Lightweight batch authentication method based on long and short term historical information fusion

Abstract

The invention discloses a lightweight batch authentication method based on long-term and short-term historical information fusion, and the method comprises the steps: collecting and recording long-term and short-term historical information with cooperation time parameters in a process that edge equipment executes a distributed cooperation task; establishing a memory bank dynamic updating mechanism based on comprehensive storage scores, and performing self-adaptive management on collaboration historical data; selecting historical information fragments which are uniformly distributed in time and are difficult to eavesdrop, predict and analyze by a third party from the memory bank by adopting a time hierarchical selection strategy as authentication information; performing linear transformation on local parameters in the authentication information by utilizing dynamic linear transformation to generate behavior tokens, and broadcasting the behavior tokens to group members; and after receiving all the tokens, the edge device reversely restores the local parameters locally, reconstructs the global parameters, and determines whether the authentication is passed by analyzing whether the global parameters are consistent with the global parameters in the authentication information. According to the invention, the calculation and communication overhead of edge cooperation authentication is effectively reduced, and the security and efficiency of authentication are improved.

Description

Technical Field

[0001] This invention relates to the fields of network security and distributed computing technology, and in particular to a lightweight batch authentication method based on the fusion of long-term and short-term historical information. Background Technology

[0002] In distributed edge collaborative networks, establishing a reliable identity authentication mechanism is a core prerequisite for ensuring system security and trust. With the proliferation of decentralized scenarios such as drone swarms and vehicle-to-everything (V2X) collaborative positioning, how to verify the legitimacy of devices in environments without a central certification authority (CA) and where nodes are highly dynamic has become a research hotspot in the field of cybersecurity. Traditional identity authentication schemes typically rely on pre-shared keys or complex public key infrastructures (PKI). However, these encryption mechanisms are often independent of collaborative tasks and cannot perceive the behavioral characteristics of nodes during algorithm iteration. In resource-constrained edge environments, traditional heavy cryptographic primitives (such as bilinear pairing and large number modular exponentiation) not only incur heavy computational overhead but also struggle to cope with security failures caused by physical impersonation of node identities or theft of credentials. Especially when multi-agent collaboratively execute distributed algorithms (such as ADMM and federated learning), the system generates massive amounts of intermediate parameters with spatiotemporal correlation characteristics. Existing authentication methods fail to fully exploit these highly identity-bound "algorithmic behavioral fingerprints," resulting in authentication mechanisms lacking dynamic evolution capabilities and forward security. Therefore, exploring an endogenous authentication method that can deeply couple collaborative historical information and has low-overhead batch verification characteristics is key to solving the problems of identity fraud and unauthorized access in edge collaboration systems. Summary of the Invention

[0003] The purpose of this invention is to address the problems of excessive reliance on centralized trust institutions and excessive computational and communication overhead in existing group authentication technologies. It provides a lightweight batch authentication method based on the fusion of long and short-term historical information. By establishing a memory bank to dynamically manage collaboration history information, selecting long and short-term historical information, and merging collaboration time to generate behavior tokens, it achieves low-overhead, decentralized batch identity verification, effectively improving the security and efficiency of authentication in edge collaboration systems.

[0004] The technical solution adopted in this invention is:

[0005] A lightweight batch authentication method based on the fusion of long-term and short-term historical information is applied to a distributed collaborative network without a centralized authentication authority, consisting of several edge devices. The method includes the following steps:

[0006] Long-term and short-term historical information collection: Recording The local parameters and global consensus parameters generated by each edge device during the execution of distributed collaborative tasks form a sequence with collaboration time. The parameter record is used as a long-term or short-term historical information. It is stored in the local memory of each edge device;

[0007] Long-term and short-term historical information management: Establish a dynamic update mechanism for the local memory bank to manage long-term and short-term historical information. Subdivided into multiple historical information fragments based on collaborative iteration rounds. When the memory storage space is saturated, a comprehensive storage score is calculated for each historical information segment. And remove low-value historical information fragments based on the comprehensive storage score;

[0008] Authentication information selection: A time-stratified selection strategy is adopted, selecting information from the local memory. A historical information fragment evenly distributed along the historical timeline. As authentication information;

[0009] Authentication token construction based on the fusion of long-term and short-term historical information: edge devices extract selected... Local parameters corresponding to each historical information fragment are integrated with the iteration rounds and collaboration times corresponding to each fragment. The dynamic obfuscation parameters are linearly transformed to obtain the transformed local parameters to generate a behavior token, which is then broadcast to other edge devices in the group.

