Cross-station identity authentication method and system for trust continuity and behavior evolution

Abstract

The invention belongs to the technical field of identity authentication and access control of platform security, and particularly relates to a cross-site identity authentication method and system for trust continuity and behavior evolution, and the method comprises the steps: S1, collecting anonymous cross-site user information, and carrying out the unified standardization processing; s2, constructing a heterogeneous structure for collaborative modeling of attribute features and behavior features to realize joint characterization of identity features; s3, constructing an identity structure consistency alignment ISCAM module, evaluating the consistency deviation degree of the identity features and a historical identity structure, and extracting cross-station identity feature representation under structural constraints; s4, constructing a multi-task constrained association hypothesis to generate an AHGM module, and adaptively generating a multi-path cross-station identity association hypothesis when a consistency conflict exists; s5, constructing a trust continuity backtracking verification TCC-BVM module, and realizing evaluation and backtracking correction of time continuity and behavior evolution consistency of the candidate paths; according to the method, cross-station identity drift and misassociation are effectively inhibited, so that reliable identification of cross-station identities in an anonymous network environment is realized.

Description

Technical Field

[0001] This invention relates to the field of identity authentication and access control technology for platform security, specifically a cross-site identity authentication method and system based on trust continuity and behavioral evolution. Background Technology

[0002] With the rapid development of anonymous networks, dark web forums, and decentralized communities, it has become commonplace for users to operate with multiple accounts and identities across different anonymous sites. These online environments typically lack unified identity management and access control mechanisms, allowing users to freely change usernames, modify profiles, or participate in discussions intermittently, resulting in highly fragmented and heterogeneous identity representations across multiple sites. While this characteristic enhances user privacy to some extent, it also poses significant challenges to cybersecurity governance, tracing illegal activities, and cross-platform access control.

[0003] To address the aforementioned issues, existing cross-site identity authentication methods have achieved some research results. Related technologies typically rely on static identity attributes such as usernames and profiles, combined with textual semantic similarity calculations or cluster analysis, to achieve preliminary association of user identities across different sites. Some studies further incorporate user-generated content, posting time, and contextual information, improving the robustness of identity authentication through behavioral feature modeling or multimodal embedding. Furthermore, the introduction of multi-task learning and contrastive learning techniques allows multiple identity-related sub-tasks to collaboratively optimize within a shared feature space, improving the overall accuracy and generalization ability of cross-site identity recognition to some extent. However, most of these methods focus on local feature similarity or static discrimination results, and still have certain limitations. First, existing behavioral modeling methods are mostly based on discrete time series, which makes it difficult to effectively characterize the continuous evolution of users under the influence of cross-site migration, intermittent activity, or sudden events. Second, there is a lack of explicit modeling of users' historical identity structure, making it difficult to assess the degree of consistency deviation between current identity features and historical identities from the overall structural level. Third, when there are structural conflicts in identity features across different sites, existing methods usually only output a single matching result, lacking multiple hypothesis generation and backtracking verification mechanisms, which makes it difficult to meet the requirements of network security scenarios for the stability and reliability of identity authentication results. Summary of the Invention

[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0005] A cross-site identity authentication method based on trust continuity and behavioral evolution includes the following steps:

[0006] S1: Collect user-related data from multiple anonymous websites, standardize and process user attribute information and user-generated content in a unified manner, and eliminate differences in data format, expression method and time scale among different websites;

[0007] S2: Construct a static identity attribute feature modeling branch and a behavioral feature evolution modeling branch to perform static-dynamic heterogeneous dual-branch modeling of user identity features on different sites;

[0008] S3: Construct the Identity Structure Consistency Alignment (ISCAM) module to assess the degree of consistency deviation between cross-site identity features and the user's historical identity structure reference distribution, and generate cross-site identity feature representations;

[0009] S4: Construct the AHGM module to generate association hypotheses with multi-task constraints, so as to obtain multiple candidate cross-site identity association hypotheses after identity structure conflict occurs;

[0010] S5: Under the structural deviation constraint feedback of the ISCAM module, construct the Trust Continuity Backtracking Verification TCC-BVM module to realize the backtracking verification and correction of candidate cross-site identity association assumptions;

[0011] S6: Output cross-site user identity association results that simultaneously satisfy trust continuity constraints and identity structure stability, realizing cross-site identity authentication in an anonymous network environment.

