Method, device and system for constructing and deploying virtual multiple tasks, equipment and medium

Abstract

Distributed multi-task computing is very wide in application, typical applications comprise an out-of-order network (mix net) and other anonymous communication networks, and a plurality of sub-tasks deployed in a plurality of computing power facilities (servers) and internetworks thereof can be virtualized to one or a few computing power facilities. Limited computing power resources are used as much as possible, and intercommunication is carried out on an internal communication channel. A plurality of sub-tasks deployed on the same computing power facility can be isolated from one another through the trusted execution environment to protect respective privacy security. If necessary, remote attestation functions of some trusted execution environments can be used to verify the correct execution of each sub-task, for example to ensure that data transmitted in the anonymous communication network is not deleted. The structure and the function of the computing power facility bearing the multiple sub-tasks can be modified and expanded to adapt to different application scenes.

Description

Technical Field

[0001] This application involves cross-domain complex computing in multiple locations, multiple sources, multiple clouds or multiple servers, exploring methods and devices for virtualizing multiple computing tasks in the same computing facility; a multi-task system that can virtualize multiple routing nodes in an anonymous communication network in one or a few servers; and solving the information security issues therein, reducing trust assumptions through a trusted execution environment and ensuring data privacy and operation verifiability of each virtual task; such methods can be modified and expanded to support other applications where multiple parties and multiple facilities jointly complete computing and communication tasks. Background Art

[0002] With the development of computing devices and communication networks, complex computing tasks increasingly use diverse computing power from the end, edge, cloud, and network, and multiple computing facilities (such as multiple cloud servers or network communication nodes) connected by communication networks jointly complete computing tasks. This method of using multi-source computing power has certainly promoted the development of computing power networks and big data applications, but it has also increased the cost of using computing power facilities and network communications, often occupying multiple paid external servers and low-speed, high-priced external channels.

[0003] For example, there is an anonymous communication network called a mix net, which uses multiple external network routing nodes (servers) to provide multiple disorder operations to disrupt the batch data sent to a recipient, ensuring that they cannot be traced back to the data provider, thereby protecting their identity and address and other personal privacy information. For applications, although the use of multiple computing facilities will increase security redundancy and at least appear to be more conducive to data privacy protection, the use of multiple external communication servers and the external network bandwidth between them is a high price, not to mention that the network control policies of some countries may not allow such external network anonymous services. Another commonly used anonymous communication network is the onion routing network, which also has similar weaknesses, and the corresponding invention content will be covered later.

[0004] As a typical application, the disordered network reflects the characteristics of the scenario of using external computing power of (multiple) independent servers. Its technical background is briefly introduced as follows. Data privacy protection laws such as GDPR require that data privacy not only includes the data content itself, but also involves personal information such as the source of the data and the identity of the data provider. As an anonymous communication network, the purpose of the disordered network is not limited to protecting the data content itself, but also to prevent any data content from being traced back to the data owner after leakage, thereby protecting their personal privacy information such as identity and address. The implementation method of the disordered network is that multiple routing nodes in the network sequentially reorder (permutate) and re-encrypt (re-encrypt) or partially decrypt (partially) the encrypted data of multiple network users (data owners). decrypt), and the work of each routing node is called shuffling. The last routing node outputs all the data that have been shuffled multiple times, so as long as at least one routing node does not disclose the permutation it uses, the original order of all the data can be guaranteed to be unrecoverable. Generally speaking, shuffled networks are verifiable, that is, each routing node must be able to prove that its operation only changes the order of batch data without deleting their content. Specific users or third-party bystanders can verify the proof of the routing node to confirm the correctness of the shuffled network. During the proof process, the permutation and other secrets (such as decryption keys) used by the routing node must be kept confidential and cannot be leaked. Existing shuffled network solutions use multiple external network servers as routing nodes. In an external untrusted environment, they must prove that their respective shuffling operations are correct; this proof is usually high-cost and low-efficiency (for example, using complex zero-knowledge proofs), which is much more expensive than the shuffling operation itself in terms of computing and communication, and often becomes the efficiency bottleneck of the shuffled network and restricts its application in big data low-latency scenarios. Summary of the invention Technical issues

[0005] In the out-of-order network solution, when the computing power of the receiver or its receiving port is sufficient to complete all the reordering operations, it is a simple, efficient and low-cost solution for its own internal server to provide the computing power resources and communication pipelines required for the reordering operations. Of course, security should not be sacrificed. On the one hand, internal communication reduces the attack surface and is beneficial to communication security; but on the other hand, reducing the security redundancy of multiple distributed servers means that alternative means are needed to ensure that multiple reordering operations on the same server can be safely isolated from each other and ensure their respective information privacy. At the same time, the correctness proof of the reordering operation, as an indispensable security guarantee, cannot be omitted but is still a performance bottleneck.