[0010] Distributed batch authentication process: After receiving all behavior tokens, each edge device performs an inverse transformation on the behavior tokens to obtain the restored local parameters, and reconstructs the global consensus parameters based on the restored local parameters; when the reconstructed global consensus parameters match the selected... If the corresponding global consensus parameters stored in each historical information segment match, it is determined that all devices in the group have passed authentication; otherwise, it is determined that an attacker exists.

[0011] Furthermore, the algorithm used to perform distributed cooperative tasks is applicable to a variety of distributed cooperative algorithms, including Alternating Direction Multiplier Method (ADMM), Federated Learning, Multi-Agent Reinforcement Learning, etc., and there is no Central Certification Authority (CA) in the group.

[0012] Furthermore, local parameters It is generated by multiple edge devices in multiple iterations, for the first... An edge device, the specific structure of its local parameters is as follows: ,in, For iterative rounds in distributed collaboration, Indicates the first The edge device in the first Local parameters generated during round iteration.

[0013] Furthermore, global consensus parameters It was generated during multiple iterations, and its specific structure is as follows: ,in, For iterative rounds in distributed collaboration, Indicates the first step in the collaboration process The global consensus parameters generated by the round of iteration.

[0014] Furthermore, the stored long-term and short-term historical information structure is as follows: parameter records of all members. Constitutes a long-term and short-term historical information Local memory The storage structure is ,in, The number of members in the group. For iterative rounds in distributed collaboration, For distributed collaboration, For the first The device in the Local parameters in round iteration, For the first Global parameters in round iteration, For the first All parameters in this collaboration.

[0015] Furthermore, historical information fragments It consists of long-term and short-term historical information According to iteration rounds Further subdivision reveals that the smallest storage and authentication unit in the local memory bank has the following specific structure: ,in, Indicates the first step in the collaboration process The global consensus parameters generated by the round of iterations For the first The device in the Local parameters in round iteration.

[0016] Furthermore, comprehensive storage score Based on global parameters The calculation is based on the relevant time and usage statistics, without involving local parameters. It is jointly determined by the collaboration value item, time value item, and risk penalty item through linear weighting. The calculation formula is as follows:

[0017] ;

[0018] in, Represents global parameters during collaboration. The range of change, The weighting coefficient for collaborative value is related to the iteration round at which historical information was generated. Related; This represents the difference between the current collaboration count and the collaboration count when the historical information fragment was generated, used to indicate the timeliness of the information. As the basic weight for the value of time, The time decay coefficient, is the base of the natural logarithm; This indicates the cumulative number of times this historical information has been selected as an authentication credential, used to indicate the risk of repeated use. As the basic weight for risk penalty, This is a risk growth index.

[0019] Furthermore, the formula for the magnitude of change Specifically, it is determined by calculating the absolute logarithm of the ratio of the parameter update step size between two adjacent iterations, which is used to quantify the abrupt changes in the parameter evolution process. The calculation formula is as follows:

[0020] ;

[0021] in, No. The global consensus parameters generated in each round of iterations Represents the Euclidean distance between vectors. To prevent the minimum constant with a denominator of zero.

[0022] Furthermore, the specific execution process of the time-stratified selection strategy is as follows: The long-term and short-term historical information stored in the local memory bank are arranged in chronological order of their generation time, and then evenly divided according to the number of historical information fragments. A series of consecutive time intervals, selecting from each time interval The largest historical information fragments are ultimately combined to form a sequence that is evenly distributed over time and contains the maximum parameter evolution characteristics within each time interval. A fragment of historical information As authentication information.

[0023] Furthermore, the linear transformation uses dynamic confusion parameters. The linear transformation is calculated using the collaboration time calculation method, and the calculation formula is as follows:

[0024] ;

[0025] in, For the first The first edge device targets the first A behavior token element generated from a historical information fragment. These are local parameters within this historical information segment. This refers to the iteration round corresponding to this historical information fragment. This refers to the collaboration session corresponding to this historical information fragment. and Updated independently with each authentication task, the parameter seed is initially distributed by the group. , Through random number generation function generate:

[0026] ;

[0027] ;

[0028] Furthermore, the process of reconstructing global parameters is as follows: after receiving all tokens, each device utilizes the shared collaboration time parameter. , and obfuscated parameters , The process involves reversing the steps to reconstruct all local parameters and repeating the collaboration process. The specific formula is as follows:

[0029] ;

[0030] in, The number of members in the group. For the first The aggregate weight of each device in collaboration These are the restored local parameters.