[0012] As a preferred embodiment of the cross-site identity authentication method based on trust continuity and behavioral evolution described in this invention, the specific steps of S2 are as follows:

[0013] S21: Input the username and personal profile into the static identity attribute feature modeling branch to obtain the identity attribute representation that reflects the stability of the user's identity. ;

[0014] S22: Input user-generated content, behavioral time information, and contextual information into the behavioral feature evolution modeling branch to obtain an identity behavior representation that reflects the evolution of user behavior patterns and content style characteristics. .

[0015] As a preferred embodiment of the cross-site identity authentication method based on trust continuity and behavioral evolution described in this invention, the specific steps of S21 are as follows:

[0016] S211: Input the text set consisting of the user's username and personal profile into the cross-language pre-trained semantic encoding model for unified semantic mapping, and obtain a high-dimensional static attribute semantic vector. ;

[0017] S212: Under the constraint of information bottleneck, for the... By performing variational inference and information decoupling modeling, we can obtain the core identity components that represent the user's essential personality preferences. and the environmental noise component that characterizes the expressive style and contextual bias of a site ;

[0018] S213: Regarding the above Discretization mapping, in the global codebook Select the cluster center vector with the smallest distance from it. Quantified static identity representation is obtained At the same time, the aforementioned and Input adversarial learning network, suppress Despite interference with identity discrimination features, the output environment remains unchanged, resulting in a static identity fingerprint representation. ;

[0019] S214: Static identity fingerprints of the same user across different sites Model as a distribution of latent identity variables Furthermore, a commitment loss constraint is introduced, with minimizing the distribution of cross-site identity latent variables as the primary optimization objective. This achieves cross-site alignment of static identity attributes at the distribution level, ultimately outputting a steady-state identity attribute representation. ;

[0020] The specific steps of S22 are as follows:

[0021] S221: Perform content semantic encoding, time period encoding, and context association encoding on the discrete behaviors of users across multiple anonymous websites to obtain user... Behavioral state vector sequence at discrete time points ;

[0022] S222: The above In the continuous-time behavior evolution model with input structure consistency constraints, the user's cross-site behavior over continuous time intervals... Internal modeling yields a representation of the behavioral evolution trajectory. ;

[0023] S223: The above Mapping to the public behavior feature space yields the identity and behavior feature representations of the same user on different anonymous sites. Furthermore, cross-site consistency comparison constraints are used to suppress relevant noise in the common behavior feature space.

[0024] As a preferred embodiment of the cross-site identity authentication method based on trust continuity and behavioral evolution described in this invention, the specific steps of S3 are as follows:

[0025] S31: Construct the Identity Structure Consistency Alignment (ISCAM) module;

[0026] S32: Referencing the distribution of historical identity structures and bi-branch feature representation , Inputting the data into the ISCAM module yields the consistency deviation of the structural deviation constraint information. and structural deviation direction constraints ; and based on the Alignment and The cross-site identity feature representation modulated by identity structure consistency constraints is obtained. ;

[0027] S32 utilizes the Identity Structure Consistency Alignment (ISCAM) module to perform consistency correction on the current cross-site identity features, which includes the following specific steps:

[0028] S321: Represent the bi-branch feature and They are all mapped to a reference space defined by the user's historical identity structure, combined with the user's... In the Cross-site identity feature representation of the fusion of historical observation moments Construct a reference distribution of historical identity structures Achieve consistent alignment;

[0029] S322: Regarding the above The resulting state jump increment Constructing the jump-induced uncertainty measurement matrix , get in Under the constraints of the current observation point Obtained identity status support points With the Support points Consistency deviation between Then through The optimal transmission solution yields the structural deviation direction of the current state pointing towards the historical reference center. ;

[0030] S323: The aforementioned and and By performing directional consistency alignment and weighted fusion to suppress feature components inconsistent with historical identity structures, a cross-site identity feature representation modulated by identity structure consistency constraints is obtained. .