[0006] In other scenarios where computing resources on external networks are used to complete multi-task computing, such as onion routing networks, application systems can also virtualize external servers (computing resources) and the network communications between them into tasks (applications, processes, threads or containers) on an internal server and the internal communications between them. By transforming and expanding the functions of servers to support more applications, their computing costs will be reduced and communication efficiency will be improved. However, when running multiple tasks on a server, attention must be paid to security isolation to protect confidential information. At the same time, there are also challenges in using new methods to improve efficiency in scenarios where the correctness of operations needs to be verified. Solution

[0007] Taking the out-of-order network as an example, all external routing servers can be virtualized to the receiver's system or its communication port, such as on the receiver's access device (such as a data receiving server), each running as an independent sub-task, and the required communication between them can also be achieved by a fast internal channel. At this time, two points should be noted. ● The privacy security guarantee of the original disordered network lies in: unless all routing nodes leak the secret information used for their disordered operations, the distributed multiparty n-out-of-n redundant security mechanism can ensure that the information transmitted through the disordered network is anonymous and untraceable. In other words, as long as at least one routing server does not leak privacy, the disordered network can achieve anonymity. After virtualization, the disordered network is realized with only one server, which becomes the only guarantee for the security of the disordered network. When the owner who controls it (such as the receiver or its system administrator) or the attacker who cracks it (such as an intruding hacker) masters all its virtualization tasks, it can extract all the secret information used for the disordered operations, thereby cracking the anonymity protection of the disordered network to track the privacy information of all data providers. This security weakening is unacceptable and needs to be addressed with appropriate countermeasures. ●As mentioned above, the correctness proof and verification of multiple reordering operations is a necessary security guarantee and also a bottleneck of efficiency that consumes a lot of computing and communication. When they are concentrated on one server, it will obviously form a greater computing pressure, which is likely to conflict with the premise assumption of the virtualization solution required by the invention content 0005: "When the computing power of the receiver is sufficient to complete all reordering operations." Therefore, the traditional correctness proof and verification methods need a simpler and more efficient alternative.

[0008] The virtualized out-of-order network needs to solve the two requirements in the invention content 0007. A feasible solution is to deploy a special multi-task isolation system on the receiving party's data receiving server, such as a trusted execution environment TEE (Trusted Executive Environment). This is a security protection technology based on trusted hardware (CPU and other computing chips with multi-task isolation protection capabilities). It can isolate the computing environment (data, etc.) of multiple computing tasks (usually referred to as TA, such as processes, etc.) running in a computing facility (server), so that neither can obtain the secret information of other tasks. At the same time, TEE also provides a technology called remote attestation, which can verify that the operation of each task is correct and does not deviate from the original design. This remote attestation does not require complex and inefficient tedious operations (such as zero-knowledge proof commonly used in traditional disordered networks), and usually only requires a few summary and digital signature operations, which is very efficient. Common TEE systems include SGX and TDX based on INTEL chips, SEV based on AMD chips, both of which are based on X86 architecture, and TRUSTZONE and CCA (Confidential Compute Architecture) under ARM architecture. Various computing architectures already have or are developing and improving TEE functions.

[0009] The biggest feature of TEE is that the security foundation only needs to be built on trusted hardware (chips with TEE functions). At most, trust is attributed to a very small number of chip manufacturers during remote proof, without the need to trust various server administrators or service providers (even they cannot access the private data of TEE tasks). This avoids the problem of reduced security redundancy and stronger trust assumptions when multiple distributed servers are virtualized to one server. Even if there is only one server, whether it is an internal system administrator or an external attacker, even if they control the server and obtain unlimited permissions, they cannot steal the protected secret information in TEE.

[0010] In virtualized solutions, TEE's remote attestation function can also help applications that need to prove and verify the correctness of operations; this is very important in many scenarios (disordered networks are used in many applications that require multi-task collaborative computing that does not need to trust each other, such as network elections that require multiple tasks to cooperate to process votes); generally speaking, multi-tasks (whether on multiple servers or on a single server) are often used to complete multi-party computing, and the scenarios where there is no trust between each other and the need to verify the correctness of each other's operations are very common.

