Systems and methods for determining whether a uniform resource locator is present in a blocklist

GB2706069APending Publication Date: 2026-11-11NETWORK GUARD PTE LTD
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
GB2025011968
Authority / Receiving Office
GB · GB
Patent Type
Applications
Current Assignee / Owner
Filing Date
2025-07-23
Publication Date
2026-11-11

Smart Images

  • Figure 00000000_0000_ABST
    Figure 00000000_0000_ABST
Patent Text Reader

Abstract

A method for determining whether a Uniform Resource Locator, URL, is present in a blocklist, wherein a URL is hashed and initially categorized as either being on or not on the blocklist. Upon categori
Need to check novelty before this filing date? Find Prior Art

Description

TECHNICAL FIELD

[0001] The disclosure relates to systems, methods, data processing apparatuses and computer-readable storage media for determining whether a uniform resource locator (URL) is present in a blocklist. BACKGROUND

[0002] The exponential growth of malicious web content (e.g., phishing, malware, and child sexual abuse content (CSAM)) has created an urgent need for URL filtering services. Traditional blocklistbased approaches suffer from fundamental architectural limitations that compromise both security and privacy. SUMMARY

[0003] One or more aspects of the present disclosure are set out in the independent claims. Other aspects and features of the present disclosure are set out in the dependent claims and in the description below. BRIEF DESCRIPTION OF THE DRAWINGS

[0004] The disclosure will now be described in more detail in connection with a number of exemplary embodiments shown in the accompanying drawings, in which: Fig. 1 shows a schematic diagram of the system of the present disclosure; Fig. 1a shows a schematic diagram of a computing device; Fig. 2 shows a schematic diagram overview of the system of the present disclosure performing a method for determining whether a Uniform Resource Locator, URL, is present in a blocklist; Fig. 3 shows a schematic diagram overview of a part of the system performing a part of the method for determining whether a URL is present in a blocklist; Fig. 4 shows a flowchart depicting the method for determining whether a URL is present in a blocklist performed by the system of the present disclosure; Fig. 5 shows a flowchart depicting the steps of the method for determining whether a URL is present in a blocklist performed by a user device; Fig. 6 shows a flowchart depicting the steps of the method for determining whether a URL is present in a blocklist performed by a service provider; and Fig. 7 shows a flowchart depicting the steps of the method for determining whether a URL is present in a blocklist performed by a blocklist provider. Like reference numerals are used for like components throughout the detailed description when appropriate. DETAILED DESCRIPTION

[0005] In overview, the disclosure relates to systems, computer-implemented methods, data processing apparatuses and computer-readable storage media for determining whether a uniform resource locator (URL) is present in a blocklist.

[0006] Herein is provided a computer-implemented method for determining whether a Uniform Resource Locator, URL, is present in a blocklist, the method comprising: hashing the URL; categorizing the hashed URL as either not being on the blocklist or possibly being on the blocklist; upon categorizing the hashed URL as possibly being on the blocklist: deriving from the hashed URL: at least one keyvalue pair, a first key and a second key; transmitting the at least one key-value pair and the first key to a first service provider; transmitting the at least one key-value pair and the second key to a second service provider; receiving, from the first service provider, a first result based on the at least one keyvalue pair and the first key; receiving, from the second service provider, a second result based on the at least one key-value pair and the second key; and determining whether the URL is present in the blocklist based on the first result and the second result.

[0007] Herein is also provided a computer-implemented method for determining whether a Uniform Resource Locator, URL, is present in a blocklist, the method comprising: receiving the blocklist; encoding the blocklist; receiving at least one key-value pair and a first key; querying the encoded blocklist based on the at least one key-value pair; based on the query selecting at least one hashed URL from the encoded blocklist and adding the at least one selected URL to a bucket; generating a first result by evaluating the first key over the bucket; and transmitting the first result to a user device.

[0008] Accordingly, advantageously, the present methods enable secure and efficient URL filtering, where neither the user’s inputted URL nor the blocklist’s entries are exposed during determination. Further, advantageously, the present methods ensure that any query from the user is not linkable to the user, therefore, behavioural tracking is avoided. Moreover, advantageously, the present methods support large-scale blocklists with millions of entries while maintaining low latency for real-time checks.

[0009] This overview is provided to introduce, in simplified form, a selection of concepts that are further described herein. The presence of a feature in this overview does not imply that the feature is essential to the disclosure.

[0010] The present disclosure will now be described in general terms before implementations of the disclosure are described by way of example with reference to the drawings. The words “may” is used to referto a feature that is optional, i.e. in some implementations ofthe disclosure, the feature is present, and in some implementations ofthe disclosure, the feature is not present.

[0011] Described herein is a computer-implemented method for determining whether a Uniform Resource Locator, URL, is present in a blocklist. The method may be performed by a system comprising a user device, a service provider and a blocklist provider. The service provider may comprise two noncolluding servers.

[0012] The user device may be a computing device. The computing device may be any suitable electronic device, e.g. a personal computer or computing device, a laptop, a tablet, a smart phone etc. It will be understood that this is a non-exhaustive and non-limiting list of example devices. Some of the components of the user device may comprise at least one processor coupled to memory. The at least one processor may comprise one or more of: a microprocessor, a microcontroller, and an integrated circuit. The memory may comprise volatile memory, such as random access memory (RAM), for use as temporary memory, and / or non-volatile memory such as Flash, read only memory (ROM), or electrically erasable programmable ROM (EEPROM), for storing data, programs, or instructions, for example. The computing device typically comprises at least one user interface for user input (e.g. input of the URL). The at least one user interface may take any appropriate form, e.g. a keyboard, a mouse, a touchpad or other input device or a computer screen. Similarly, the computing device typically comprises at least one display for providing the results (e.g. blocking access to the website represented by the URL; or accessing the website represented by the URL).

[0013] The user device may communicate with the service provider via the Internet.

[0014] The user device may be configured to perform the steps of the method depicted in Fig. 5.

[0015] The service provider may also be a computing device. The computing devices may comprise at least one processor coupled to at least one memory. The computing device may comprise other elements known in the art.

[0016] The service provider may comprise two servers. The service provider may comprise two noncolluding servers. The service provider may comprise at least two non-colluding service provider servers. It will be appreciated that more than one service provider may be used to implement the methods of the present disclosure. Each service provider may comprise at least one server. The servers may be non-colluding servers. Herein, the terms “first service provider” and “second service provider” may refer to two non-colluding servers of one service provider.

[0017] For instance, a first service provider may comprise a first service provider server and the second service provider may comprise another service provider server. In one or more examples, more than two servers may be used to implement the methods of the present disclosure. Each server may be hosted in independent cloud platform such as Amazon Web Services and Microsoft Azure. Within this extension, the privacy-preserving blocklist-based detection service can benefit from robustness against server failures.

[0018] To implement the method of the present techniques on more than two non-colluding servers, a threshold function secret sharing (FSS) may be used. This extends the two-party FSS (described in more detail below) to a t-out-of-T threshold scheme, which may include two algorithms: tFSS.Gen, and tFSS.Eval.

[0019] In particular, tFSS.Gen (11, f, t, T) —> (kt, k2, ■ ■ ■ , kr ): Given the security parameter A, the function description / , the threshold t, and the total number of servers T, generate T function shares, where any t - 1 function shares reveal nothing about the function / .

[0020] In particular, tFSS.Eval (kb, x) —> yb: Given an input x and the function share kb, output yb, which may correspond to party b’s share of / (x), b e [T], The t-out-of-T threshold FSS scheme ensures that at least t servers, which sum up their evaluation values at a given index a, can conclude the value of / (a). Thus, the present techniques can define the threshold t and the number T of servers in the setup.