[0031] As a preferred embodiment of the cross-site identity authentication method based on trust continuity and behavioral evolution described in this invention, the specific steps of S4 are as follows:

[0032] S41: Design the AHGM module by generating the associated assumptions of multi-task constraints;

[0033] S42: Deviate from the consistency of S32 Input into the AHGM module, when When the threshold is exceeded, multiple candidate cross-site identity association hypotheses are obtained;

[0034] S42 generates the AHGM module using the association assumption of multi-task constraints, realizing the cross-site identity association assumption under the joint constraints of identity structure consistency and behavioral evolution. The specific steps are as follows:

[0035] S421: The above Cross-site identity features exceeding a preset threshold are represented as a set of conflict features, and the set of conflict features is mapped to a dynamic evolutionary manifold space defined by the behavioral evolution model to uniformly characterize the evolutionary constraints of identity features in the continuous time dimension.

[0036] S422: For any candidate cross-site identity association path Calculate the consistency cost of static identity attributes respectively. The cost of consistency between behavioral content and style evolution Consistency cost of interest evolution distribution ;

[0037] S423: Encode each search entity in the optimized search algorithm as a candidate cross-site identity association path. and make the random walk step size Deviation from consistency Adaptive association; when a user's behavior sequence is interrupted due to prolonged offline time, random walks are used to resolve the issue. Achieve probabilistic jumps across unobserved time intervals and establish potential identity association paths between discontinuous time nodes;

[0038] S424: Regarding the above The consistency score is calculated by inputting it into the multi-task consistency cost, and then filtered and ranked by the evolution span of candidate paths over time to obtain a set of multi-path cross-site identity association hypotheses. .

[0039] As a preferred embodiment of the cross-site identity authentication method based on trust continuity and behavioral evolution described in this invention, the specific steps of S5 are as follows:

[0040] S51: Design a Trust Continuity Backtracking Verification (TCC-BVM) module;

[0041] S52: Under the constraints of path-level identity feature evolution and time continuity, trust continuity assessment, adaptive threshold decision and backtracking correction are performed on candidate identity association paths;

[0042] S52 utilizes the Trust Continuity Backtracking Verification TCC-BVM module to perform path-level verification of candidate cross-site identity association paths, which includes the following specific steps:

[0043] S521: For any candidate cross-site identity association path Reconstructing the path-level identity feature sequence in chronological order In the Construct the corresponding covariance matrix under the induction of ; Regarding the Constructing the directional projection operator Local trust continuity deviation in the distribution of identity features of adjacent stations along the calculation path ; and combine the perceived weights based on the time span between adjacent stations in the path. Regarding the Weighted aggregation is used, and a path-level decision threshold that adaptively varies with the deviation from the consistency of the identity structure is introduced. The path-level trust continuity scoring function is obtained. ;

[0044] S522: According to the above For candidate paths Perform path-level structural consistency backtracking verification to obtain the set of paths that satisfy the trust continuity constraint. .

[0045] A cross-site authentication system based on trust continuity and behavioral evolution includes:

[0046] The dual-branch identity feature module is used to perform heterogeneous dual-branch modeling of user identity information from different anonymous websites, so as to achieve collaborative representation of user identity stability features and behavioral evolution features.

[0047] The ISCAM module is used to evaluate cross-site identity features and the reference distribution of user historical identity structure. The degree of consistency deviation between them, and the degree of consistency deviation in generating structural deviation constraint information. and structural deviation direction constraints and in structural deviation constraints The following is a representation of identity attribute features. Representation of identity and behavioral characteristics Alignment is performed to obtain a cross-site identity feature representation modulated by identity structure consistency constraints. ;

[0048] The AHGM module is used to detect deviations in identity structure consistency. When the threshold is exceeded, multiple candidate cross-site identity association hypotheses are generated under multi-task constraints;

[0049] The TCC-BVM module, based on feedback from the consistency deviation output of the identity structure consistency alignment module, combines path-level identity feature evolution and temporal continuity constraints to evaluate the trust continuity of candidate cross-site identity association hypotheses, calculates a path-level trust continuity score, and incorporates subsequent... Adaptive decision threshold The hypothesis of candidate identity association is backtracked for verification and correction, so as to output the cross-site user identity association result that simultaneously satisfies the trust continuity constraint and the identity structure stability.