[0011] A virtualized out-of-order network is reflected as multiple subtasks (such as protected TEE tasks) on the receiving party's data receiving server. Each task corresponds to a routing node to complete a reordering operation. TEE will allocate its own storage space to each task, and the access control based on the trusted root of the TEE chip ensures that only the corresponding TEE task can access its own space. Many TEE systems also provide memory encryption functions, which can encrypt the memory information of each task and even provide them with different encryption keys, which further strengthens data isolation and ensures that the reordering operation details of each virtual routing node are unknown to the outside world. Even on a single server, it can still be guaranteed that unless all TEE routing tasks leak the secret information used for their own reordering operations, the distributed multiparty n-out-of-n redundant security mechanism can still be implemented, thereby ensuring that the information transmitted through the reordering network is anonymous and untraceable. In other words, as long as at least one TEE routing task does not leak privacy, the reordering network can be anonymized. The reordering n-out-of-n redundant security mechanism of the reordering network can continue to be effective in virtualization.

[0012] Invention content 0011 The virtualized disordered network described in the invention, in addition to each sub-task completing a reordering operation, the communication between them (the output of the previous reordering operation is passed to the next reordering operation as input) is carried out on the channel inside the data receiving server. There are two options for verifying the correctness of the reordering operation. The traditional proof (each routing task proves that its reordering operation is correct but does not disclose any secret information) and verification (selected by a specific stakeholder or any third party) can of course be inherited; any other verification method (such as self-verification led by the data provider) can also be adopted for compatibility; but it can also rely only on the remote proof function of the TEE system to prove to the participants or bystanders who care about the correctness that the execution code of each routing task is a correct reordering operation (for example, the TEE remote proof module measures the code of the routing operation to form its collision-resistent function summary, and after comparing and confirming the consistency with the standard summary, the measurement result is digitally signed based on the TEE trusted root). In this way, the communication and computing efficiency of the disordered network can be greatly improved, and its correctness is still verifiable.

[0013] After the processes described in Invention Content 0011 and Invention Content 0012, the input data of the same batch have all undergone the reordering operations of all virtual routing tasks, and these reordering operations can also be verified to be correct, and the permutation (permutation) and other secret information in the reordering are not leaked. In other words, in the virtual environment of a server, a non-disclosure and verifiable disordered network is still achieved. These encrypted data that are disrupted and no longer traceable can be processed as anonymous data. Information privacy protection laws and regulations, including GDPR, generally recognize that anonymized data will not be leaked Personal privacy can be processed in a non-protected environment. Therefore, this data can be decrypted and used, and can also be used in various applications outside the TEE environment. For example, in the most common application of random networks, online elections, votes are anonymized through a random network and then counted in a public environment. Traditionally, all voters encrypt their ballots, and after their identities are verified as legitimate, these encrypted ballots are submitted to the random network, and multiple routing servers in the random network perform reordering operations to disrupt the order of the ballots. The shuffled ballots are then handed over to the counting server to decrypt the ballots and count the election results. Voters can vote with their real names. The votes can be decrypted into plain text to ensure that they are legitimate and have not voted repeatedly. The votes can also be decrypted into plain text for direct statistics, which is convenient and fast. The encryption mechanism and the random network in the middle can ensure that each voter's secret information (vote) will not be linked to his identity and will not disclose his privacy. As described in Background Technology 0002 and Background Technology 0003, using multiple external routing servers to shuffle the ballots and using external network communications between them is high-cost and low-efficiency; as described in Background Technology 0004, each routing server must prove that there is no tampering with or missing ballots and cannot disclose any secret information, and must also use complex proof and verification technology It will become an efficiency bottleneck. If the virtualization technology of the present invention is used, multiple routing servers that shuffle the votes are virtualized to the counting server or its receiving port, and become the order-changing tasks on it; and a secure multi-task management system such as TEE is set on the computing power facility to place each order-changing task in it for isolation and protection, the server usage cost and communication cost can be reduced without sacrificing security. If the remote proof function of TEE is used to verify the correctness of the operation of each virtualized routing task and avoid other complex and inefficient proof methods, electronic elections based on virtualized disordered networks can further improve efficiency and avoid bottlenecks.

