Trusted data management method for multi-party evolutionary game in high dynamic environment

Abstract

The invention discloses a credible data management method for a multi-party evolution game in a high dynamic environment, and relates to a multi-party participation dynamic evolution game safety deduplication method and a cross-domain cooperation system facing an air-space-ground-sea fusion network. The problems of trust loss and benefit conflict among TPS, CSP, edge nodes and users in existing cross-domain cooperation are solved. According to the method, different strategy combinations are obtained, the incomes of users, CSP and TPS are analyzed, and corresponding income values under the different strategy combinations are obtained; solving and analyzing the corresponding income values, obtaining a copy dynamic equation of the user, the CSP and the TPS, solving characteristic values under different equilibrium points, and judging the stability of the equilibrium points according to the positive and negative characteristics of the characteristic values; and determining an evolution stable strategy, and realizing data deduplication management and the like. The method also has good adaptability and popularization value in the heterogeneous environments of an emerging space-air-ground-sea converged communication system, an intelligent manufacturing system, an edge node of the Internet of Vehicles and the like.

Description

Technical Field

[0001] This invention relates to a dynamic evolutionary game-theoretic secure deduplication method and cross-domain collaboration system for multi-party participation in integrated air-space-ground-sea networks. For highly dynamic, multi-participant ubiquitous network environments (including airborne satellite nodes, space-based UAV platforms, ground-based vehicle-to-everything (V2X) terminals, and marine edge computing nodes), it addresses the issues of trust deficiency and conflict of interest among third-party service providers (TPS), cloud service providers (CSP), edge nodes, and users in cross-domain collaboration by constructing a reputation-based dynamic evolutionary game-theoretic model. Background Technology

[0002] With the rapid development of intelligent transportation, low-Earth orbit satellite communication, and ubiquitous network technologies, the Internet of Vehicles (IoV) and satellite networks are gradually merging to build an integrated information perception and transmission system encompassing air, space, and sea. In this multi-dimensional collaborative network, vehicle-mounted terminals, roadside units (RSUs), low-Earth orbit satellites, unmanned aerial vehicles (UAVs), edge servers, and cloud platforms constitute multiple layers of heterogeneous nodes, participating in data collection, processing, and forwarding. Especially in typical scenarios such as autonomous driving, collaborative perception, traffic guidance, and disaster emergency response, various nodes frequently generate and interact with multimodal perception data, including location, speed, images, radar, and video. This data exhibits characteristics of high-frequency updates, redundancy, and cross-domain transmission within the network.

[0003] Against this backdrop, effectively achieving data deduplication has become a key challenge restricting system performance. On the one hand, duplicate data transmission and storage not only waste bandwidth resources but also significantly increase the load on edge nodes and satellite links. On the other hand, while existing secure deduplication technologies (such as deduplication methods based on random message lock encryption) achieve redundancy identification while ensuring privacy, they still suffer from high computational and storage overhead and limited efficiency. To reduce matching complexity, researchers have proposed a scheme based on autoencoder models to generate digest labels. By comparing the similarity of digest labels, unnecessary fingerprint comparisons are filtered out in advance, significantly improving the efficiency of secure deduplication. However, existing schemes typically require users or edge devices to train the autoencoder locally, resulting in high computational and storage burdens. To alleviate the burden on terminals, some studies have introduced TPS to assist in managing digest models and label indexes. However, TPS itself may lack sufficient credibility guarantees. Once digest tampering, data leakage, or coordinated attacks occur, it will seriously threaten the overall security of the system and user privacy.

[0004] Furthermore, in highly dynamic environments with multiple participants, incomplete information, and time-varying node strategies, the relationships of interest among users, TPS, CSP, and edge nodes are complex. When choosing between autonomous training, relying on TPS, or delegating deduplication tasks to CSP, each party is often influenced by current gains, the behavior of others, and long-term trust mechanisms. Therefore, a mechanism capable of dynamically modeling the multi-party behavioral choices and trust evolution process is urgently needed.