[0050] Compared with existing technologies:

[0051] 1. This invention first jointly models the continuous evolution characteristics of static identity attributes and user behavior, then introduces historical identity structure consistency constraints, and combines multiple hypothesis association generation and backtracking verification mechanisms in the case of identity conflict to achieve stable association and determination of cross-site user identities;

[0052] 2. This invention improves the ability of existing cross-site identity authentication technologies to characterize the continuous evolution of user identities and assess the consistency of identity structure. On the other hand, it improves the stability of identity determination results in scenarios with conflicting identity features, thereby enhancing the reliability and practicality of cross-site identity authentication in anonymous network environments. Attached Figure Description

[0053] Figure 1 The main flowchart of a cross-site identity authentication method based on trust continuity and behavioral evolution;

[0054] Figure 2 A flowchart illustrating a cross-site identity authentication method that considers trust continuity and behavioral evolution.

[0055] Figure 3 A schematic diagram of the structure of a static identity attribute feature modeling branch for a cross-site identity authentication method that considers trust continuity and behavioral evolution;

[0056] Figure 4 A schematic diagram of the structure of the AHGM module, which is a cross-site identity authentication method based on trust continuity and behavioral evolution;

[0057] Figure 5 This is a schematic diagram of the TCC-BVM module, which is a cross-site identity authentication method based on trust continuity and behavioral evolution. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0059] This invention provides a cross-site identity authentication method based on trust continuity and behavioral evolution. Please refer to [link / reference]. Figures 1-5 This includes the following steps:

[0060] S1: Collect user-related data from multiple anonymous websites, standardize and process user attribute information and user-generated content in a unified manner, and eliminate differences in data format, expression method and time scale among different websites;

[0061] S2: Construct a static identity attribute feature modeling branch and a behavioral feature evolution modeling branch to perform static-dynamic heterogeneous dual-branch modeling of user identity features on different sites;

[0062] The specific steps of S2 are as follows:

[0063] S21: Input the username and personal profile into the static identity attribute feature modeling branch to obtain the identity attribute representation that reflects the stability of the user's identity. ;

[0064] The system models user identity from the perspective of identity stability. It takes the usernames and personal profiles used by users on various anonymous sites as input and introduces a static identity attribute modeling branch to characterize the user's long-term identity features. Specifically, the system encodes and concatenates the usernames and personal profiles to obtain a high-dimensional static attribute semantic vector. Then, by introducing a variational inference decoupling operator constrained by information bottlenecks, the system decomposes the semantic vector into core identity components representing the user's essential personality preferences and long-term identity characteristics. And environmental noise components that characterize differences in site expression style, contextual habits, and interactive environment. To improve the alignability of identity representation in cross-site scenarios, the system... Perform discretization mapping to obtain quantized static identity representation and the and The data is fed into an adversarial learning network to suppress environmental noise interference and obtain an environmentally invariant static identity fingerprint representation. Based on this, the system's primary optimization objective is to minimize the potential distribution differences in the identity of the same user across different anonymous sites. By aligning the static identity fingerprint distribution level, a static identity attribute representation reflecting the long-term stability of a user's identity is obtained. .

[0065] The specific steps of S21 are as follows:

[0066] S211: Input the text set consisting of the user's username and personal profile into the cross-language pre-trained semantic encoding model for unified semantic mapping, and obtain a high-dimensional static attribute semantic vector. ;

[0067] S212: Under the constraint of information bottleneck, for the... By performing variational inference and information decoupling modeling, we can obtain the core identity components that represent the user's essential personality preferences. and the environmental noise component that characterizes the expressive style and contextual bias of a site ;

[0068] S213: Regarding the above Discretization mapping, in the global codebook Select the cluster center vector with the smallest distance from it. Quantified static identity representation is obtained :

[0069]

[0070] At the same time, the above and Input adversarial learning network, suppress Despite interference with identity discrimination features, the output environment remains unchanged, resulting in a static identity fingerprint representation. ;