[0031] The present invention adopts the above technical solution, and compared with the traditional method, the beneficial effects of the present invention are as follows:

[0032] 1. Extremely low computational overhead: This invention abandons complex cryptographic techniques and only involves basic linear algebra operations, with computational overhead at the millisecond level, greatly reducing the computational burden on edge devices.

[0033] 2. Intrinsic security: It uses historical information generated by distributed collaboration algorithms as authentication credentials and integrates collaboration time to generate tokens. If an attacker has not participated in real historical collaboration, it is difficult to construct a token that can restore the real historical information.

[0034] 3. Dynamic forward security: The authentication target and the parameters used for linear transformation change dynamically, so even if the parameters constituting a single token are leaked, it is impossible to deduce historical or future authentication information.

[0035] 4. Decentralized architecture: The authentication process relies entirely on the historical data and calculations of group members locally, without the need for an online CA. Attached Figure Description

[0036] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments;

[0037] Figure 1This is a schematic diagram of the overall framework of the lightweight batch authentication method based on the fusion of long and short-term historical information in the distributed edge collaboration of the present invention.

[0038] Figure 2 This is a schematic diagram comparing the online computational overhead performance of traditional cryptographic schemes with the increase of the number of edge devices in the embodiment;

[0039] Figure 3 This is a schematic diagram illustrating the distribution and evolution of authentication information across historical collaboration rounds in the example. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0041] like Figures 1 to 3 As shown in any of the accompanying figures, this invention discloses a lightweight batch authentication method based on the fusion of long-term and short-term historical information, applied to a distributed collaborative network without a centralized authentication authority, consisting of several edge devices. The method includes the following steps:

[0042] Long-term and short-term historical information collection: Record the local parameters and global consensus parameters generated by N edge devices during the execution of distributed collaborative tasks, forming a parameter record with a collaboration time t as a long-term and short-term historical information. It is stored in the local memory of each edge device;

[0043] Long-term and short-term historical information management: Establish a dynamic update mechanism for the local memory bank to manage long-term and short-term historical information. Subdivided into multiple historical information fragments based on collaborative iteration rounds. When the memory storage space is saturated, a comprehensive storage score is calculated for each historical information segment. And remove low-value historical information fragments based on the comprehensive storage score;

[0044] Authentication information selection: A time-stratified selection strategy is adopted, selecting information from the local memory. A historical information fragment evenly distributed along the historical timeline. As authentication information;

[0045] Authentication token construction based on the fusion of long-term and short-term historical information: edge devices extract selected... The local parameters corresponding to each historical information fragment are linearly transformed by integrating the iteration round, collaboration time t, and dynamic obfuscation parameters corresponding to each fragment to obtain transformed local parameters to generate behavior tokens, which are then broadcast to other edge devices in the group.

[0046] Distributed batch authentication process: After receiving all behavior tokens, each edge device performs an inverse transformation on the behavior tokens to obtain the restored local parameters, and reconstructs the global consensus parameters based on the restored local parameters; when the reconstructed global consensus parameters match the corresponding global consensus parameters stored in the selected L historical information fragments, it is determined that all devices in the group have passed authentication; otherwise, it is determined that an attacker exists.

[0047] Furthermore, the algorithm used to perform distributed cooperative tasks is applicable to a variety of distributed cooperative algorithms, including Alternating Direction Multiplier Method (ADMM), Federated Learning, Multi-Agent Reinforcement Learning, etc., and there is no Central Certification Authority (CA) in the group.

[0048] Furthermore, local parameters It is generated by multiple edge devices in multiple iterations, for the first... An edge device, the specific structure of its local parameters is as follows: ,in, For iterative rounds in distributed collaboration, Indicates the first The edge device in the first Local parameters generated during round iteration.

[0049] Furthermore, global consensus parameters It was generated during multiple iterations, and its specific structure is as follows: ,in, For iterative rounds in distributed collaboration, Indicates the first step in the collaboration process The global consensus parameters generated by the round of iteration.