[0005] In summary, how to achieve a reliable, secure, and efficient data deduplication mechanism in a highly dynamic, heterogeneous, and cross-domain collaborative air-space-ground-sea network environment, and how to coordinate the policy evolution and trust relationships among multiple participants, has become a key technical problem that urgently needs to be solved. Summary of the Invention

[0006] To address the issues of trust deficiency and conflict of interest among third-party service providers (TPS), cloud service providers (CSP), edge nodes, and users in cross-domain collaboration, this invention provides a trusted data management method for multi-party evolutionary game in a highly dynamic environment.

[0007] A reliable data management method for multi-party evolutionary games in highly dynamic environments, which is implemented by the following steps:

[0008] Step 1: Make assumptions about the three entities, namely, user, CSP, and TPS, and obtain different combinations of strategies;

[0009] Step 2: Analyze the revenue of users, CSPs, and TPS based on different strategy combinations to obtain the corresponding revenue values ​​under different strategy combinations;

[0010] Step 3: Solve and analyze the corresponding revenue values ​​under different strategy combinations obtained in Step 2 to obtain the replication dynamic equations for users, CSP, and TPS.

[0011] Step 4: Construct the Jacobian matrix based on the replication dynamic equations of users, CSP, and TPS obtained in Step 3;

[0012] Step 5: Based on the Jacobian matrix from Step 4, obtain the eigenvalues ​​at different equilibrium points, and determine the stability of the equilibrium points based on the sign of the eigenvalues; determine the evolutionary stabilization strategy to achieve data deduplication management.

[0013] The beneficial effects of this invention are:

[0014] This invention explores the possibility of untrustworthy behavior in the TPS (Transaction Processing System) during secure deduplication. A lack of integrity in the TPS could lead to data leakage or tampering, threatening data security. Therefore, this invention introduces a dynamic evolutionary game theory model. Through establishing mutual trust mechanisms and rationally allocating costs and benefits during a multi-party game, it encourages stable cooperation among all parties, thereby achieving dual guarantees of system security and processing efficiency.

[0015] First, this invention constructs a dynamic evolutionary game model involving users, CSPs, and TPS. Under conditions of incomplete information and bounded rationality, it systematically analyzes the costs, benefits, and risks of each party under different strategy choices. Through Jacobi matrix and equilibrium point analysis, stable equilibrium points under different strategy combinations are identified, particularly the ideal cooperative equilibrium point K8(1,1,1), which represents the model's stability when each party chooses its optimal strategy. Simultaneously, the model demonstrates how users, CSPs, and TPS continuously adjust their strategies based on changes in their own payouts and the behavior of other parties, reflecting the complex interactions and dynamic process of strategy selection within the game model. This provides predictions for the long-term evolutionary trend of the game model and offers a reference for strategy selection by each party.

[0016] In terms of security, this invention addresses the risk of data leakage or tampering caused by the lack of integrity of third-party service providers (TPS) during secure deduplication. It innovatively introduces dynamic evolutionary game theory to construct a dynamic evolutionary game model for secure deduplication involving users, CSPs, and TPS. By establishing a replication dynamic equation and using the Jacobian matrix for equilibrium stability analysis, the evolutionary patterns of strategies among the participants under different parameter conditions are studied in depth. The results show that when the model converges to the ideal equilibrium point K8(1,1,1), an optimal secure collaboration mode is formed. That is, users actively adopt professional secure deduplication services provided by CSPs and TPSs, rather than training their own models; CSPs actively establish strategic partnerships with TPSs to jointly improve the security and efficiency of the deduplication process; and TPSs consciously adopt integrity strategies and strictly implement security audit and verification services. This equilibrium effectively solves the trust deficiency problem in traditional secure deduplication. Through the design of the game mechanism, it achieves self-reinforcement and collaborative optimization of the three-party security strategies, providing a new paradigm for data security in cloud storage environments.