[0071] S214: Static identity fingerprints of the same user across different sites Model as a distribution of latent identity variables Furthermore, a commitment loss constraint is introduced, with minimizing the distribution of cross-site identity latent variables as the primary optimization objective. This achieves cross-site alignment of static identity attributes at the distribution level, ultimately outputting a steady-state identity attribute representation. ;

[0072]

[0073]

[0074] in, To stop the gradient operator, Weighting for commitment loss;

[0075] S22: Input user-generated content, behavioral time information, and contextual information into the behavioral feature evolution modeling branch to obtain an identity behavior representation that reflects the evolution of user behavior patterns and content style characteristics. ;

[0076] Simultaneously, the system models user identity from a behavioral evolution perspective, introducing a behavioral feature evolution modeling branch to characterize user behavior patterns across different anonymous sites and their changes over time. Specifically, the system encodes user-generated content, the time of behavior occurrence, and contextual information to obtain a sequence of behavioral state vectors at discrete time points. Furthermore, continuous-time evolution modeling is performed on the data to obtain a continuous behavioral trajectory representation that depicts the overall evolution trend of user cross-site behavior. The evolutionary process simultaneously characterizes the deterministic trend of behavioral changes, random perturbations, and state transitions caused by site migration or sudden events, adapting to non-uniform sampling and behavioral interruption scenarios. Subsequently, the system maps the continuous behavioral trajectories to a common behavioral feature space and constructs a behavioral feature consistency optimization objective by introducing cross-site consistency constraints. This approach aims to suppress behavioral noise, enhance the consistency of cross-site behavioral representations for the same user, and ultimately obtain a stable identity behavior representation that reflects the evolution of user behavior patterns and content style. .

[0077] The specific steps of S22 are as follows:

[0078] S221: Perform content semantic encoding, time period encoding, and context association encoding on the discrete behaviors of users across multiple anonymous websites to obtain user... Behavioral state vector sequence at discrete time points ;

[0079] S222: The above In the continuous-time behavior evolution model with input structure consistency constraints, the user's cross-site behavior over continuous time intervals... Internal modeling yields a representation of the behavioral evolution trajectory. :

[0080]

[0081] in, This represents the user's behavioral state vector at the initial observation time. This represents a deterministic dynamic function that characterizes the evolution of user behavior. Represents the evolution constraint matrix. The deterministic drift function representing the behavioral evolution, Represents the random diffusion intensity function. Represents a continuous-time random noise process. This indicates the time when a user migrates across sites or when an unexpected event occurs. The resulting behavioral state jump increment;

[0082] S223: The above Mapping to the public behavior feature space yields the identity and behavior feature representations of the same user on different anonymous sites. ; and by suppressing relevant noise in the common behavior feature space through cross-site consistency comparison constraints, its optimization objective function is defined as:

[0083]

[0084] In the above formula, Represents a set of behavioral feature samples. This represents the behavioral characteristics of the same user across different anonymous sites. Represents the sample set The identity behavior characteristics of other users or other candidate identities are represented. This represents the similarity measurement function. This represents the temperature coefficient.

[0085] S3: Construct the Identity Structure Consistency Alignment (ISCAM) module to assess the degree of consistency deviation between cross-site identity features and the user's historical identity structure reference distribution, and generate cross-site identity feature representations;

[0086] In obtaining static identity attribute representation Representation of identity behavior Subsequently, an Identity Structure Consistency Alignment (ISCAM) module is constructed to perform consistency correction between current cross-site identity features and the user's historical identity structure. Specifically, the system utilizes the historical identity structure reference distribution constructed from the user's identity features formed over multiple historical trusted time intervals. And assign a weight that decays over time to each historical observation point; when a user's historical reliable observation points are insufficient, the... It degenerates into a reference distribution that introduces a global identity structure as a priori supplement. Subsequently, the system uses the currently observed... and Mapped to the aforementioned reference space, in the behavioral evolution uncertainty measurement matrix Under constraints, calculate the overall consistency deviation of the current identity state relative to the historical identity structure. At the same time, determine the corresponding structural deviation direction constraints. Subsequently, the system deviates along the direction of the structure. and Weighted alignment and fusion are performed to suppress feature components inconsistent with historical identity structures, resulting in a cross-site identity feature representation modulated by identity structure consistency constraints. .