[0050] Furthermore, the stored long-term and short-term historical information structure is as follows: parameter records of all members. Constitutes a long-term and short-term historical information Local memory The storage structure is ,in, The number of members in the group. For iterative rounds in distributed collaboration, For distributed collaboration, For the first The device in the Local parameters in round iteration, For the first Global parameters in round iteration, For the first All parameters in this collaboration.

[0051] Furthermore, historical information fragments It is the smallest unit of storage and authentication in the local memory bank, with the following specific structure: ,in, Indicates the first step in the collaboration process The global consensus parameters generated by the round of iterations For the first The device in the Local parameters in round iteration.

[0052] Furthermore, comprehensive storage score Based on global parameters The calculation is based on the relevant time and usage statistics, without involving local parameters. It is jointly determined by the collaboration value item, time value item, and risk penalty item through linear weighting. The calculation formula is as follows:

[0053] ;

[0054] in, Represents global parameters during collaboration. The range of change, The weighting coefficient for collaborative value is related to the iteration round at which historical information was generated. Related; This represents the difference between the current collaboration count and the collaboration count when the historical information fragment was generated, used to indicate the timeliness of the information. As the basic weight for the value of time, The time decay coefficient, is the base of the natural logarithm; This indicates the cumulative number of times this historical information has been selected as an authentication credential, used to indicate the risk of repeated use. As the basic weight for risk penalty, This is a risk growth index.

[0055] Furthermore, the formula for the magnitude of change Specifically, it is determined by calculating the absolute logarithm of the ratio of the parameter update step size between two adjacent iterations, which is used to quantify the abrupt changes in the parameter evolution process. The calculation formula is as follows:

[0056] ;

[0057] in, No. The global consensus parameters generated in each round of iterations Represents the Euclidean distance between vectors. To prevent the minimum constant with a denominator of zero.

[0058] Furthermore, the specific execution process of the time-stratified selection strategy is as follows: The long-term and short-term historical information stored in the local memory bank are arranged in chronological order of their generation time, and then evenly divided according to the number of historical information fragments. A series of consecutive time intervals, selecting from each time interval The largest historical information fragments are ultimately combined to form a sequence that is evenly distributed over time and contains the maximum parameter evolution characteristics within each time interval. A fragment of historical information As authentication information.

[0059] Furthermore, the linear transformation uses dynamic confusion parameters. The linear transformation is calculated using the collaboration time calculation method, and the calculation formula is as follows:

[0060] ;

[0061] in, For the first The first edge device targets the first A behavior token element generated from a historical information fragment. These are local parameters within this historical information segment. This refers to the iteration round corresponding to this historical information fragment. This refers to the collaboration session corresponding to this historical information fragment. and The parameter seeds are updated independently with each authentication task and distributed to each edge device by the registration authority during the system initialization phase. , Through random number generation function generate:

[0062] ;

[0063] ;

[0064] Furthermore, the process of reconstructing global parameters is as follows: after receiving all tokens, each device utilizes the shared collaboration time parameter. , and obfuscated parameters , The process involves reversing the steps to reconstruct all local parameters and repeating the collaboration process. The specific formula is as follows:

[0065] ;

[0066] in, The number of members in the group. For the first The aggregate weight of each device in collaboration For the restored local parameters, These are the global parameters for reconstruction.

[0067] The specific principles of this invention will be explained in detail below:

[0068] This embodiment constructs a decentralized authentication authority (CA)-less edge collaboration system comprising N edge devices. All devices collaborate in a distributed cooperative localization task using the Alternating Direction Multiplier Method (ADMM). Figure 1 As shown, it includes the following steps:

[0069] 1) Initialization: During the group startup phase, the following initialization parameters are distributed to each legitimate device via a temporary Registration Manager (RM): Parameter seed: , This is used to generate pseudo-random numbers later. Hash function: This is used for subsequent calculation of pseudo-random numbers. After distribution is complete, the RM goes offline, and the system enters a fully decentralized operating mode.

[0070] 2) Historical information collection: Group execution Step-by-step ADMM algorithm iteration. For each device... In the During the iteration, the following standard ADMM update steps are performed:

[0071] Local parameter update: ;

[0072] The device is based on the objective function and global parameters Local parameters Update local parameters ,

[0073] ;

[0074] The device is based on global parameters and local parameters Update local parameters ;

[0075] Global parameter update: ;

[0076] Devices exchange their respective local parameters Calculate global consensus parameters ,in, As the penalty parameter for the ADMM algorithm, the device will Parameter recording of each iteration As a piece of historical information, both long-term and short-term Stored in local memory bank middle.