[0017] The method described in this invention is not only applicable to traditional cloud storage systems, but also has good adaptability and promotion value in emerging heterogeneous environments such as air-space-ground-sea integrated communication systems, intelligent manufacturing systems, and vehicle-to-everything (V2X) edge nodes. In particular, against the backdrop of increasingly prominent needs for trusted sharing and collaborative processing of multi-source heterogeneous data, the dynamic evolutionary game model it constructs can provide decision support and security assurance for related systems. Attached Figure Description

[0018] Figure 1 The flowchart is a process for a trusted data management method for multi-party evolutionary game in a highly dynamic environment, as described in this invention.

[0019] Figure 2 This is a schematic diagram of the trusted data management method for multi-party evolutionary game in a highly dynamic environment as described in this invention.

[0020] Figure 3 This is a three-way evolution trend diagram when all feature values ​​are less than 0 in the trusted data management method for multi-party evolutionary games in a highly dynamic environment as described in this invention.

[0021] Figure 4 This is a trend diagram of user strategy changes under different CSP strategies and TPS integrity behavior in the trusted data management method for multi-party evolutionary game in a highly dynamic environment described in this invention; wherein, (a) is a two-dimensional planar curve effect diagram, and (b) is a three-dimensional curved surface effect diagram.

[0022] Figure 5 This is a trend diagram of CSP strategy changes under different user strategies and TPS integrity behavior in the trusted data management method for multi-party evolutionary game in a highly dynamic environment described in this invention; wherein, (a) is a two-dimensional planar curve effect diagram, and (b) is a three-dimensional curved surface effect diagram.

[0023] Figure 6 This is a trend diagram of TPS integrity behavior changes under different CSP strategies and user strategies in the trusted data management method for multi-party evolutionary games in a highly dynamic environment described in this invention; wherein, (a) is a two-dimensional planar curve effect diagram, and (b) is a three-dimensional curved surface effect diagram. Detailed Implementation

[0024] Combination Figures 1 to 6 This embodiment describes a trusted data management method for multi-party evolutionary game in a highly dynamic environment, which includes three entities: user, CSP, and TPS.

[0025] like Figure 1 and Figure 2 As shown, users want to obtain high-quality data processing services at the lowest cost while ensuring data security.

[0026] CSPs need to select TPS reasonably and control costs while ensuring data security in order to maximize their own interests.

[0027] TPS: The goal is to minimize one's own costs while receiving sufficient compensation.

[0028] The specific implementation process of the method described in this embodiment is as follows:

[0029] Step 1: Make strategic assumptions for the three entities in this method respectively;

[0030] User's assumed strategy: When the user processes data, there are two options: 1) Self-train an autoencoder model, which requires a high cost; 2) Do not train by oneself. If the CSP provides a TPS solution, the user needs to bear the TPS usage cost.

[0031] CSP's assumed strategy: When the CSP provides services, there are two strategies: 1) Use TPS, which requires bearing costs such as cooperation negotiation, contract signing, and service supervision; 2) Use random information lock encryption (R-MLE), which requires more investment in technology R & D and maintenance costs to ensure service effectiveness.

[0032] TPS's assumed strategy: When TPS provides services, it can choose an honest or dishonest strategy: 1) Honest strategy: It needs to invest in technical equipment, staff training, and data security management and other costs to ensure service quality and security. 2) Dishonest strategy: In the short term, it may profit through illegal operations such as cutting security investment, but it will bring serious consequences in the long term.

[0033] In this embodiment, based on the above assumed conditions, symbols and formulas are used to strictly define the above logic to obtain different strategy combinations;

[0034] User's choice: The cost when the user self-trains the autoencoder model is C U1 , at this time, the potential benefit brought by the user to himself is R1. The cost when not self-training the autoencoder model is C U2 (C U2 < C U1 - R1). If at this time the CSP selects a TPS solution for the user, the user also has the TPS usage cost C U3 . At this time, if TPS is dishonest, TPS brings the risk cost C U4 of data leakage to the user, and the CSP will provide the user with compensation R2 (R2 < C U4 ).