[0014] The above solution of virtualizing multiple servers to the same computing facility and protecting them with a trusted execution environment such as TEE when necessary is also applicable to many other scenarios. Taking the anonymous network as an example, another anonymous communication network, the Onion Routing network (TOR), also relies on the routing of multiple external servers, and the communication is also on the external network. Compared with the disordered network, the characteristic of onion routing is that each anonymized data is encrypted through multiple layers to form a data packet, and then multiple routing servers in the external network relay the data and decrypt it layer by layer. The key is that the transmission path of each data packet can be different, that is, different routing server groups can be used, that is, each data transmission can be transmitted through an independent secret path instead of forming a common batch with other data through a public and identical path. However, the anonymity principle of onion routing is still: unless all routing nodes leak their own secret routing operations, the distributed multiparty n-out-of-n redundant security mechanism can ensure that the input information through the anonymous network is anonymous and untraceable. Since each routing server does not operate publicly, it is even necessary to keep the path secret to prevent path analysis attacks (traffic According to the analysis, the onion routing network usually does not require routing nodes to provide proof of the correctness of their operations; and once someone wants to verify the correctness of routing operations, it will be more difficult than in a disordered network. In addition to being more difficult to achieve verifiability, the onion routing network, like the disordered network, consumes too much energy to occupy multiple external servers and use a large amount of external network communications, which affects efficiency. When the multiple routing servers of the onion routing are virtualized into multiple sub-tasks on a computing power facility, server resources can also be saved and low-speed and high-cost communication connections on the external network can be avoided. In addition, by deploying the TEE mechanism on this computing power facility, those virtualized routing tasks can also be isolated and protected to ensure their data privacy. The remote proof function provided by TEE can also verify the correctness of all routing tasks to ensure that no data has been deleted or modified.

[0015] The centralized virtualization implementation of the disordered network and onion routing network described in Invention Contents 0007 to 0014 can also adopt a compromise solution. Multiple routing servers (nodes) of traditional anonymous routing can be virtualized to a few servers instead of one server. For example, 9 traditional routing nodes that undertake one-stop routing tasks can be virtualized to 3 centralized servers, each server has 3 routing tasks. For another example, 5 traditional routing nodes that undertake one-stop routing tasks can be virtualized to 2 centralized servers, one server has 3 routing tasks, and the other server has 2 routing tasks. Each server that concentrates multiple routing tasks can deploy a multi-task management system or even TEE, isolate and protect the routing tasks on it when necessary, and use remote proof to ensure that the correctness of the routing operation can be verified. In this way, security and privacy can be doubly protected by TEE and multi-server redundancy, and sometimes a certain ideal balance can be achieved between security and efficiency.

[0016] In many other scenarios, there are also common situations where high-security computing or communication applications need to distribute tasks to a large number of servers to be completed jointly by them and may rely on n-out-of-n redundancy to ensure data privacy. At this time, the virtualization method described in Invention Contents 0007 to 0015 can concentrate too many distributed tasks on a few (or even one) servers by expanding applications (for example, when there is no anonymous communication task, the anonymous communication server is transformed to perform other operations or a dedicated server with the same virtualization setting is provided for special applications). On the one hand, the communication cost of the external network is saved and the network attack surface is reduced. On the other hand, multi-task management systems such as TEE deployed on the server can be used to isolate and protect data privacy, and no longer rely entirely on the n-out-of-n redundant protection mechanism of multiple servers. Each virtualized distributed task can also rely on the remote proof of TEE to realize the correctness verification without leaking data privacy, and achieve verifiability that was not easy to achieve efficiently in the past. The association between multiple distributed tasks can be serial (one after another) as in the disordered network and onion routing network, or concurrent (can be performed simultaneously) or have a more complex logical relationship. In order to further protect data privacy, the communication between each sub-task can also be encrypted (even multi-layer encryption) as in the disordered network and onion routing network, and a similar decryption mechanism can be used to obtain plaintext data (such as calculation results) when necessary.

[0017] There are endless ways to run multiple tasks on a server and isolate and protect their private data and operations when necessary, and they are not limited to solutions that call themselves virtualization or TEE, and their correctness verification methods are not necessarily remote proof. The present invention covers similar technologies. As long as multiple tasks can be run on a computing power facility, the anonymous communication network or other multi-task applications are essentially virtualized to one or a few computing power facilities. No matter how they are named, they are valid. When each sub-task needs security protection, it can be based on the multi-task isolation provided by the computing power equipment, or on other methods such as encryption protection and non-decryption calculation. The correctness verification of each sub-task can be based on hardware trust similar to remote proof, or other proof technologies such as zero-knowledge proof can be used, or even omitted when there is no verification requirement. Advantageous Effects of the Invention