[0021] Then, the algorithm for the second layer of the precise detection (described below) may be modified as follows: FineDete (m, sbh) —> (kt, fo). The methods may use FineDete*(m, sbh) —> (kt, fa, ■ ■ ■, kr) for the extension so as to distinguish from the original version. Specifically, FineDete*(m, sbh) —> (kt, k2, ■ ■ ■ , kr) may work as follows: The user device may be configured to sample sl / 2 indexes and corresponding bits of sbh to construct key-value pairs <f> = {(inda : vala)}, a e [sZ], |a| = sl / 2, where each index inda may correspond to the bit vala of sbh. The key-value pairs <f> may be delivered to t servers Sb, b e [T], |fe| = t. The user device may be configured to generate T keys (kt, k2, ■ ■ ■, kr) <— DPF.Gen (H(m), 1) and deliver the keys to the T servers. Upon receiving ¢, t servers Sb, b e [T], |fe| = t search BH based on each index bkt, i e n, respectively. In particular, if each corresponding bit of bkt indexed by inda equals vala, the hash value hi may be selected to join a small bucket smallB. For the small bucket smallB, t servers Sb, b e [T], |fe| = t evaluate kb over each hash value and sum up the evaluation values, i.e., ^be [T], \b\=t (^h^ EsmallB DPF.Eval(fa, hi)). If the sum equals one, it may mean that the URL of the user is on the blocklist. The t values can be returned to the user device which may be configured to check the sum of the two values, so only the user device learns the determination result.

[0022] In one or more examples, the service provider may comprise five servers across different providers / jurisdictions, where two servers are maintained by cloud provider 1 and cloud provider 2, and a server may be held by the government, a server may be held by an academic institute, and a server may be hosted by a non-governmental organization (NGO). The system may set the threshold tto three for the above instance, which means that the service needs responses from any three servers. Thus, advantageously, up to two servers can be offline or compromised without breaking the service. It will be appreciated that the number choices of the threshold and the number of servers are a trade-off between security and performance. More servers and a higher threshold guarantee stronger robustness, while the communication and computation overheads of the user device will increase. In some examples, a 2-out-of-3 threshold scheme may be viable fault tolerance, while a 3-out-of-5 threshold scheme may be a good balance of security and performance.

[0023] The service provider may communicate with the user device via the Internet. The service provider may communicate with the blocklist provider via the Internet. Other communication methods may be employed.

[0024] Each server may be configured to perform the steps of the method depicted in Fig. 6.

[0025] The blocklist provider may also be a computing device. The computing devices may comprise at least one processor coupled to at least one memory. The computing device may comprise other elements known in the art.

[0026] The blocklist provider may be configured to collect and / or audit and / or update URLs to build at least one blocklist. The blocklist provider may be configured to hash the blocklist. The blocklist provider may be configured to hash the blocklist to produce fixed-length (e.g. 256-bit) hashes of URLs by a collision-resistant hash function. The Collusion-resistant hashing functions are a class of keyed hash functions that ensure output unpredictability and privacy in the presence of adversarial collusion among multiple parties. These functions may be used in cryptographic protocols where multiple users operate independently, yet the users must not be able to infer information about the output of other users’ hash functions, even if they collude.

[0027] The blocklist may comprise at least one URL that is harmful (e.g. URLs that are associated with websites that comprise CSAM content.) The blocklist may comprise at least one URL. The blocklist may comprise at least one hashed URL.

[0028] The blocklist provider may be configured to protect the blocklist against external and / or internal leakages. The blocklist provider may provide more than one blocklist. The system may comprise more than one blocklist provider.

[0029] The blocklist provider may be configured to perform the steps of the method depicted in Fig. 7.

[0030] The system may further comprise an enclave. The enclave may be configured to store at least one blocklist. The enclave may be configured to hash each URL in the at least one blocklist. The enclave may be in communication with at least one service provider. The enclave may be configured to perform same method steps as the blocklist provider. The enclave may be used instead of the blocklist provider. In this example, the enclave may receive at least one blocklist from the blocklist provider. In one or more examples, the enclave may be used in connection with the blocklist provider. Advantageously, the enclave provides extremely secure environment for the storage of at least one blocklist.

[0031] Described herein are the steps of the method for determining whether a URL is present in a blocklist performed by a user device.

[0032] The method may comprise obtaining a URL. The URL may be input by a userto the user device. The URL may be associated with a website that the user is trying to access. The URL may be input by a user via a user interface of the user device.

[0033] The method may comprise hashing the URL. Thus, advantageously, the privacy of the user is increased. A lightweight protocol may be used to generate the hashed URL.

[0034] The method may comprise categorizing the hashed URL as either not being on the blocklist or possibly being on the blocklist.

[0035] The step of categorizing the hashed URL as either not being on the blocklist or possibly being on the blocklist may comprise encoding the hashed URL. The step of categorizing the hashed URL as either not being on the blocklist or possibly being on the blocklist may comprise performing a local query on the user device using the encoded hashed URL with an encoded filter to categorize the hashed URL. Thus, advantageously, the user device does not have to have full blocklist saved locally which saves computing power and makes the determination more efficient. The method may further comprise obtaining a filter. The filter may be obtained from a service provider. The filter may be downloaded from a service provider.

[0036] Encoding the hashed URL may comprise encoding the hashed URL in association with a service provider. Encoding the hashed URL may comprise transmitting the hashed URL to a service provider server for encoding; and receiving the encoded hashed URL. In some examples, encoding the hashed URL may comprise encoding the hashed URL in association with a service provider, wherein the service provider holds a secret key for encoding. Any service provider server may encode the hashed URL with the secret key. The secret key may be used by the service provider to encode the blocklist. The secret key may be used by the service provider to encode the filter. The secret key may be termed “a blocklist key”. The blocklist key is not known to the user device. Only the service provider servers hold the secret key.

[0037] A filter may be used to categorize the hashed URL as either not being on the blocklist or possibly being on the blocklist. The filter may be a probabilistic data structure. The filter may be a probabilistic data structure for a membership test. The filter may be encoded. The filter may be encoded by a service provider.

[0038] The filter may be a Bloom filter or Counting Bloom filter or Cuckoo filter or XOR filter or Ribbon filter. The filter may be comprise a combination of at least one of: a Bloom filter, Counting Bloom filter, Cuckoo filter, XOR filter, and Ribbon filter.

[0039] The step of obtaining a filter from a service provider may comprise obtaining the encoded filter from the first service provider or the second service provider. The encoded filter may be encoded Bloom filter. The encoded Bloom filter may comprise a compressed blocklist. Thus, advantageously, approximate filtering may be performed locally on the user device without the need of having the entire blocklist on the user device.

[0040] A Bloom filter is a space-efficient probabilistic data structure for high-speed set membership tests. The Bloom filter may use an m-bit array to represent a set of at most n items. For each query item x, it will be mapped into j positions, which is calculated by a set of; independent uniform hash functions He = {Hi, H2, ■ ■ ■, Hj}. If / v Hi (x) = 1 where a is the bitwise AND operation, x is probably in the set (with a tunable probability e of being wrong); otherwise, x is not in the set (with no false negatives). The lower the false positive rate e is, the more bits the filter costs. The Bloom filter may be used for filtering, which allows a tolerable false positive rate.

[0041] Alternative probabilistic data structures may be used for filtering instead of the Bloom filter. These alternatives include variants such as the Counting Bloom filter, Cuckoo filter, XOR filter, and Ribbon filter. While all support approximate membership testing, they exhibit distinct trade-offs in memory efficiency, computational speed, deletion support, and false-positive rates. Thus, advantageously, the characteristics of the respective filters may be leveraged to configure different blocklist-based determination systems and methods in accordance with specific requirements, taking into account the various types of blocklists and corresponding determination objectives.

[0042] The Bloom filter may be advantageous for high-speed, memory-efficient membership queries where a controlled false-positive rate is permissible. It performs well in caching systems, database optimizations, and network applications.

[0043] The Counting Bloom filter extends the Bloom filter by incorporating reference counters, thereby enabling deletion operations. However, this enhancement comes at a significant memory overhead (typically 3 - 4x that of a conventional Bloom filter), making it suitable for use cases where dynamic modifications (e.g., database caches requiring item removal) justify the additional memory cost.

[0044] The Cuckoo filter improves upon the Bloom filter and the Counting Bloom filter by supporting deletions without counters for better space efficiency and lower false positive rates. However, its insertion mechanism may fail under high loads due to excessive relocations, imposing a practical load factor limit (~ 95%). Consequently, it is ill-suited for high-insertion workloads or extremely large datasets, where resizing operations could degrade performance.

[0045] The XOR filter achieves near-optimal space efficiency with fast lookups, requiring only three memory accesses per query. Employing XOR-based fingerprinting reduces memory usage compared to Bloom and Cuckoo filters while maintaining a low false-positive rate. However, its static nature necessitates preprocessing, rendering it unsuitable for dynamic datasets. Thus, it is best applied in scenarios with fixed data, such as embedded systems or read-intensive applications, where memory constraints are critical.