[0035] CSP's strategy: The cost when the CSP uses TPS is C C1 , the income obtained from the user is R3, and the remuneration given to TPS is F1. If TPS is honest, the CSP gives TPS a reward R4, and TPS brings potential benefit E1 to the CSP. If TPS is dishonest, it gives TPS a penalty P1 (P1 < R4) and compensates the user R2. When the CSP uses R-MLE, the CSP cost is C C2 (C C2 > C C1 ), and the income obtained from the user is R6. When the user self-encodes, the CSP cost is C C3 (C C3 > C C1The revenue generated from users is R7.

[0036] TPS behavior: The cost of TPS integrity is C. T1 The cost of TPS dishonesty is C. T2 (C T2 <C T1 This will result in a certain loss of reputation costs (L1).

[0037] In this embodiment, a probability distribution is also included: the probability that a user does not train an autoencoder model is X, and the probability that the user trains an autoencoder model is 1-X; the probability that the CSP uses TPS is Y, the probability that it uses R-MLE is 1-Y, the probability that the TPS is honest is Z, and the probability that the TPS is dishonest is 1-Z.

[0038] Step 2: For the different strategy combinations mentioned above, conduct a detailed analysis of the benefits for users, CSPs, and TPS, and construct the corresponding benefit matrix, as shown in Table 1. This table provides a detailed analysis of the benefits under different strategy combinations, so as to intuitively show the benefits of each party under different strategy choices.

[0039] Table 1

[0040]

[0041] Step 3: Construct the replication dynamic equations for users, CSP, and TPS to determine the game equilibrium state under different strategy choices;

[0042] User: Expectations when the user does not train the autoencoder themselves Expectations during self-training and self-encoding for:

[0043]

[0044] ;

[0045] Average Expected User Strategy for:

[0046] ;

[0047] The user's replication dynamic equation is:

[0048] ;

[0049] CSP: Expectations when selecting TPS in CSP And expectations when choosing R-MLE for:

[0050] ;

[0051] CSP Strategy Average Expectation for:

[0052] ;

[0053] The replication dynamic equation of CSP is:

[0054] ;

[0055] TPS: TPS Integrity Expectations And expectations when dishonesty for:

[0056] ;

[0057] ;

[0058] TPS strategy average expectation for

[0059] ;

[0060] The replication dynamic equation of TPS is:

[0061] ;

[0062] Step 4: Construct the Jacobian matrix from the replication dynamic equations of the user, CSP, and TPS as follows:

[0063]

[0064]

[0065]

[0066]

[0067]

[0068]

[0069]

[0070]

[0071]

[0072]

[0073] Step 5: Find the eigenvalues ​​at different equilibrium points , , As shown in Table 2, the stability of a point is determined by the eigenvalues ​​of the Jacobian matrix at the equilibrium point.

[0074] Table 2

[0075]

[0076] Step 6: Analyze the sign of these eigenvalues. If all eigenvalues ​​are less than 0, the equilibrium point is stable; if there is an eigenvalue greater than 0, the equilibrium point is unstable. Analyzing the eigenvalues ​​of the equilibrium point provides a comprehensive understanding of the stability of the dynamic evolutionary game model under different equilibrium states. If the current state of the dynamic evolutionary game model is close to an unstable equilibrium point, each entity can take measures in advance to adjust its strategy to avoid the dynamic evolutionary game model falling into an unstable state.

[0077] In this embodiment, in-depth analysis of the eigenvalues ​​of the Jacobi matrix enables precise determination of the evolutionary stable strategy of the dynamic evolutionary game model. In this embodiment, when all eigenvalues ​​are less than 0, the dynamic evolutionary game model will stabilize at the equilibrium point K8(1,1,1), as shown below. Figure 3 At the equilibrium point K8(1,1,1), the probability that a user chooses not to train their own autoencoder model is 1, the probability that the CSP chooses to use TPS is 1, and the probability that the TPS chooses the integrity strategy is 1. The condition that all feature values ​​are less than 0 is as follows:

[0078] , ;

[0079] , ;

[0080] , .