[0087] The specific steps of S3 are as follows:

[0088] S31: Construct the Identity Structural-ConsistencyAlignment (ISCAM) module;

[0089] S32: Referencing the distribution of historical identity structures and bi-branch feature representation , Inputting the data into the ISCAM module yields the consistency deviation of the structural deviation constraint information. and structural deviation direction constraints ; and based on the Alignment and The cross-site identity feature representation modulated by identity structure consistency constraints is obtained. ;

[0090] S32 utilizes the Identity Structure Consistency Alignment (ISCAM) module to perform consistency correction on the current cross-site identity features, which includes the following specific steps:

[0091] S321: Represent the bi-branch feature and They are all mapped to a reference space defined by the user's historical identity structure, combined with the user's... In the Cross-site identity feature representation of the fusion of historical observation moments Construct a reference distribution of historical identity structures Achieving consistent alignment:

[0092]

[0093] In the above formula, These are normalized weighting coefficients determined based on time decay or identity credibility. To characterize a discrete set of reference anchor points for historical identity features;

[0094] S322: Regarding the above The resulting state jump increment Constructing the jump-induced uncertainty measurement matrix , get in Under the constraints of the current observation point Obtained identity status support points With the Support points Consistency deviation between Then through The optimal transmission solution yields the structural deviation direction of the current state pointing towards the historical reference center.

[0095]

[0096]

[0097]

[0098] In the above formula, Regularization coefficients to ensure the positive definiteness and invertibility of a matrix; It is the identity matrix; The variable is the optimal transmission quality variable, and its row sum and column sum are equal to the probability quality of the corresponding distribution, respectively. The optimal transmission solution obtained through calculation;

[0099] S323: The aforementioned and and By performing directional consistency alignment and weighted fusion to suppress feature components inconsistent with historical identity structures, a cross-site identity feature representation modulated by identity structure consistency constraints is obtained. :

[0100]

[0101] In the above formula, The adaptive fusion weights are for the dual-branch features.

[0102] S4: Construct the AHGM module to generate association hypotheses with multi-task constraints, so as to obtain multiple candidate cross-site identity association hypotheses after identity structure conflict occurs;

[0103] When the overall consistency deviation obtained by the identity structure consistency alignment step When the threshold is exceeded, the system determines that the current cross-site identity association result is uncertain and constructs a multi-task constrained association hypothesis generation (AHGM) module to generate multiple candidate identity association paths. Specifically, under the constraints of identity feature evolution and temporal continuity, the system maps the current cross-site identity features to the behavior evolution space, generating multiple candidate cross-site identity association paths representing the possible association relationships between users across different sites or at different times. For any candidate path, the system calculates its static identity attribute consistency cost. The cost of consistency between behavioral content and style evolution Consistency cost of interest evolution distribution The aforementioned multi-task consistency cost is used as a joint optimization objective; simultaneously, the search step size is... and Adaptive association, using a horned lizard optimization search mechanism to iteratively update candidate paths. This system aims to establish potential identity association paths between discontinuous time points in identity structure conflict scenarios, thereby covering possible cross-site identity associations while maintaining the rationality of temporal evolution. Through consistency scoring and screening of candidate paths, the system ultimately obtains a set of multi-path cross-site identity association hypotheses for subsequent trust continuity backtracking verification. .

[0104] The specific steps of S4 are as follows:

[0105] S41: Design an AHGM (Multi-Task constrained association hypothesis generation model) module for generating association hypotheses for multi-task constraints;

[0106] S42: Deviate from the consistency of S32 Input into the AHGM module, when When the threshold is exceeded, multiple candidate cross-site identity association hypotheses are obtained;

[0107] S42 generates the AHGM module using the association assumption of multi-task constraints, realizing the cross-site identity association assumption under the joint constraints of identity structure consistency and behavioral evolution. The specific steps are as follows:

[0108] S421: The above Cross-site identity features exceeding a preset threshold are represented as a set of conflict features, and the set of conflict features is mapped to a dynamic evolutionary manifold space defined by the behavioral evolution model to uniformly characterize the evolutionary constraints of identity features in the continuous time dimension.