[0046] The Ribbon filter optimizes both space and query performance through a linear-algebraic approach. Although it offers superior space efficiency, its construction involves solving a linear system, resulting in higher initialization overhead compared to other filters. Additionally, its lookup speed may marginally lag behind the XOR filter due to computational complexity. While highly effective for static, read-heavy workloads (e.g., genomic data or large-scale storage systems), its inflexibility and construction latency make it less adaptable than Bloom or Cuckoo filters for dynamic environments.

[0047] If the encoded hash of the URL does not exist in the filter, it means that the URL is not on the blocklist, and the user can normally access the URL through the networking service. Therefore, the method may comprise when the URL is not on the blocklist, accessing a website associated with the URL. Thus, advantageously, only the query hash of a URL that does not pass the local check will perform oblivious service provider side exact detection which improves the efficiency and speed of the method.

[0048] The method may further comprise upon categorizing the hashed URL as possibly being on the blocklist: deriving from the hashed URL: at least one key-value pair, a first key and a second key. The first key may comprise first data partially representing the hashed URL wherein the second key may comprise second data partially representing the hashed URL. The first key and the second key together may comprise data representing the hashed URL. The first key and the second key in combination may uniquely define the hashed URL and the hashed URL cannot be determined in isolation from either: the first key or the second key. Thus, advantageously, the user’s URL is not exposed during determination and any query from the user device is not linkable back to the user of the user device so behavioural tracking is avoided.

[0049] Deriving the at least one key-value pair from the hashed URL may comprise sampling the hashed URL to generate the at least one key-value pair.

[0050] Sampling the hashed URL to generate the at least one key-value pair may comprise using a lightweight mechanism to generate the at least one key-value pair. Sampling the hashed URL to generate the at least one key-value pair may comprise applying a sampling to the length-Z index. Sampling the hashed URL to generate the at least one key-value pair may comprise choosing a designated number sl / 2 of bit indices in a random and uniform distribution. Sampling the hashed URL may comprise sampling and indexing the hashed URL. Deriving the at least one key-value pair from the hashed URL may comprise sampling and indexing the hashed URL.

[0051] The number of key value-pairs may be a half of the length of the hashed URL. For example, if the hashed URL has 32 bits, the number of key-value pairs is 16. The number of the key-value pairs may be sl / 2 key value pairs.

[0052] Each key value pair may comprise data representing a part of sampled and indexed hashed URL. Each key value pair may comprise data which enables the service provider to generate a bucket. Each key value pair may comprise data representing a partial index of the hashed URL.

[0053] Deriving the first key and the second key from the hashed URL may comprise encoding the hashed URL to generate the first key and the second key.

[0054] The method may further comprise transmitting the at least one key-value pair and the first key to a first service provider. The method may further comprise transmitting the at least one key-value pair and the second key to a second service provider.

[0055] Transmitting the at least one key-value pair may comprise transmitting the designated number sl / 2 of bits their indices to the first service provider. Transmitting the at least one key-value pair may comprise transmitting the designated number sl / 2 of bits their indices to the second service provider.

[0056] The first key and the second key may be used for two-party function secret sharing scheme. Two-party Function Secret Sharing (FSS) scheme decomposes a function / into two compact shares. Each share reveals no information about the function f, and only the sum of the evaluation outputs at a given index x results in the corresponding value of / (x). Formally, a two-party FSS scheme consists of a pair of algorithms (FSS.Gen, FSS.Eval): (1) FSS.Gen (11, / ) -> (Zci, fo): Given a security parameter A and a function description / , the method may comprise generating two function shares Zci and fo that define two function keys for / , (2) FSS.Eval (kb, x) —> yb. Given an input x and the function share kb, output yb, which corresponds to party b’s share of / (x). Distributed Point Function (DPF) scheme may be special FSS that guarantees DPF.Eval (fci, x) + DPF.Eval (fa, x) = p only if x = a for kt, k2 <— DPF.Gen (a, P) where a, p e {0, 1}' and kt, k2 e {0, 1};.

[0057] The method may further comprise receiving, from the first service provider, a first result based on the at least one key-value pair and the first key. The method may further comprise receiving, from the second service provider, a second result based on the at least one key-value pair and the second key. The method may further comprise determining whether the URL is present in the blocklist based on the first result and the second result. In some examples, determining whether the URL is present in the blocklist based on the first result and the second result may comprise determining that the sum of the first result and the second result equals one. Thus, advantageously, only the user device will know the result of the determination which enhances the security and privacy.

[0058] The method may further comprise preventing access to a website associated with the URL if it is determined that the URL is present in the blocklist. The method may further comprise accessing the website associated with the URL if it is determined that the URL is not present on the blocklist.

[0059] Described herein are also the steps of the method for determining whether a URL is present in a blocklist performed by each service provider server.

[0060] The method may comprise receiving the blocklist. The blocklist may be received from a blocklist provider. The blocklist may be received from an enclave. Receiving the blocklist may comprise receiving the blocklist through a secure and authenticated channel. Thus, advantageously, privacy and security is enhanced. It will be appreciated that more than one blocklist may be received.

[0061] The method may comprise encoding the blocklist. The blocklist may comprise at least one hashed URL. The blocklist may comprise at least one hashed URL which is associated with a harmful website. Encoding the blocklist may comprise indexing each hashed URL with a blocklist key. The blocklist key may be an encoded index. The blocklist key may not be known to the user device. Thus, advantageously, the user device cannot learn the blocklist.

[0062] The method may further comprise encoding a filter based on the encoded blocklist; and optionally transmitting the encoded filter to the user device. The encoding the filter based on the encoded blocklist may comprise generating encrypted hashes for the encoded blocklist and encoding the filter with the encrypted hashes. The encoded filter may be an encoded Bloom filter. It will be appreciated that the encoded filter may be any other similar data structure described herein (e.g. XOR filter). The filter may be chosen based on desired parameters. The encoded filter may be transmitted to the user device when the user device signs up for a service. The encoded filter may be transmitted to the user device when the user device signs up for a service and requests a download.

[0063] The method may comprise receiving at least one key-value pair and a first key. The first key may comprise first data partially representing the hashed URL. It will be appreciated that if the method steps were performed by a second service provider server the method may comprise receiving at least one key-value pair and a second key. The second key may comprise second data partially representing the hashed URL. The first key and the second key together comprise data representing the hashed URL. The key distribution (i.e. first key to a first service provider and a second key to a second service provider etc.) increases security and privacy of the method because the service provider cannot reconstruct the URL.

[0064] The method may comprise querying the encoded blocklist based on the at least one key-value pair. The method may comprise based on the query selecting at least one hashed URL from the encoded blocklist and adding the at least one selected URL to a bucket. The bucket may be shared with a second service provider. The bucket may comprise a sub-portion of the encoded blocklist. Thus, advantageously, the detection speed is increased. Thus, the methods described herein are scalable for large-scale blocklists.

[0065] The method may comprise generating a first result by evaluating the first key over the bucket. The method may comprise transmitting the first result to a user device. Thus, advantageously, the first service provider evaluates the first key over the bucket and the second service provider evaluates the second key over the bucket and return the results to the user device, without learning anything about the hashed URL. This increases security and privacy.

[0066] The method may further comprise updating the blocklist. The method may comprise regularly updating at least one blocklist. The method may comprise periodically synchronizing the filter with the user device’s local environment. The step of updating the blocklist may comprising hashing newly added URLs. The step of updating the blocklist may comprise encoding the filter by delta encoding. The step of updating the blocklist may comprise transmitting the updated encoded filter to the user device. Thus, advantageously, the blocklists may be updated to ensure detection against the latest abusive URLs.