[0077] 3) Historical Information Management: This involves managing each piece of historical information in the memory bank. Subdivided into multiple historical information fragments Calculate the comprehensive storage score for each historical information fragment. :

[0078] ;

[0079] ;

[0080] When the memory bank storage space is saturated, remove low-rated historical information fragments.

[0081] 4) Authentication Information Selection: The device arranges the long and short-term historical information stored in the local memory bank in chronological order of generation time, and evenly divides them according to the number of historical information fragments. A series of consecutive time intervals are used to select the magnitude of global parameter changes within each time interval. The largest fragment of historical information will be this A fragment of historical information This serves as the certification information for this certification.

[0082] 5) Authentication token construction: The device selects... Extracting local parameters from historical information Composition of plaintext tokens : ;

[0083] And utilize dynamic obfuscation parameters And linearly transform the collaboration time:

[0084] ;

[0085] in, For each piece of historical information, there is a separate collaboration session. For each iteration round in the collaboration, Calculated by edge devices:

[0086] ;

[0087] ;

[0088] The device will use the token Broadcast to other members of the group.

[0089] 6) Batch authentication:

[0090] S1: After receiving a token broadcast by other devices, the device uses the shared parameters. Calculate and restore local parameters :

[0091] ;

[0092] S2: The device uses the restored local parameters and substitutes them into the aggregation formula of the ADMM algorithm to reconstruct the global parameters. :

[0093] ;

[0094] ;

[0095] S3: Check the reconstructed global parameters Is it consistent with the selected If the real global parameters match in a historical information fragment, the authentication is successful; otherwise, at least one attacker exists in the group.

[0096] Verification experiment:

[0097] Experimental conditions: In a 700m × 700m outdoor environment, 5 edge devices and the positioning target were randomly distributed; the signal transmission power was 20dBm; and the learning error limit was [missing information]. .

[0098] Baseline Comparison: An ECC-based identity authentication scheme was used as the baseline comparison method.

[0099] Experimental steps: Implement a lightweight batch authentication method based on the fusion of long and short-term historical information in distributed edge collaboration according to the following process.

[0100] Distributed collaboration: Five edge devices determine their distances to the cooperative localization target by measuring Received Signal Strength Indication (RSSI), run the distributed ADMM algorithm, and calculate, update, and save the learned parameters. Until convergence.

[0101] Token building and broadcasting: Each edge device performs dynamic management of historical information in its local memory based on a comprehensive storage score, and employs a time-tiered selection strategy to filter out... A historical information fragment covering the collaboration history, using random numbers generated from a pre-set seed. And the collaboration time corresponding to each message. Local parameters for selected historical information segments Perform a linear transformation to generate a behavior token that fuses long-term and short-term historical information. And broadcast it to other devices in the group.

[0102] Batch authentication of devices: The device receives tokens broadcast by other devices. Then, the local parameters are reverse-engineered. And substitute them into the ADMM aggregation formula to reconstruct the global parameters. The system verifies whether the reconstructed global parameters match the true global parameters in the selected long and short-term historical information. If the conditions are met, the authentication is successful.

[0103] Repeat the above steps 5 times, for a total of 5 collaborative positioning and 5 authentications.

[0104] Experimental results: such as Figure 2 As shown, the computational overhead performance of a lightweight batch authentication method based on the fusion of long-term and short-term historical information in distributed edge collaboration is compared with that of a traditional baseline scheme based on ECC (Elliptic Curve Cryptography). To visually demonstrate the significant difference in computational overhead between the two methods, the vertical axis uses the "computational overhead exponent" (i.e., the common logarithm of computation time in milliseconds). The figure shows that, relying on lightweight linear algebra operations, the average online computational overhead per authentication in this embodiment remains stable at around 0.2 (corresponding to an actual time of approximately 0.24 milliseconds), represented as a horizontal line close to the bottom. Furthermore, for batch authentication, the computational overhead does not increase with the number of edge devices, remaining consistently at the millisecond level. In contrast, the baseline scheme involves complex cryptographic operations and requires pairwise authentication between devices, resulting in a rapid increase in computational overhead with the number of devices. This demonstrates the significant efficiency advantage of this invention on edge devices with limited computing resources.