[0081] The method described in this embodiment is implemented through a dynamic evolutionary game model consisting of users, CSPs, and TPSs.

[0082] In the actual collaborative process of secure deduplication, users, CSPs, and TPSs, as core stakeholders, face complex issues across multiple dimensions and levels. Coordinating the interests of users, CSPs, and TPSs is key to achieving a win-win situation for all three parties. While ensuring stable revenue for CSPs, they can incentivize users to more readily accept and adopt their solutions by offering discounts. This discount mechanism should be designed to consider user needs and expectations to increase user engagement and loyalty. Furthermore, CSPs can establish reward and penalty mechanisms based on the integrity provided by TPSs, prompting TPSs to improve their integrity and ensuring the efficient operation of the entire service chain. Through game theory analysis, the interaction and game process between users, CSPs, and TPSs can be systematically simulated and evaluated, exploring the distribution of benefits under different strategies. Within this framework, the three parties can adjust their behaviors through dynamic game theory to maximize their respective interests and promote the stability and long-term development of the entire dynamic evolutionary game model. Through the design of reward and penalty mechanisms, CSPs can incentivize TPSs to provide better services while ensuring users receive more value, thus forming a virtuous cycle and promoting the balance and growth of interests among the three parties.

[0083] like Figures 4 to 6 As shown, in the game model described in this embodiment, the strategy choices of each game entity are not only driven by its own interests but also influenced by the strategy choices of other entities. This interdependent relationship makes the game process complex and subtle. Figure 4 In (a) and (b) of the diagram, when choosing whether to train an autoencoder model themselves, users need to consider the probability of the CSP selecting TPS and the integrity of TPS. If the probability of the CSP selecting TPS is high and the integrity of TPS is guaranteed, users may be more inclined not to train the autoencoder model themselves to reduce costs and risks. Conversely, if the probability of the CSP selecting TPS is low, or if the integrity of TPS is highly uncertain, users may choose to train the autoencoder model themselves to ensure the quality and security of data processing. Figure 5 In (a) and (b) of the examples, when choosing between TPS and R-MLE, CSPs need to comprehensively consider the user's strategy choices and the impact of TPS's integrity on their own revenue. If users prefer to train their own autoencoder models, CSPs may need to adjust their strategies to better meet user needs and maximize their own profits. Figure 6In (a) and (b), when choosing between honesty and dishonesty, TPS needs to weigh the relationship between short-term interests and long-term reputation, as well as the possible countermeasures taken by CSP and users. The choices of users, CSP, and TPS are mutually influential. Under different strategy combinations, the choices of each party will change, thus affecting their payoffs. This strategic interaction and mutual influence among multiple parties constitutes a complex dynamic evolutionary game model, which requires in-depth analysis and research to reveal its inherent laws and evolutionary trends.

[0084] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0085] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A reliable data management method for multi-party evolutionary games in highly dynamic environments, characterized by: The method is implemented by the following steps: Step 1: Make assumed strategies for three entities, namely the user, CSP, and TPS, respectively, to obtain different strategy combinations; Step 2: Analyze the benefits of the user, CSP, and TPS according to different strategy combinations to obtain the corresponding benefit values under different strategy combinations; Step 3: Solve and analyze the corresponding benefit values under different strategy combinations obtained in Step 2 to obtain the replicator dynamic equations of the user, CSP, and TPS; Step 4: Construct a Jacobian matrix according to the replicator dynamic equations of the user, CSP, and TPS obtained in Step 3; Step 5: According to the Jacobian matrix in Step 4, obtain the eigenvalues at different equilibrium points, and judge the stability of the equilibrium points according to the positive and negative of the eigenvalues; Determine the evolutionary stable strategy to achieve the management of data deduplication.

2. The trusted data management method for multi-party evolutionary games in a highly dynamic environment according to claim 1, characterized in that: The assumed strategies of the three entities, namely the user, CSP, and TPS, are respectively: The assumed strategy of the user is: When the user processes data, there are two choices: self-training the autoencoder model or not training the autoencoder model by oneself; The assumed strategy of the CSP is: When the CSP provides services, there are two strategies: using the TPS or using the random information lock to encrypt the R-MLE; The assumed strategy of the TPS is: When the TPS provides services, it chooses an honest or dishonest strategy.