[0067] To ensure detection against the latest abusive URLs, the blocklists may require regular updates once the blocklist providers (BPs) release new versions. The updates may be reflected in the encoded filter. For instance, Bloom filter BF may be synchronized to users’ local environment periodically while ensuring the privacy of the system. In one example, to add new URLs from the blocklist to a filter may comprise updating the Bloom filter BF (or any other similar filter) by inserting newly prepared encoded short hashes. Specifically, for each new URL us, the BP may compute the hash value hs and the BP may deliver hs to the SP (service provider). The SP may perform the same step with hs as in the steps of encoding the filter (for instance, EncoBFilter). For instance, an SP may run bks <— SH(hs) and sbks <— Hq (k ■ Hw (bks)). Then the SP may insert sbks into the Bloom filter BF. For clarity, we may use BFo and BFi to represent the original Bloom filter and the updated Bloom filter, respectively. To update the BFi to the user device storage, it is desirable to avoid any unnecessary user device storage and to reduce bandwidth. To achieve this, Delta encoding may be used. Delta encoding is a compression technique that stores only the differences (deltas) between sequential elements rather than full values. When the delta encoding is applied to filters (e.g. Bloom filter), delta encoding reduces storage and transmission costs, especially in dynamic or distributed systems where updates are frequent (e.g., phishing blocklists like PhishTank). In particular, the delta encoding may generate Delta filters A(only the changes from the previous Bloom filter BFo) by △ <— BFo ® BF^, where △ is delivered to the user device, who runs BF^ = BFo ® △ to get the updated Bloom filter.

[0068] The method may comprise removing at least one URL from at least one blocklist. In practice, such an update does not always add URLs but may need to remove outdated ones, e.g. PhishTank. To address this issue, Counting Bloom Filter (CBF) may be available for adding and deleting items, which uses counters (instead of bits) to support deletions. However, CBF may require more memory due to counters (typically 4-8 bits per counter). Thus, the system may comprise selecting a filter based on requirements. In some examples, the system may use one or more data structures.

[0069] The method may further comprise periodically refreshing the data structure. The method may further comprise periodically refreshing the filter. The method may further comprise periodically refreshing the Bloom filter. The Bloom filters like any data structure may fill up once too many URLs have been added to the data structure. To overcome this disadvantage, the method may comprise refreshing the data structure periodically so that userdevice can download the latest version of the filter.

[0070] Described herein are also data processing apparatuses comprising means for carrying out the methods described herein.

[0071] Described herein are also computer-readable storage media comprising the instructions which, when executed by a computer, cause the computer to carry out the methods described herein.

[0072] Each possible combination of any of the above-described features forms part of the present disclosure. A plurality of instances of any feature may also be present. By way of example, implementations of the disclosure will now be described with reference to the drawings. Each possible combination of any features described below with any features described above also forms part of the disclosure.

[0073] Fig. 1 shows a schematic diagram of a system 1 of the present disclosure. The system comprises a blocklist provider 10, a service provider comprising two non-colluding servers 20a, 20b and a user device 30. The first service provider server 20a is herein also denoted “S1”, “Si”. The second service provider server 20b is herein also denoted “S2”,“S2”. The blocklist provider 10 is in communication with both servers 20a,20b. The user device 30 is in communication with both servers 20a, 20b. The user device 30 is not in communication with the blocklist provider 10. The communication is via the Internet.

[0074] Fig. 1a shows a schematic diagram of a processing system 400. The processing system 400 may be a computing device. The computing device may be the blocklist provider 10. The computing device may be the first service provider server Si . The computing device may be the second service provider server S2. The computing device may be the user device 30.

[0075] With reference to Fig. 1a, processing system 400 suitable for carrying out the methods described herein will now be described. Fig. 1a shows a block diagram of one implementation of a processing system 400 in the form of a computing device within which a set of instructions for causing the computing device to perform any one or more of the methods described herein may be executed. In some implementations, the computing device may be connected (e.g., networked) to other machines in a Local Area Network (LAN), an intranet, an extranet, or the Internet. The computing device may operate in the capacity of a server or a client machine in a client-server network environment, or as a peer machine in a peer-to-peer (or distributed) network environment. The computing device may be a personal computer (PC), a tablet computer, a set-top box (STB), a Personal Digital Assistant (PDA), a cellular telephone, a web appliance, a server, a network router, switch or bridge, or any machine capable of executing a set of instructions (sequential or otherwise) that specify actions to be taken by that machine. Further, while only a single computing device is illustrated, the term ‘computing device’ shall also be taken to include any collection of machines (e.g., computers) that individually or jointly execute a set (or multiple sets) of instructions to perform any one or more of the methods described herein.

[0076] The example processing system 400 includes a processor 402, a main memory 404 (e.g., readonly memory (ROM), flash memory, dynamic random-access memory (DRAM) such as synchronous DRAM (SDRAM) or Rambus DRAM (RDRAM), etc.), a static memory 406 (e.g., flash memory, static random-access memory (SRAM), etc.), and a secondary memory (e.g., a data storage device 418), which communicate with each other via a bus 430.

[0077] Processor 402 represents one or more general-purpose processors such as a microprocessor, central processing unit, or the like. More particularly, the processor 402 may be a complex instruction set computing (CISC) microprocessor, reduced instruction set computing (RISC) microprocessor, very long instruction word (VLIW) microprocessor, processor implementing other instruction sets, or processors implementing a combination of instruction sets. Processor 402 may also be one or more special-purpose processors such as an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a digital signal processor (DSP), network processor, or the like. Processor 402 is configured to execute the processing logic (instructions 422) for performing the operations and steps described herein.

[0078] The processing system 400 may further include a network interface device 408. The processing system 400 also may include any of a video display unit 410 (e.g., a liquid crystal display (LCD) or a cathode ray tube (CRT)), an alphanumeric input device 412 (e.g., a keyboard or touchscreen), a cursor control device 414 (e.g., a mouse or touchscreen), and an audio device 416 (e.g., a speaker).

[0079] It will be apparent that some features of the processing system 400 shown in Fig. 1a may be absent. For example, the processing system 400 may have no need for display device 410 (or any associated adapters). This may be the case, for example, for particular server-side computer apparatuses which are used only fortheir processing capabilities and do not need to display information to users. Similarly, user input device 412 may not be required. In its simplest form, processing system 400 comprises processor 402 and main memory 404.

[0080] The data storage device 418 may include one or more machine-readable storage media (or more specifically one or more non-transitory computer-readable storage media) 428 on which is stored one or more sets of instructions 422 embodying any one or more of the methods or functions described herein. The instructions 422 may also reside, completely or at least partially, within the main memory 404 and / or within the processor 402 during execution thereof by the processing system 400, the main memory 404 and the processor 402 also constituting computer-readable storage media 428.

[0081] Fig. 2 shows a schematic diagram overview of the system 1 performing a method for determining whether a Uniform Resource Locator, URL, is present in a blocklist. Fig. 3 shows a schematic diagram overview of a part of the system 1 performing a part of the method for determining whether a URL is present in a blocklist. In particular, the part shown in Figure 3 relates to a coarsegrained layer which allows secure user-side approximate filtering.

[0082] Broadly speaking, the method comprises two main layers: 1) a coarse-grained layer which allows secure user-side approximate filtering, and 2) a fine-grained layer which supports oblivious exact detection.

[0083] The inventors recognized that maintaining full, encrypted blocklist on a user device would offer privacy and it would reduce the costs of the servers of the service providers, but at the expense of userside storage. To overcome this disadvantage, the method of the present disclosure implements the coarse-grained layer which allows secure user-side approximate filtering. The user device 30 is configured to store a compressed version of the blocklist so as to locally check the query hash of the URL (e.g. URLa inputted by User A) at a modest storage cost. A probabilistic data structure, such as a Bloom filter may represent the compressed blocklist.

[0084] However, the inventors also recognized that using the hashes of the URLs on the blocklist directly to generate a probabilistic data structure such as the Bloom filter would be vulnerable. Bloom filter, as a probabilistic data structure may be susceptible to differential attacks by adversary users who may be capable of inferring some hashes of URLs that have been inserted into the filter, given specific auxiliary information. To address this disadvantage, the present method uses an encoded Bloom filter where the items are encoded to hide the original distribution. The encoded Bloom filter is a custom filter.

[0085] It will be appreciated that one direct way to encode the hashes may be to leverage latency-aware homomorphic encryption schemes to encrypt the hashes of the blocklist. The service provider may generate a Bloom filter by using these encrypted hashes and then put it on the user device, where the user device produces the encrypted hash by using the same holomorphic encryption scheme and checks whether or not the encrypted hash is on the blocklist via the custom filter. Nevertheless, the inventors recognized that such a solution would incur heavy computation overhead on the user device side. Moreover, how to securely distribute the encryption key to distributed users is a question where one compromised user could break the whole privacy guarantees. Thus, the inventors recognized the need to find a solution that enables users to generate the encrypted hash without learning the encryption key.