[0105] like Figure 3 The figure illustrates the evolution of the distribution of authentication information selection as the collaboration rounds progress (from round 2 to round 5) in this embodiment of the invention. The four sub-graphs (a–d) show the sources of authentication information for the authentication token in rounds 2, 3, 4, and 5, respectively. The horizontal axis represents the number of historical collaborations, and the vertical axis represents the number of authentication information selected in that collaboration. It can be seen that as collaboration continues, the authentication information is not concentrated solely on the most recent collaborations, but is widely and dynamically distributed across various historical stages from the early stages to the present. Specifically, in the early stages of collaboration, the authentication information mainly comes from newly generated historical information; while in the later stages, the selected authentication information includes both the latest collaboration information and earlier collaboration information. This distribution characteristic intuitively verifies the long-short-term historical information fusion mechanism of this invention, indicating that the system can effectively utilize the full lifecycle data in the memory bank to construct authentication tokens with high spatiotemporal complexity, thereby significantly increasing the difficulty for attackers to commit fraud by forging data from a single stage and improving the inherent security of the system.

[0106] This invention addresses the technical problems of existing solutions, which heavily rely on centralized infrastructure, incur high computational overhead, and struggle to achieve dynamic intrinsic security. The solution presented in this invention exhibits significant decentralization, low computational load (involving only linear algebra operations), and requires no complex cryptographic mechanisms. Simultaneously, it possesses intrinsic security against replay and fraud, and can flexibly adapt to edge device groups of varying sizes and computing capabilities. It can provide efficient and reliable security authentication support for a wide range of edge computing applications, including drone swarm formations, vehicle-to-everything (V2X) collaborative positioning, and distributed sensor networks.

[0107] The present invention adopts the above technical solution, and compared with the traditional method, the beneficial effects of the present invention are as follows:

[0108] 1. Extremely low computational overhead: This invention abandons complex cryptographic techniques and only involves basic linear algebra operations, with computational overhead at the millisecond level, greatly reducing the computational burden on edge devices.

[0109] 2. Intrinsic security: It uses historical information generated by distributed collaboration algorithms as authentication credentials and integrates collaboration time to generate tokens. If an attacker has not participated in real historical collaboration, it is difficult to construct a token that can restore the real historical information.

[0110] 3. Dynamic forward security: The authentication target and the parameters used for linear transformation change dynamically, so even if the parameters constituting a single token are leaked, it is impossible to deduce historical or future authentication information.

[0111] 4. Decentralized architecture: The authentication process relies entirely on the historical data and calculations of group members locally, without the need for an online CA.

[0112] Obviously, the described embodiments are only a part of the embodiments of this application, not all of them. Without conflict, the embodiments and features in the embodiments of this application can be combined with each other. The components of the embodiments of this application described and illustrated herein can generally be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of this application is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

Claims

1. A lightweight batch authentication method based on the fusion of long-term and short-term historical information, applied to a distributed collaborative network without a centralized authentication authority, consisting of several edge devices, characterized in that... The method includes the following steps: Long-term and short-term historical information collection: Recording The local parameters and global consensus parameters generated by each edge device during the execution of a distributed collaborative task form a parameter record with a collaboration time t, serving as a short-term and long-term historical record. It is stored in the local memory of each edge device; Long-term and short-term historical information management: Establish a dynamic update mechanism for the local memory bank to manage long-term and short-term historical information. Subdivided into multiple historical information fragments based on collaborative iteration rounds. When the memory storage space is saturated, a comprehensive storage score is calculated for each historical information segment. And remove low-value historical information fragments based on the comprehensive storage score; Authentication information selection: A time-stratified selection strategy is adopted, selecting information from the local memory. A historical information fragment evenly distributed along the historical timeline. As authentication information; Authentication token construction based on the fusion of long-term and short-term historical information: edge devices extract selected... Local parameters corresponding to each historical information fragment are integrated with the iteration rounds and collaboration times corresponding to each fragment. The dynamic obfuscation parameters are linearly transformed to obtain the transformed local parameters to generate a behavior token, which is then broadcast to other edge devices in the group. Distributed batch authentication process: After receiving all behavior tokens, each edge device performs an inverse transformation on the behavior tokens to obtain the restored local parameters, and reconstructs the global consensus parameters based on the restored local parameters; when the reconstructed global consensus parameters match the selected... If the corresponding global consensus parameters stored in each historical information segment match, it is determined that all devices in the group have passed authentication; otherwise, it is determined that an attacker exists.