[0018] One of the main functions of virtualized anonymous communication networks (including disordered networks and onion routing networks) is data anonymity. As long as at least one virtualized routing node does not disclose the routing operation it uses (for example, by revealing the key or other secrets it uses), the data can no longer be traced back to the corresponding data owner, thereby achieving data anonymization; and the virtualized anonymous communication network using the present invention can improve efficiency and performance and ensure privacy and verifiability. Even if the data is used in plain text and its content is known to a certain extent, because the source cannot be traced, it can avoid damaging the personal privacy of the data owner. Information protection laws and regulations, including GDPR, generally recognize that anonymized data will not disclose personal privacy. Therefore, in many anonymous applications, users are often allowed to submit data through disordered networks to protect their privacy, such as in the following scenarios. ● In an electronic auction, bidders may not want to reveal their bids, which is the so-called sealed-bid auction scenario. If the content of the bid must be in plain text to select the winning bidder, at least bidders who care about personal privacy do not want their bids to be traced back to themselves. In this case, anonymous communication networks can be used to anonymize the bidders' encrypted bids and then decrypt them to determine the winning bid; the winning bidder can prove that the bid came from himself and then obtain the auction item, while other bidders can protect their bids from being disclosed. ●Many journals and academic conferences require that all manuscripts be reviewed anonymously in blind review, so anonymous communication networks are suitable for anonymizing manuscripts. When conducting online questionnaire surveys (such as various polls), participants often need to be anonymous. This is because participants want to remain anonymous for privacy reasons, and the organizers will encourage participants to submit their answers anonymously to ensure the objectivity and fairness of the results. At this time, the electronic answers can be encrypted first, anonymized through an anonymous communication network, and then decrypted and counted, so that anonymous questionnaires can be implemented. Similar applications are popular online voting activities for the public in society (such as selecting excellent works or popular public figures). When anonymous forms are required, anonymous communication networks are also needed to anonymize votes. Various statistical analyses and various operations of anonymous inputs can be built on anonymous communication networks.

[0019] The more extensive application field of virtualized anonymous communication networks is electronic elections, including corporate boards, student unions, labor unions, industry associations, sports associations and other civil groups’ leadership bodies, as well as representatives of the people or heads of government. In electronic elections, voters submit digital ballots to the election agency through electronic communication networks, which calculate the election results. Anonymous communication networks can be used to process a large number of encrypted ballots, thereby protecting voters’ voting privacy and correctly counting election results. After anonymization, the encrypted ballots can no longer be traced, and at this time, they can be decrypted to count the election results. As long as at least one routing node in the anonymous communication network does not leak the permutation used, the privacy of the ballot can be guaranteed. The routing nodes of the anonymous communication network used in electronic elections can be multiple distributed workstations set up by the election agency, or they can be controlled separately by the election agency and candidates (party) representatives and independent observers. When each workstation uses an independent external server and they use external network communication, the cost is high and the exposed attack surface is large; after virtualizing them to one or a few servers, the computing cost can be reduced, the communication efficiency can be improved, the exposure surface can be reduced, and internal trusted technologies such as TEE isolation and remote attestation can be used to ensure privacy security and verifiability.

[0020] Virtualized anonymous communication networks can also be used in other situations where the user's identity needs to be protected, such as secretly using cloud services or accessing network resources, invisible browsing, anonymous email, etc., and the same optimization effect can be achieved. Through more extensive transformation, multi-source computing power applications on independent servers on multiple external networks can be virtualized to one or a few computing power facilities (servers, such as anonymous communication servers with expanded functions) and rely more on internal communications, which can save costs and improve efficiency; using functions such as data isolation and remote proof such as TEE can ensure their privacy security and verifiability. As emphasized in the invention content 0016 and the drawings of the specification, the relationship between the virtualized tasks can be various, not limited to serial operations like anonymized data. As described in the invention content 0017, the virtualized environment and its isolation protection technology and correctness verification technology may also have various names and choices, but they must all support multi-task operation and meet security requirements when necessary. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 -Multi-server anonymous communication network and multi-tasking virtualization for extended applications.

Claims

1. A method for virtualizing a multi-server disordered network (a typical anonymous communication network used to anonymize the data passing through it so that it cannot be traced). Each routing operation of the disordered network is traditionally undertaken by a routing node on an external communication network, usually requiring an independent external routing server and an external channel connecting it; in the future, it can be virtualized into a routing task and deployed on an access device (such as a data receiving server) of the data receiver. Multiple routing tasks can be deployed on this access device (server) (forming multiple sub-tasks on a computing device), which are connected to each other through internal channels and managed through multi-task scheduling. Some multi-task management mechanisms, such as the trusted execution environment TEE, can isolate and protect the privacy data and secret operations of each routing sub-task, thereby ensuring the n-out-of-n redundant security of the disordered network: there is no need to assume the trustworthiness or security of any server or its administrator, as long as at least one routing sub-task is not leaked (for example, when its TEE isolation protection is effective), it can be ensured that the data passing through the virtualized disordered network is anonymized and cannot be traced.