[0086] In this problem, the user device has the hash of a URLa, and each service provider server Si, S2 has the encrypted hashes for a blocklist where the encryption key is only known by the service provider server Si, S2. The user device wants to obtain the encrypted hash encrypted by the same encryption key while the server cannot learn the original hash of the user. To address this disadvantage, the method of the present disclosure uses a cryptographic primitive oblivious pseudorandom function (OPRF) protocol. In brief, the user device 30 is configured to run an OPRF protocol with the service provider server Si, S2 to obtain the encrypted hash without revealing the original hash and encryption key to each other. Thus, the service provider server Si, S2 is able to generate a custom filter such as an encoded Bloom filter where each item is encrypted by the encryption key k (i.e. the secret key), and the user device 30 cannot infer any additional blocklist information when obtaining the encoded Bloom filter (or any other similar encoded data structure), since the user device 30 cannot know the encryption key k. In this context, the user device 30 can locally check the encoded hash over the custom filter, and if it does not exist, the user device 30 can access the URL since the underlying Bloom filter has no false negatives. If it exists, regarding the false positives of the Bloom filter, the method proceeds to the second layer: 2) the fine-grained layer which supports oblivious detection.

[0087] Regarding the second layer: 2) the fine-grained layer which supports oblivious detection, for a user device 30 whose URL requires exact detection over a blocklist, it is inefficient to scan the whole blocklist. To overcome this issue, the method of the present disclosure generates a small bucket, wherein the small bucket is shared between the service providers / servers Si, S2. The bucket comprises a sub-portion of the encoded blocklist. However, the inventors faced an issue on how to select a small bucket from the large-scale blocklist. To be specific, if the user device gives the index (i.e., a short hash of the hash of the URL; or the hash of the URL) of the URL directly to the service provider server, an adversarial server may attempt to learn, from the indices generated for some URLs drawn from an adversary-known distribution, a predicate about the underlying URL of the user. To overcome this disadvantage, instead of using the full index of the hashed URL to retrieve the hash on the blocklist and to add the URL from the blocklist to the smaller bucket, the present methods use a lightweight mechanism that increases adversarial uncertainty. For instance, the methods of the present disclosure apply a sampling to the length-Z index, choose a designated number sl / 2 of bit indices in a random and uniform distribution, and send the resulting sl / 2 bits and their indices to the service provider servers. A URL on the service provider server blocklist is included in the small bucket if it has the relevant sl / 2 bits of its index matching with those that are sent from the user device.

[0088] Thus, the service provider servers Si, S2 share identical small bucket where each entry includes the hash of a harmful URL taken from the blocklist. The user device 30 is configured to transmit encoded data (i.e. partial index which may be the key value pair) to the servers so as to determine whether a match on the blocklist exists, bearing in mind that the servers Si, S2 should not inferthe detection result, and the service provider servers Si, S2 should not infer which URL is being detected by the user device 30. With that, the present methods configure the user device 30 to encode the hash h as a key pair for a point function fa=h,p='\, where the key pair can be evaluated by two service provider servers Si, S2 on the blocklist B. The evaluation values are returned to the user device so as to reveal the determination result. The URL is blocked if the sum of the evaluation values is 1. Thus, further, the present methods prevent the user of the user device from accessing the website associated with the URL if it is determined that the URL is on the blocklist.

[0089] Accordingly, advantageously, the present methods enable secure and efficient URL filtering, where neither the user’s inputted URL nor the blocklist’s entries are exposed during determination. Further, advantageously, the present methods ensure that any query from the user is not linkable to the user, therefore, behavioural tracking is avoided. Moreover, advantageously, the present methods support large-scale blocklists with millions of entries while maintaining low latency for real-time checks.

[0090] Fig. 4 shows a flowchart of the system 1 performing the method for determining whether a Uniform Resource Locator, URL, is present in a blocklist.

[0091] The method comprises S100: storing, by a blocklist provider (BP), a blocklist. The blocklist comprises at least one URL. Each URL in the blocklist is associated with a harmful website.

[0092] The method comprises S105: hashing, by the BP, each URL in the blocklist. The blocklist provider may be configured to use collusion-resistant hashing function to hash each URL in the blocklist.

[0093] The method comprises S110: transmitting, from the BP to a first service provider (S1 / 20a), the blocklist with hashed URLs. The blocklist is transmitted through a secure and authenticated channel.

[0094] The method comprises S115: transmitting, from the BP to a second service provider (S2 / 20b), the blocklist with hashed URLs. The blocklist is transmitted through a secure and authenticated channel.

[0095] The method comprises S120:encoding, by the S1 and S2, the blocklist. Each service provider server encodes the blocklist independently. In this example, encoding the blocklist comprises indexing each hashed URL with a blocklist key. The blocklist key is not known to the user device.

[0096] The method comprises S125: encoding, by the S1 and S2, a filter based on the encoded blocklist. Each service provider server S1, S2 independently encodes a filter based on the encoded blocklist. In this example, encoding a filter based on the encoded blocklist comprises generating encrypted hashes for the encoded blocklist and encoding the filter with the encrypted hashes. In this example, the encoded filter is an encoded Bloom filter.

[0097] The method comprises S135: transmitting, from the S1 or S2 to the user device, the encoded filter. Upon registering with the service, the user device downloads the encoded filter from S1 or S2.

[0098] The method comprises S140: hashing, by the user device, the URL. In this example, the URL is input by a user to the user device. The URL is subsequently hashed.

[0099] The method comprises S145: categorizing, by the user device, the hashed URL as either not being on the blocklist or possibly being on the blocklist. In this example, categorizing the hashed URL as either not being on the blocklist or possibly being on the blocklist comprises encoding the hashed URL and performing a local query of the encoded hashed URL with an encoded filter to categorize the hashed URL. The encoded filter is the Bloom filter obtained from the S1 or S2. If it is categorized that the hashed URL is not being on the blocklist, the user device is configured to access the website associated with the URL.

[0100] The method comprises S150: upon categorizing, by the user device, the hashed URL as possibly being on the blocklist: deriving, by the user device, from the hashed URL: at least one keyvalue pair, a first key and a second key.

[0101] The first key comprises first data partially representing the hashed URL. The second key comprises second data partially representing the hashed URL. The first key and the second key together comprise data representing the hashed URL.

[0102] In this example, deriving the at least one key-value pair from the hashed URL comprises sampling the hashed URL to generate the at least one key-value pair.

[0103] In this example, deriving the first key and the second key from the hashed URL comprises encoding the hashed URL to generate the first key and the second key.

[0104] The method comprises S155: transmitting, from the user device to S1, the at least one keyvalue pair and the first key.

[0105] The method comprises S160: transmitting, from the user device to S2, the at least one keyvalue pair and the second key.

[0106] The method comprises S165: querying, by S1, the encoded blocklist based on the at least one key-value pair; and based on the query selecting at least one hashed URL from the encoded blocklist and adding the at least one selected URL to a bucket.

[0107] The method comprises S170: querying, by S2, the encoded blocklist based on the at least one key-value pair; and based on the query selecting at least one hashed URL from the encoded blocklist and adding the at least one selected URL to the bucket.

[0108] The bucket is shared between S1 and S2. The bucket comprises a sub-portion of the encoded blocklist.

[0109] The method comprises S175: generating, by S1, a first result by evaluating the first key over the bucket and transmitting the first result to the user device.

[0110] The method comprises S180: generating, by S2, a second result by evaluating the second key over the bucket and transmitting the second result to the user device.

[0111] The method comprises S185: receiving, by the user device, the first result and the second result.

[0112] The method comprises S190: determining whether the URL is present in the blocklist based on the first result and the second result. In this example, determining whether the URL is present in the blocklist based on the first result and the second result comprises determining that the sum of the first result and the second result equals one. If is it determined that the URL is on the blocklist, the user device is configured to block the access to the website associated with the URL. If it is determined that the URL is not on the blocklist, the user device is configured to access the website associated with the URL. The result of the determination may be displayed to a user of the user device.

[0113] Fig. 5 shows a flowchart depicting the steps of the method for determining whether a URL is present in a blocklist performed by a user device. To facilitate understanding of which entity performs each step, the steps in this figure are numbered differently from those in Fig. 4, even though the underlying actions may be the same.

