Based on the experimental design, update one or more parameters of the first eligibility rule to specify a second range of values ​​for the one or more parameters; and The first server of the secure MPC system, in cooperation with the second server of the secure MPC system, performs a counterfactual selection process to select a second winning selection criterion from the set of selection criteria by applying the set of eligibility rules, which includes the updated first eligibility rule.

[0216] Aspect 7. The method according to any one of Aspects 1 to 6, wherein the transmission of the response to the request for digital content from the first server of the secure MPC system to the application on the user device comprises: The encrypted secret share of the response to the request for digital content is received by the first server of the secure MPC system from the second server of the secure MPC system; The first server of the secure MPC system encrypts a second secret share of the response to the request for digital content to generate an encrypted second secret share of the response; and The first server of the secure MPC system transmits the encrypted secret share of the response and the encrypted second secret share of the response to the application on the user device.

[0217] Aspect 8. The method according to any one of Aspects 1 to 7, wherein the set of eligibility rules comprises one or more eligibility rules. The set of eligibility rules includes at least one affirmative condition and at least one negative condition, and The selection criteria set is stored in the secure MPC system.

[0218] Aspect 9. The method according to any one of Aspects 1 to 8, wherein attaching the unique experimental identifier associated with the selection criteria to the response to the request for digital content includes adding the unique experimental identifier to a list of unique identifiers included in the response, and

[0219] Specifically, attaching the unique control identifier associated with the selection criteria to the response to the request for digital content includes adding the unique control identifier to the list of unique identifiers, and

[0220] The list of unique identifiers is represented using a Bloom filter.

[0221] Aspect 10. The method according to aspect 9 further includes: A first report, including a unique identifier, is transmitted by the application to the demand-side platform associated with the selected digital content; and A batch of one or more reports, including the first report, is uploaded by the demand-side platform associated with the selected digital component to the aggregated reporting server.

[0222] Aspect 11. A system comprising: The first server of the secure multi-party computation system, and The second server of the secure multi-party computation system The system is configured to perform the method according to any one of aspects 1 to 10.

Claims

1. A method for monitoring cross-domain experiments with privacy protection, comprising: A first server of a secure multi-party computation system receives a request for digital content from an application on a user device, wherein the request includes a first secret share identifying an application instance associated with the user device or an application instance identifier identifying the user device itself. In response to receiving the request, the first server of the secure multi-party computation system: In collaboration with a second server of the secure multi-party computation system, a privacy-preserving selection process is performed to select a first winning digital component from a set of digital components, wherein each digital component has a corresponding (i) unique experimental identifier and (ii) unique control identifier, the privacy-preserving selection process comprising: For each digital component in the set of digital components: Collaborating with the second server and determining a first secret share of candidate parameters representing whether the digital component is a qualified candidate for selection in the privacy-preserving selection process, based on a set of eligibility rules; Collaborating with the second server, and based on the first secret share of the application instance identifier and the corresponding first secret share of the candidate parameter value, a secret share of the experimental parameter indicating whether the application is in the experimental group of the digital component is determined. Based on the first secret share of the candidate parameters of each digital component and the selection value of each digital component, a first secret share representing the selection result of the first winning digital component is generated; and Generate a response to the request for digital content, the response including: the first secret share of the selection result, and, for each of one or more digital components including the first winning digital component, indicating whether the application is in the experimental group or the control group of the digital component; and The first server of the secure multi-party computation system transmits the response to the request for digital content to the application on the user device.

2. The method of claim 1, wherein, The data indicating whether the application is in the experimental group or the control group of the digital component is represented by a Bloom filter.

3. The method of claim 2, wherein, The privacy-preserving selection process for choosing the first winning number component includes: For each digital component in the set of digital components: Add a corresponding (i) unique experiment identifier if the application is in the experiment group, or (ii) unique control identifier if the application is in the control group to the Bloom filter.

4. The method of claim 1, wherein, Determining whether the application is part of a control group or an experimental group for each digital component includes: For each experiment in the set of experiments associated with the digital component, determine whether the application is part of a control group or an experimental group related to the digital component.

5. The method of claim 1, wherein, The eligibility rules may include two or more eligibility rules. In this context, one or more parameters of the first eligibility rule are determined based on an experimental design, which is based on several mutually orthogonal Latin matrices of a specific size, and The first eligibility rule specifies a first range of values ​​for the one or more parameters.

6. The method of claim 5, further comprising: Based on the experimental design, update one or more parameters of the first eligibility rule to specify a second range of values ​​for the one or more parameters; as well as The first server and the second server of the secure multi-party computation system collaborate to perform a counterfactual selection process to select a second winning digital component from the set of digital components by applying the set of eligibility rules, including the updated first eligibility rule.

7. The method of claim 1, wherein, The response transmitted by the first server of the secure multi-party computation system to the application on the user device for the request for digital content includes: The encrypted secret share of the response to the request for digital content is received by the first server of the secure multi-party computation system from the second server of the secure multi-party computation system; The first server of the secure multi-party computation system encrypts a second secret share of the response to the request for digital content to generate an encrypted second secret share of the response; and The first server of the secure multi-party computation system transmits the encrypted secret share of the response and the encrypted second secret share of the response to the application on the user device.

8. The method of claim 1, wherein, The set of eligibility rules includes one or more eligibility rules. The set of eligibility rules includes at least one affirmative condition, and The set of eligibility rules is stored in the secure multi-party computation system.