2. A verification method based on remote attestation can verify the correctness of virtualized out-of-order networks. In situations where the correctness of an out-of-order network needs to be verified (correctness means that the data passing through the out-of-order network has not been deleted or modified), the virtualized out-of-order network does not necessarily have to rely on the traditional correctness proof and verification mechanism, but can also provide a remote attestation mechanism for a multi-task management module (such as TEE) on the device where it is deployed. Moreover, it is not necessary to implement the verifiability of operations on each routing server of the traditional out-of-order network separately, but only to verify the correctness of all subtasks on that access device. Remote attestation can measure the execution code of each routing subtask and compare it with the standard code to verify its correctness, and its fundamental trust assumption does not lie in the server, the tasks on it, or their controllers, but falls on the computing chip through digital signatures. Specifically, because the measurement results are authenticated by digital signatures, and the signature private key is attributed to the built-in key of the server's main computing chip, as long as the computing chip is trustworthy (or the chip manufacturer is trustworthy and provides a qualified and compliant computing chip), it is possible to verify whether the operation of the virtualized out-of-order network is correct.

3. A more general method for virtualizing various anonymous communication networks. Any anonymous communication network, such as the onion routing network, can also be virtualized onto a communication server as long as it also uses multiple external routing nodes to relay the data that needs to be anonymized, which can not only save server resources, but also transmit data on a faster and more secure internal channel. Similarly, each routing node is virtualized as a sub-task on the communication server; and a multi-task management system is set up on the communication server, which can use hardware-trusted isolation measures such as TEE to protect the data privacy and secret operations of each sub-task. In addition to traditional correctness verification methods, this security mechanism can also include verification methods such as remote proof to verify the correctness of each routing sub-task on the communication server.

4. A method for transforming and expanding virtualized servers. A server that deploys virtualized multi-tasks can be adjusted by enhancing computing power or expanding functions to support multi-task computing scenarios of varying scales and complexities and carry their multiple sub-tasks. For example, for the encryption and decryption operations of asymmetric cryptography widely used in disordered networks, the server can be upgraded to use a main computing chip that supports hardware encryption and decryption or auxiliary software and hardware encryption and decryption modules to support scenarios with large data volumes. For another example, AI computing chips such as GPU or NPU can be deployed on the server to cooperate with the main computing chip to process more complex sub-tasks (such as intelligently identifying illegal users in anonymous communication networks and deanonymizing them or supporting other computing and communication tasks). Of course, other more suitable servers can also be used to replace it to support various multi-task computing power requirements. In this way, many distributed applications that originally needed to be deployed on multiple computing facilities (servers) can be virtualized into multiple sub-tasks on one computing facility (server). In scenarios where data privacy protection is required, a trusted environment such as TEE can be set up for this computing facility to isolate and protect each sub-task. In scenarios where verifiability needs to be met, remote proof can be implemented in the trusted environment of this computing facility to verify the correct execution of each sub-task.

5. A method for flexibly using a virtualized server. The virtualized computing power described in claims 1 to 4 does not necessarily have to be implemented on one computing power facility. Multiple computing power facilities can also be used to carry virtualized sub-tasks, and multiple but not necessarily all sub-tasks are deployed on each computing power facility. Multiple sub-tasks within the same computing power facility use internal channels to communicate with each other and are protected and verifiable in the same trusted environment (such as TEE); sub-tasks on different computing power facilities use external network communications, and the different trusted environments in which they are located can independently provide protection and verifiability for their own sub-tasks, or they can be combined to provide a unified security protection mechanism. Of course, the number of computing power facilities commonly used is less than the number of sub-tasks. This method can be regarded as a compromise between the traditional one-task-one-facility solution and the multi-task-one-facility solution described in claims 1 to 4. According to the characteristics and scale of specific tasks and the capabilities of computing power facilities, a few computing power facilities are used to carry most sub-tasks to achieve limited distributed deployment, which can not only optimize computing power resources and communication pipelines, but also flexibly use existing computing power facilities and communication networks.

Images