[0114] The method comprises S200: hashing the URL. The URL is input by the user to the user device. The user may intend to visit a website associated with the URL. The input URL is hashed.

[0115] The method comprises S205: categorizing the hashed URL as either not being on the blocklist or possibly being on the blocklist. In this example, categorizing the hashed URL as either not being on the blocklist or possibly being on the blocklist comprises encoding the hashed URL and performing a local query of the encoded hashed URL with an encoded filter to categorize the hashed URL.

[0116] Encoding the hashed URL comprises transmitting the hashed URL to a random service provider for encoding. The service provider is configured to encode the hashed URL and return the encoded hashed URL to the user device. Encoding the hashed URL comprises receiving the encoded hashed URL.

[0117] The encoded filter is the Bloom filter obtained from the S1 or S2. Thus, categorizing the hashed URL as either not being on the blocklist or possibly being on the blocklist comprises obtaining the encoded filter from the first service provider or the second service provider; wherein the encoded filter is an encoded Bloom filter; wherein the encoded Bloom filter comprises a compressed blocklist. If it is categorized that the hashed URL is not being on the blocklist, the user device is accesses the website associated with the URL.

[0118] The method comprises S210: upon categorizing the hashed URL as possibly being on the blocklist: deriving from the hashed URL: at least one key-value pair, a first key and a second key. The first key comprises first data partially representing the hashed URL. The second key comprises second data partially representing the hashed URL. The first key and the second key together comprise data representing the hashed URL.

[0119] In this example, deriving the at least one key-value pair from the hashed URL comprises sampling the hashed URL to generate the at least one key-value pair.

[0120] In this example, deriving the first key and the second key from the hashed URL comprises encoding the hashed URL to generate the first key and the second key.

[0121] The method comprises S215: transmitting the at least one key-value pair and the first key to a first service provider.

[0122] The method comprises S220: transmitting the at least one key-value pair and the second key to a second service provider.

[0123] The method comprises S225: receiving, from the first service provider, a first result based on the at least one key-value pair and the first key.

[0124] The method comprises S230: receiving, from the second service provider, a second result based on the at least one key-value pair and the second key.

[0125] The method comprises S235: determining whether the URL is present in the blocklist based on the first result and the second result. In this example, determining whether the URL is present in the blocklist based on the first result and the second result comprises determining that the sum of the first result and the second result equals one. If it is determined that the URL is present on the blocklist, the user device is denied access to a website associated with the URL. If it is determined that the URL is not present on the blocklist, the user device is allowed to access the website associated with the URL.

[0126] Fig. 6 shows a flowchart depicting the steps of the method for determining whether a URL is present in a blocklist performed by a service provider. To facilitate understanding of which entity performs each step, the steps in this figure are numbered differently from those in Fig. 4, even though the underlying actions may be the same.

[0127] The method comprises S300: receiving the blocklist. The blocklist is received through a secure and authenticated channel. The blocklist comprises at least one hashed URL.

[0128] The method comprises S305: encoding the blocklist. Encoding the blocklist comprises indexing each hashed URL with a blocklist key. The blocklist key is not known to the user device.

[0129] The method comprises S306: encoding a hashed URL received from the user device. The user device is configured to transmit the hashed URL to a random service provider. The service provider is configured to encode the hashed URL and return the encoded hashed URL to the user device.

[0130] The method comprises S307: encoding a filter based on the encoded blocklist. In this example, encoding a filter based on the encoded blocklist comprises generating encrypted hashes for the encoded blocklist and encoding the filter with the encrypted hashes. In this example, the encoded filter is an encoded Bloom filter. In this example, the method also comprises transmitting the encoded filter to the user device.

[0131] The method comprises S310: receiving at least one key-value pair and a first key. It will be appreciated that in the case of the second service provider server, the method comprises receiving at least one key-value pair and a second key.

[0132] The first key comprises first data partially representing the hashed URL. The second key comprises second data partially representing the hashed URL. The first key and the second key together comprise data representing the hashed URL.

[0133] The method comprises S315: querying the encoded blocklist based on the at least one keyvalue pair.

[0134] The method comprises S320: based on the query selecting at least one hashed URL from the encoded blocklist and adding the at least one selected URL to a bucket. The bucket is shared with a second service provider. The bucket comprises a sub-portion of the encoded blocklist.

[0135] The method comprises S325: generating a first result by evaluating the first key overthe bucket. It will be appreciated that in case of the second service provider, the method comprises generating a second result by evaluating the second key overthe bucket.

[0136] The method comprises S330: transmitting the first result to a user device. It will be appreciated that in case of the second service provider, the method comprises transmitting the second result to the user device.

[0137] Fig. 7 shows a flowchart depicting the steps of the method for determining whether a URL is present in a blocklist performed by a blocklist provider. To facilitate understanding of which entity performs each step, the steps in this figure are numbered differently from those in Fig. 4, even though the underlying actions may be the same.

[0138] The method comprises S400: storing the blocklist. The blocklist comprises at least one URL. Each URL in the blocklist is associated with a harmful website.

[0139] The method comprises S405: hashing each URL in the blocklist. The blocklist provider may be configured to use collusion-resistant hashing function to hash each URL in the blocklist.

[0140] The method comprises S410: transmitting the blocklist with at least one hashed URL to a first service provider. The transmission is through a secure channel.

[0141] The method comprises S415: transmitting the blocklist with at least one hashed URL to a second service provider. The transmission is through a secure channel.

[0142] In this exemplary implementation of the method for determining whether a Uniform Resource Locator, URL, is present in a blocklist, the filter is a Bloom filter.

[0143] A Bloom filter is a space-efficient probabilistic data structure for high-speed set membership tests. The Bloom filter is configured to use an m-bit array to represent a set of at most n items. For each query item x, it will be mapped into j positions, which is calculated by a set of j independent uniform hash functions He = {Eh, H2, ■ ■ ■, Hj}. If / V Hi (x) = 1 where a is the bitwise AND operation, x is probably in the set (with a tunable probability e of being wrong); otherwise, x is not in the set (with no false negatives).

[0144] In this exemplary implementation of the method for determining whether a Uniform Resource Locator, URL, an Oblivious Pseudorandom Function (OPRF) is implemented.

[0145] The Oblivious Pseudorandom Function (OPRF) protocol enables two parties, say party A holding an input string m and party B holding a secret key k, to jointly and securely compute a pseudorandom function (PRF) F( / c, m), i.e., {0,1}*x {0,1}*—> {0,1}'. This two-party computation protocol is operated obliviously in the sense that party A only learns the output value, while party B learns nothing from the interaction. The method described herein uses OPRF instantiation based on the security of the Elliptic Curve Discrete Logarithm Problem (ECDLP), namely EC-OPRF: F( / c, m) = Hq (k ■ Hw (m)), where Hw: {0,1}*—> E is a secure hash function that maps a binary string m of arbitrary length to a point G on an elliptic curve E, Hq: E —> {0, 1}; is a secure hash function that derives a binary string from an elliptic curve point G e E, k is a randomly selected secret key, and c ■ G denotes the multiplication of a scalar integer c with a point G on an elliptic curve. In particular, party A picks a random integer r and sends X = r ■ Hw (m) to party B. Then party B sends back Y = k ■ X. Lastly, party A computes Z = r “1 ■ Y = k ■ Hw (m) and obtains Hq (Z), i.e., Hq (k ■ Hw (m)), as the PRF output. Note that party A has access to the elliptic curve point Hw (m) and k ■ Hw (m), but party A is unable to recover party B’s secret k because this is the ECDLP.

[0146] In this exemplary implementation of the method for determining whether a Uniform Resource Locator, URL, a Collusion-resistant Hashing Function is used.

[0147] The Collusion-resistant hashing functions are a class of keyed hash functions that ensure output unpredictability and privacy in the presence of adversarial collusion among multiple parties. These functions are useful where multiple users operate independently, yet must not be able to infer information about the output of other users’ hash functions, even if they collude.

[0148] In this exemplary implementation of the method for determining whether a Uniform Resource Locator, URL, Function Secret Sharing (FSS) is implemented.