2. The lightweight batch authentication method based on the fusion of long-term and short-term historical information as described in claim 1, characterized in that, The algorithms used in distributed collaborative tasks include alternating direction multiplier method, federated learning, and multi-agent reinforcement learning.

3. The lightweight batch authentication method based on the fusion of long-term and short-term historical information according to claim 1, characterized in that, Local parameters are generated by the edge device during multiple iterations; the first Local parameters of an edge device ,in, ; For iterative rounds in distributed collaboration, Indicates the first The edge device in the first Local parameters generated during round iteration; Global consensus parameters It was generated during multiple iterations. ,in, For iterative rounds in distributed collaboration, Indicates the first step in the collaboration process The global consensus parameters generated by the round of iteration.

4. The lightweight batch authentication method based on the fusion of long-term and short-term historical information according to claim 1, characterized in that, A short-term and long-term historical record stored in the local memory bank Including parameter records of all members Local memory The storage structure is ,in, For the number of edge devices, For iterative rounds in distributed collaboration, This represents the total number of distributed collaborations. For the first The edge device in the first Local parameters in round iteration, For the first Global parameters in iteration For the first All parameters in this collaboration.

5. A lightweight batch authentication method based on the fusion of long-term and short-term historical information as described in claim 1, characterized in that, Historical information fragments It consists of long-term and short-term historical information According to iteration rounds Further subdivision reveals that historical information fragments, as the smallest storage and authentication unit in the local memory bank, The structure is ,in, Indicates the first step in the collaboration process The global consensus parameters generated by the round of iterations For the first The device in the Local parameters in round iteration.

6. The lightweight batch authentication method based on the fusion of long-term and short-term historical information according to claim 1, characterized in that, Overall Storage Rating Based on global parameters The collaboration value, time value, and risk penalty are calculated using a linear weighted average, as shown in the following formula: ； in, Represents global parameters during collaboration. The range of change, This is a weighting factor for the value of collaboration; This represents the difference between the current collaboration sequence and the collaboration sequence when the corresponding historical information fragment was generated; As the basic weight for the value of time, The time decay coefficient, is the base of the natural logarithm; This indicates the cumulative number of times historical information has been selected as authentication credentials; As the basic weight for risk penalty, This is a risk growth index.

7. A lightweight batch authentication method based on the fusion of long-term and short-term historical information as described in claim 6, characterized in that, global parameters range of change The parameter update step size is determined by calculating the absolute logarithm of the ratio of the parameter update step sizes between two adjacent iterations, as shown in the following formula: ； in, No. The global consensus parameters generated in each round of iterations No. The global consensus parameters generated in each round of iterations No. The global consensus parameters generated in each round of iterations Represents the Euclidean distance between vectors. To prevent the minimum constant with a denominator of zero.

8. A lightweight batch authentication method based on the fusion of long-term and short-term historical information as described in claim 1, characterized in that, The time-stratified selection strategy is implemented as follows: Long-term and short-term historical information stored in the local memory bank are arranged in chronological order of their generation time; the historical information fragments are evenly divided according to their quantity. A series of consecutive time intervals; within each time interval, select The largest historical information fragment; ultimately combined into a structure that is evenly distributed over time and contains the maximum parameter evolution characteristics within each time interval. A fragment of historical information As authentication information.

9. A lightweight batch authentication method based on the fusion of long-term and short-term historical information as described in claim 1, characterized in that, Using dynamic obfuscation parameters Perform linear transformation calculations, the first linear transformation... The first edge device The token element generated from each historical information fragment is: ； in, For the first The first edge device targets the first A behavior token element generated from a historical information fragment. These are local parameters within historical information fragments. For the iteration rounds corresponding to historical information fragments, This refers to the collaboration order corresponding to historical information fragments. and The parameter seeds are updated independently with each authentication task and distributed to each edge device by the registration authority during the system initialization phase. and Through a random number generation function generate: ; ; This represents the number of historical information fragments.

10. A lightweight batch authentication method based on the fusion of long-term and short-term historical information as described in claim 9, characterized in that, The global parameter reconstruction is implemented as follows: After receiving all behavior tokens, each edge device reverse-engineers all local parameters and repeats the collaboration process. The specific formula is as follows: ； in, This refers to the number of edge devices within the group. For the first The aggregation weight of each edge device in collaboration For the restored local parameters, These are the global parameters for reconstruction.

Images