3. The reliable data management method for multi-party evolutionary games in a highly dynamic environment according to claim 2, characterized in that: In Step 1, the obtained strategy combinations are as follows: The cost when a user selects a self-trained autoencoder model is C. U1 The potential benefit for the user is R1; the cost for the user to choose not to train their own autoencoder model is C. U2 C U2 <C U1 -R1; If the CSP selects the TPS plan for the user at this time, the user's selection also includes the cost of using TPS, which is C. U3 If TPS is dishonest, the risk cost of data leakage it poses to users is C. U4 The CSP will provide users with compensation of R2. <C U4 ; The cost of the CSP using TPS is C. C1 The revenue obtained from users is R3, and the reward for TPS is F1; If the TPS is honest, the CSP gives the TPS a reward of R4, and the TPS brings potential benefits of E1 to the CSP; if the TPS is dishonest, the CSP gives the TPS a penalty of P1, P1 < R4, and compensates the user with R2; When CSP uses R-MLE, the CSP cost is C C2 C C2 >C C1 The revenue obtained from users is R6. The CSP cost is C when users self-code. C3 C C3 >C C1 Revenue generated from users is R7; The cost of TPS integrity is C. T1 The cost of TPS dishonesty is C. T2 C T2 <C T1 The cost of reputation loss is L1.

4. The reliable data management method for multi-party evolutionary games in a highly dynamic environment according to claim 3, characterized in that: It also includes setting probability distributions, that is: the probability that the user does not train the autoencoder model by oneself is X, and the probability of self-training the autoencoder model is 1 - X; the probability that the CSP uses the TPS is Y, the probability of using the R-MLE is 1 - Y, the probability that the TPS is honest is Z, and the probability that the TPS is dishonest is 1 - Z.

5. The reliable data management method for multi-party evolutionary games in a highly dynamic environment according to claim 4, characterized in that: In Step 3, the process of obtaining the replicator dynamic equation of the user is: Setting expectations when users do not train their own autoencoder models Expectations when training a self-encoding model for: ； ； Average expectation of user strategy for: ； The replicator dynamic equation of the user is: 。 6. The reliable data management method for multi-party evolutionary games in a highly dynamic environment according to claim 5, characterized in that: In Step 3, the process of obtaining the replicator dynamic equation of the CSP is: When setting CSP to select TPS, the expected value is... And expectations when choosing R-MLE for: ； ； Average expectation of CSP strategy for: ； The replicator dynamic equation of the CSP is: 。 7. The reliable data management method for multi-party evolutionary games in a highly dynamic environment according to claim 6, characterized in that: In Step 3, the process of obtaining the replicator dynamic equation of the TPS is: Expectations when setting TPS integrity And expectations when dishonesty for: ； ； Average expectation of TPS strategy for: ； The replicator dynamic equation of the TPS is: 。 8. The reliable data management method for multi-party evolutionary games in a highly dynamic environment according to claim 7, characterized in that: In Step 5, if all eigenvalues are less than 0, then this equilibrium point is stable; if there are eigenvalues greater than 0, then this equilibrium point is unstable.

9. The trusted data management method for multi-party evolutionary game in a high-dynamic environment according to claim 8, wherein: When the condition that all eigenvalues are less than 0 is satisfied, it will stabilize at a certain equilibrium point. At this equilibrium point, it means that the probability that the user chooses not to train the autoencoder model by oneself is 1, the probability that the CSP chooses to use the TPS is 1, and the probability that the TPS chooses an honest strategy is 1; the conditions when all eigenvalues are less than 0 are as follows: ， ； ， ； ， 。 10. The trusted data management method for multi-party evolutionary games in a highly dynamic environment according to any one of claims 1-9, characterized in that: This method is implemented through a dynamic evolutionary game model composed of the user, CSP, and TPS.

Images