[0149] Two-party Function Secret Sharing (FSS) scheme decomposes a function / into two compact shares. Each share reveals no information about the function / , and only the sum of the evaluation outputs at a given index x results in the corresponding value of / (x). Formally, a two-party FSS scheme consists of a pair of algorithms (FSS.Gen, FSS.Eval): (1) FSS.Gen (1 / / ) —► (fci, fo): Given a security parameter A and a function description / , generate two function shares Zci and fo that define two function keys for / , (2) FSS.Eval (kb, x) yb Given an input x and the function share kb , output yb , which corresponds to party b’s share of / (x). Distributed Point Function (DPF) scheme is a special FSS that guarantees DPF.Eval (fci, x) + DPF.Eval (fe, x) = p only if x = a for ki, ki <— DPF.Gen (a, p) where a, p e {0, 1};and fa, k2 e {0, 1};.

[0150] Let $ denote uniform sampling. Let Z denote the integer field and A denote a security parameter. Let H :{0, 1}*—> {0, 1}' denote a collision-resistant hash function mapping an arbitrary-length string uniformly to a fixed string of length Z. Let SH : {0,1}; —> {0,1 }s; denote a short hash function where I >si. Let c G denote the multiplication of a scalar integer c with a point G on an elliptic curve. Let Hw: {0, 1}* —> E denote a hash function that maps a string m of arbitrary length to a point G on an elliptic curve E. Let Hq: E —> {0, 1}; denote a secure hash function that derives a binary string from an elliptic curve point G e E. Let k denote a secret key used for PRF F(k, x), where k is maintained by the S1 (first service provider server) and S2 (Second service provider server) of the service provider (SP) secretly. We assume a set of j independent uniform hash functions He = {Eh, H2, ■ ■ ■ , Hj} for Bloom filter. We assume that the blocklist B = w, U2, ■ ■ ■ , Un is the blocklist, where each m, i e n is an URL.

[0151] An exemplary, non-limiting, algorithmic implementation of the method for determining whether a URL is present in a blocklist, performed by the system described herein, is shown below. Algorithm Privacy-preserving Blocklist-based Detection (PPBD) Input: The blocklist provider (BP) holds the blocklist, first service provider server (Si) and the second service provider server (S2) hold the secret key k, the user device holds the URL m. Output: The first service provider server (Si) and the second service provider server (S2) hold an identical blocklist B. The user learns the detection result. EncoBlolist: Generated encoded blocklist The BP runs hi <— H(w) for each URL ut e B, i e n. The BP delivers these hash values to the Si and S2. The Si and S2, respectively, run bki <— SH (hi), and generate the encoded blocklist BH = {bki, hi). EncoBFilter: Generated encoded Bloom filter The Si and S2, respectively, run sbki <— Hq (k ■ Hw (bki)), and generate the encoded Bloom filter BF using sbki, i e n. The user device downloads BF when joining the system. CoarseFilt: Coarse-grained secure approximate filtering The user device runs sbh <— SH(H(m) for the URL m. The user device petforms r $ Z and X = r ■ Hw (sbh), where X is delivered to a random server Sb, b e {1,2}. The Sb produces Y = k ■ X and delivers Y to the user. The user device computes Z = r “1 ■ Y and HqZ to get sbm = Hq (k ■ Hw (sbh)). The user device checks sbm over BF to j hash functions. if AJi=1 Hi (sbm) = 0 then The user device can access the URL m. else The user device calls FineDete. end if FineDete: Fine-grained oblivious exact detection The user device produces key-value pairs <p = {(inda : vala)}, a e [sZ], |a| = sU2 where each sampling index inda corresponds to the bit vala of sbh, and delivers <p to the servers Si and S2. The user device generates (Zci, fo) <— DPF.Gen(H(m), 1 and delivers kb to the server Sb, b e {1,2}. Sb initiates a small bucket smallB. for (bkt, hi) e BH, i e n, do if bki (inda) = vala, (inda. vala) £ 0 then smallB.append (hi). end if end for Sfe executes tmpz, = DPF. EvalCkphj) .where i e smallB and returns tmpb to the user device. if tmpi + tmp2 = 1 then The URL is blocked. Only the user learns the detection result. end if

[0152] EncoBlolist (B) —> (BH) is the step of encoding blocklist. The blocklist provider (BP) maintains the blocklist B (i.e. S100 in Fig. 4) and executes hi <— H(w) for each ut e B, i e n (i.e. S105). The BP delivers all hashes of URLs on the blocklist B to the Si and S2 of the SP through a secure and authenticated channel (i.e. S110 and S115). Each server Sb, b e {1,2} generates encoded blocklist BH (i.e. S120), where each value hi is indexed by the secret key bki <— SH(fe). ie., BH = {(bki : hi)}. The system may comprise more than one blocklist provider with this algorithm.

[0153] EncoBFilter (BH, k) —> (BF) is the step of encoding a filter (i.e. S125). Each server Sb, b e {1,2} generates encrypted short hashes for BH, where each encrypted short hash sbki is generated by Hq (k ■ Hw (bki)). These short hashes are built to produce the encoded Bloom filter BF, which is downloaded to the user device when the user device registers to the service. In other words, the encoded Bloom filter BF is transmitted from the S1 or S2 to the user device (i.e. S135).

[0154] CoarseFilt (m, k) —> (sbm) relates to the steps of categorizing the URL in the user device. The user device is configured to use the URL m to generate short hash sbh = SH(H(m)) (i.e. S140). The user device is configured pick a random value r $ Z. The user device is configured to generate X = r ■ Hw (sbh) and delivers X to a random servers . The server Sb produces Y = k ■ X and delivers Y to the user device. The user device computes Z = r “1 ■ Y and Hq (Z) to obtain sbm = Hq (k ■ Hw (sbh)). Then, the user device searches sbm over BF by hashing sbm to He that includes j hash functions (i.e. S145). If Aj=1 Hi (sbm) = 0 then the URL passes the detection and the user device visits the URL. Else, the user calls FineDete.

[0155] FineDete (m, sbh) —> (Zci, fo) relates to the step of oblivious server-side exact detection. The user device is configured to sample sl / 2 indexes and corresponding bits of sbh to construct key-value pairs <p = {(inda: vala)}, a e [sZ], |a| = sl / 2, where each index inda corresponds to the bit vala of sbh (i.e. S150). The key-value pairs <p are delivered to the Si and S2 (i.e. S155 and S160). The user generates two keys (Zci, fo) <— DPF.Gen(H(m), 1) and delivers them to the Si and S2 (i.e. S150, S155 and S160). Upon receiving ¢, the Si and S2 are configured to search BH based on each index bki, i e n. In particular, if each corresponding bit of bki indexed by inda equals vala, the hash value hi is selected to join a small bucket smallB (i.e. S165 and S170). For the small bucket smallB, Si and S2 evaluate ki and k2 over each hash value and sum up the evaluation values to generate a first result and a second result, i.e., h DPF. Eval(k,, ht), where i e smallB and'£hi,iDPF.Eval(k2,hi), where i e smallB (i.e. S175 and S180.) The two values are returned to the user device (i.e. S185) that is configured to determine whether the URL is present in the blocklist based on the first result and the second result. The user device is configured to check the sum of the first result and the second result, so only the user device learns the determination result. If the sum equals one, it means that the URL of the user is on the blocklist. Thus, the access to a website associated with the URL is blocked. In other words, if the sum equals one, the user device is configured to block access to a website associated with the URL. If the sum does not equal one, the user device is configured to access the website associated with the URL.

[0156] Each possible combination of any features from the implementations described above forms part of the present disclosure. It is to be understood that this description is intended to be illustrative, and not restrictive. Many other implementations will be apparent to those of skill in the art upon reading and understanding this description. Although the present disclosure has been described with reference to specific implementations, it will be recognized that the disclosure is not limited to the implementations described, but can be carried out with modification and alteration. Accordingly, the specification and drawings are to be regarded in an illustrative sense rather than a restrictive sense. The scope of protection should, therefore, be determined with reference to the claims, along with the full scope of equivalents to which the claims are entitled.

[0157] Aspects of the present disclosure are set out below. Each item below corresponds to an aspect of the disclosure, and each possible combination of any of the items corresponds to an aspect of the disclosure. When items are combined, two or more instances of an element with the same name are to be understood as referring to the same element, even if the element is introduced by the word “a” more than once. Each possible combination of any features from the items below with any features described above also forms part of the disclosure.