9. The method of claim 8, wherein, The set of eligibility rules includes at least one negative condition.

10. The method of any one of claims 1 to 9, wherein, Attaching the unique experimental identifier associated with the digital component to the response to the request for digital content includes adding the unique experimental identifier to a list of unique identifiers included in the response. Specifically, attaching the unique control identifier associated with the digital component to the response to the request for digital content includes adding the unique control identifier to the list of unique identifiers, and The list of unique identifiers is represented using a Bloom filter.

11. The method of claim 10, further comprising: The application transmits a first report, including a unique identifier, to the demand-side platform associated with the selected digital content; as well as A batch of one or more reports, including the first report, is uploaded by the platform associated with the selected digital component to the aggregated report server.

12. The method of claim 10, further comprising: The application transmits a first report, including a unique identifier, to the aggregated report server.

13. A system for monitoring privacy-preserving cross-domain experiments, the system comprising: One or more processors; as well as One or more memory elements, including instructions that, when executed, cause the one or more processors to perform operations, the operations including: A first server of a secure multi-party computation system receives a request for digital content from an application on a user device, wherein the request includes a first secret share identifying an application instance associated with the user device or an application instance identifier identifying the user device itself. In response to receiving the request, the first server of the secure multi-party computation system: In collaboration with a second server of the secure multi-party computation system, a privacy-preserving selection process is performed to select a first winning digital component from a set of digital components, wherein each digital component has a corresponding (i) unique experimental identifier and (ii) unique control identifier, the privacy-preserving selection process comprising: For each digital component in the set of digital components: Collaborating with the second server and determining a first secret share of candidate parameters representing whether the digital component is a qualified candidate for selection in the privacy-preserving selection process, based on a set of eligibility rules; Collaborating with the second server, and based on the first secret share of the application instance identifier and the corresponding first secret share of the candidate parameter value, a secret share of the experimental parameter indicating whether the application is in the experimental group of the digital component is determined. Based on the first secret share of the candidate parameters of each digital component and the selection value of each digital component, a first secret share representing the selection result of the first winning digital component is generated; and Generate a response to the request for digital content, the response including: a first secret share of the selection result, and, for each of one or more digital components including the first winning digital component, data indicating whether the application is in the experimental group or the control group of the digital component; and The first server of the secure multi-party computation system transmits the response to the request for digital content to the application on the user device.

14. The system of claim 13, wherein, The data indicating whether the application is in the experimental group or the control group of the digital component is represented by a Bloom filter.

15. The system of claim 14, wherein, The privacy-preserving selection process for choosing the first winning number component includes: For each digital component in the set of digital components: Add a corresponding (i) unique experiment identifier if the application is in the experiment group, or (ii) unique control identifier if the application is in the control group to the Bloom filter.

16. The system of claim 13, wherein, Determining whether the application is part of a control group or an experimental group for each digital component includes: For each experiment in the set of experiments associated with the digital component, determine whether the application is part of a control group or an experimental group related to the digital component.

17. The system according to any one of claims 13 to 16, further comprising the second server of the secure multi-party computation system.

18. A non-transitory computer-readable storage medium encoded with instructions, said instructions, when executed by a distributed computing system, causing the distributed computing system to perform an operation, said operation comprising: A first server of a secure multi-party computation system receives a request for digital content from an application on a user device, wherein the request includes a first secret share identifying an application instance associated with the user device or an application instance identifier identifying the user device itself. In response to receiving the request, the first server of the secure multi-party computation system: In collaboration with a second server of the secure multi-party computation system, a privacy-preserving selection process is performed to select a first winning digital component from a set of digital components, wherein each digital component has a corresponding (i) unique experimental identifier and (ii) unique control identifier, the privacy-preserving selection process comprising: For each digital component in the set of digital components: Collaborating with the second server and determining a first secret share of candidate parameters representing whether the digital component is a qualified candidate for selection in the privacy-preserving selection process, based on a set of eligibility rules; Collaborating with the second server, and based on the first secret share of the application instance identifier and the corresponding first secret share of the candidate parameter value, a secret share of the experimental parameter indicating whether the application is in the experimental group of the digital component is determined. Based on the first secret share of the candidate parameters of each digital component and the selection value of each digital component, a first secret share representing the selection result of the first winning digital component is generated; and Generate a response to the request for digital content, the response including: a first secret share of the selection result, and, for each of one or more digital components including the first winning digital component, data indicating whether the application is in the experimental group or the control group of the digital component; and The first server of the secure multi-party computation system transmits the response to the request for digital content to the application on the user device.

19. The non-transitory computer-readable storage medium of claim 18, wherein, The data indicating whether the application is in the experimental group or the control group of the digital component is represented by a Bloom filter.

20. The non-transitory computer-readable storage medium according to claim 19, wherein, The privacy-preserving selection process for choosing the first winning number component includes: For each digital component in the set of digital components: Add a corresponding (i) unique experiment identifier if the application is in the experiment group, or (ii) unique control identifier if the application is in the control group to the Bloom filter.

21. The non-transitory computer-readable storage medium according to any one of claims 18 to 20, wherein, Determining whether the application is part of a control group or an experimental group for each digital component includes: For each experiment in the set of experiments associated with the digital component, determine whether the application is part of a control group or an experimental group related to the digital component.

Citation Information

Patent Citations

  • System And Method For A Practical, Secure And Verifiable Cloud Computing For Mobile Systems

    US20150341326A1

  • Privacy preserving computation protocol for data analytics

    US20180373882A1