[0158] In one aspect, herein is provided a method for determining whether a URL is present in a blocklist, the method comprising: S400: storing the blocklist; S405: hashing each URL in the blocklist; S410: transmitting the blocklist with at least one hashed URL to a first service provider; and S415: transmitting the blocklist with at least one hashed URL to a second service provider.

[0159] In another aspect, herein is provided a method for determining whether a URL is present in the blocklist, the method comprising: obtaining, by an enclave, at least one blocklist; hashing, by the enclave, each URL in each blocklist of the at least one blocklist; transmitting the blocklist with at least one hashed URL to a first service provider; and transmitting the blocklist with at least one hashed URL to a second service provider.

[0160] In another aspect, herein is provided a system configured to implement a method for determining whether a URL is present in a blocklist, the system comprising a user device; at least two non-colluding servers; and an enclave and a blocklist provider.

[0161] In another aspect, herein is provided a system configured to implement a method for determining whether a URL is present in a blocklist, the system comprising a user device; five noncolluding servers; and enclave and / or a blocklist provider.

[0162] In another aspect, herein is provided a method for determining whether a URL is present in a blocklist, the method comprising: S110: transmitting, from the BP or an enclave to a first service provider (S1), the blocklist with hashed URLs; S115: transmitting, from the BP or an enclave to a second service provider (S2), the blocklist with hashed URLs; S120: encoding, by the S1 and S2 the blocklist; S125: encoding, by the S1 and S2, a filter based on the encoded blocklist; S140: hashing, by the user device, the URL; S145: categorizing, by the user device, the hashed URL as either not being on the blocklist or possibly being on the blocklist; S150: upon categorizing, by the user device, the hashed URL as possibly being on the blocklist: deriving, by the user device, from the hashed URL: at least one keyvalue pair, a first key and a second key; S155: transmitting, from the user device to S1, the at least one key-value pair and the first key; S160: transmitting, from the user device to S2, the at least one keyvalue pair and the second key; S165: querying, by S1, the encoded blocklist based on the at least one key-value pair; and based on the query selecting at least one hashed URL from the encoded blocklist and adding the at least one selected URL to a bucket; S170: querying, by S2, the encoded blocklist based on the at least one key-value pair; and based on the query selecting at least one hashed URL from the encoded blocklist and adding the at least one selected URL to the bucket; S175: generating, by S1, a first result by evaluating the first key over the bucket and transmitting the first result to the user device; S180: generating, by S2, a second result by evaluating the second key over the bucket and transmitting the second result to the user device; S190 determining whether the URL is present in the blocklist based on the first result and the second result. The method may further comprise encoding, by S1 or S2, a hashed URL, wherein the hashed URL was input by a user of the user device.

[0163] In another aspect, described herein is a method for determining whether a Uniform Resource Locator, URL, is present in a blocklist, the method comprising: receiving the blocklist; encoding the blocklist; receiving at least one key-value pair and a first key; querying the encoded blocklist based on the at least one key-value pair; based on the query selecting at least one hashed URL from the encoded blocklist and adding the at least one selected URL to a bucket; generating a first result by evaluating the first key over the bucket; and transmitting the first result to a user device.

[0164] In another aspect, described herein is a method for determining whether a Uniform Resource Locator, URL, is present in a blocklist, the method comprising: receiving the blocklist; encoding the blocklist; receiving at least one key-value pair and a second key; querying the encoded blocklist based on the at least one key-value pair; based on the query selecting at least one hashed URL from the encoded blocklist and adding the at least one selected URL to a bucket; generating a second result by evaluating the second key over the bucket; and transmitting the second result to a user device.

[0165] In another aspect, described herein is a computer-implemented method for determining whether a Uniform Resource Locator, URL, is present in a blocklist, the method comprising: hashing the URL; categorizing the hashed URL as either not being on the blocklist or possibly being on the blocklist; upon categorizing the hashed URL as possibly being on the blocklist: deriving from the hashed URL: at least one key-value pair, a first key and a second key; transmitting the at least one key-value pair and the first key to a first service provider; transmitting the at least one key-value pair and the second key to a second service provider; receiving, from the first service provider, a first result based on the at least one key-value pair and the first key; receiving, from the second service provider, a second result based on the at least one key-value pair and the second key; determining whether the URL is present in the blocklist based on the first result and the second result.

Claims

1. A computer-implemented method for determining whether a Uniform Resource Locator, URL, is present in a blocklist, the method comprising:hashing the URL;categorizing the hashed URL as either not being on the blocklist or possibly being on the blocklist;upon categorizing the hashed URL as possibly being on the blocklist:deriving from the hashed URL: at least one key-value pair, a first key and a second key;transmitting the at least one key-value pair and the first key to a first service provider;transmitting the at least one key-value pair and the second key to a second service provider;receiving, from the first service provider, a first result based on the at least one keyvalue pair and the first key;receiving, from the second service provider, a second result based on the at least one key-value pair and the second key;determining whether the URL is present in the blocklist based on the first result and the second result;wherein the at least one key-value pair comprises data representing sampled and indexed hashed URL;wherein the first key comprises first data partially representing the hashed URL;wherein the second key comprises second data partially representing the hashed URL; and wherein the first key and the second key together comprise data representing the hashed URL; wherein the hashed URL cannot be determined in isolation from either:the first key; orthe second key.

2. The method of claim 1,wherein categorizing the hashed URL as either not being on the blocklist or possibly being on the blocklist comprisesencoding the hashed URL;performing a local query on a user device using the encoded hashed URL with an encoded filter to categorize the hashed URL.

3. The method of claim 2, further comprising:obtaining the encoded filter from the first service provider or the second service provider;wherein the encoded filter is an encoded Bloom filter;wherein the encoded Bloom filter comprises a compressed blocklist.

4. The method of any one of the preceding claims,wherein deriving the at least one key-value pair from the hashed URL comprises sampling the hashed URL to generate the at least one key-value pair.

5. The method of any one of the preceding claims,wherein deriving the first key and the second key from the hashed URL comprises encoding the hashed URL to generate the first key and the second key.

6. The method of any one of the preceding claims,wherein determining whether the URL is present in the blocklist based on the first result and the second result comprises determining that the sum of the first result and the second result equals one.

7. A computer-implemented method for determining whether a Uniform Resource Locator, URL, is present in a blocklist, the method comprising:receiving the blocklist;encoding the blocklist;receiving at least one key-value pair and a first key from a user device;querying the encoded blocklist based on the at least one key-value pair;based on the query selecting at least one hashed URL from the encoded blocklist and adding the at least one selected URL to a bucket;generating a first result by evaluating the first key over the bucket; andtransmitting the first result to the user device for determining whether the URL is present in the blocklist based on the first result and a second result;wherein the bucket is shared with a second service provider;wherein the bucket comprises a sub-portion of the encoded blocklist;wherein the second result is generated by the second service provider by evaluating a second key over the bucket and wherein the second result is transmitted from the second provider to the user device;wherein the at least one key-value pair comprises data representing sampled and indexed hashed URL;wherein the first key comprises first data partially representing a hashed URL;wherein the second key comprises second data partially representing the hashed URL; andwherein the first key and the second key together comprise data representing the hashed URL;wherein the hashed URL cannot be determined in isolation from either:the first key; or the second key.

8. The method of claim 7, further comprisingencoding a filter based on the encoded blocklist; and optionally transmitting the encoded filter to the user device.

9. The method of claim 8,wherein encoding a filter based on the encoded blocklist comprises generating encrypted hashes for the encoded blocklist;encoding the filter with the encrypted hashes;wherein the encoded filter is an encoded Bloom filter.

10. The method of any one of the claims 7 to 9, wherein receiving the blocklist comprisesreceiving the blocklist through a secure and authenticated channel.

11. The method of any one of claims 7 to 10,wherein the blocklist comprises at least one hashed URL;and wherein encoding the blocklist comprisesindexing each hashed URL with a blocklist key;wherein the blocklist key is not known to the user device.

12. A data processing apparatus comprising means for carrying out the method of any preceding claim.

13. A computer-readable storage medium comprising the instructions which, when executed by a computer, cause the computer to carry out the method of any of claims 1 to 11.A

Citation Information

Patent Citations

  • ViewCN114844678AonEspacenetopensinnewtab