[0109] S422: For any candidate cross-site identity association path Calculate the consistency cost of static identity attributes respectively. The cost of consistency between behavioral content and style evolution Consistency cost of interest evolution distribution :

[0110]

[0111]

[0112]

[0113] In the above formula, Indicates in the global codebook The code vector that has the smallest distance to the input features; For behavioral evolution models in state The predicted behavior state at the next moment; Indicates user Behavioral evolution trajectory samples across multiple historical reliable time intervals The distribution of identity evolution trajectories obtained by statistical estimation;

[0114] S423: Encode each search entity in the optimized search algorithm as a candidate cross-site identity association path. and make the random walk step size Deviation from consistency Adaptive association; when a user's behavior sequence is interrupted due to prolonged offline time, random walks are used to resolve the issue. Achieve probabilistic jumps across unobserved time intervals and establish potential identity association paths between discontinuous time nodes; among which, in the first... Random walk step size in the next iteration and candidate path update rules for:

[0115]

[0116]

[0117] In the above formula, This is the maximum upper limit of the random walk step size, used to limit the scale of the search space; To smooth the adjustment factor, avoid When the step size is small, the step size is too small or the values ​​are unstable. The mean is zero and the covariance is A multidimensional Gaussian random vector; For the first The gradient update coefficients for the next iteration; This represents the joint gradient operator with respect to each identity feature node in the path;

[0118] S424: Regarding the above The consistency score is calculated by inputting it into the multi-task consistency cost, and then filtered and ranked by the evolution span of candidate paths over time to obtain a set of multi-path cross-site identity association hypotheses. .

[0119] S5: Under the structural deviation constraint feedback of the ISCAM module, construct the Trust Continuity Backtracking Verification TCC-BVM module to realize the backtracking verification and correction of candidate cross-site identity association assumptions;

[0120] The system constructs a Trust Continuity Backtracking Verification (TCC-BVM) module to verify the candidate path set. Verification is performed. Specifically, the system reconstructs the path-level identity feature sequence of any candidate identity association path in chronological order. And combined with the uncertainty matrix of user behavior evolution In the above Constructing projection operators under constraints Assess the continuity between adjacent identity states and calculate the local trust continuity deviation at each migration stage. Based on this, the system introduces time span-aware weights. The local continuity deviations are weighted and aggregated to obtain the path-level trust continuity score. Subsequently, the system adaptively determines the decision threshold based on the current identity structure consistency deviation. The candidate paths are filtered, and only the set of paths that meet the trust continuity constraint is retained. If the above If the result is empty or the overall score is below the preset threshold, the system triggers the Association Hypothesis Generation (AHGM) module to regenerate candidate identity association paths by increasing the search step size.

[0121] The specific steps of S5 are as follows:

[0122] S51: Design the Trust Continuity-Constrained Backtracking Verification Model (TCC-BVM) module;

[0123] S52: Under the constraints of path-level identity feature evolution and time continuity, trust continuity assessment, adaptive threshold decision and backtracking correction are performed on candidate identity association paths;

[0124] S52 utilizes the Trust Continuity Backtracking Verification TCC-BVM module to perform path-level verification of candidate cross-site identity association paths, which includes the following specific steps:

[0125] S521: For any candidate cross-site identity association path Reconstructing the path-level identity feature sequence in chronological order In the Construct the corresponding covariance matrix under the induction of ; Regarding the Constructing the directional projection operator Local trust continuity deviation in the distribution of identity features of adjacent stations along the calculation path ; and combine the perceived weights based on the time span between adjacent stations in the path. Regarding the Weighted aggregation is used, and a path-level decision threshold that adaptively varies with the deviation from the consistency of the identity structure is introduced. The path-level trust continuity scoring function is obtained. :

[0126]

[0127]

[0128]

[0129]

[0130] In the above formula, Indicates identity status The local linearization evolution sensitivity matrix; Represents the second-order Wasserstein distance; Indicates the time span between adjacent stations in the path;

[0131] S522: According to the above For candidate paths Perform path-level structural consistency backtracking verification to obtain the set of paths that satisfy the trust continuity constraint. .

[0132] S6: Output cross-site user identity association results that simultaneously satisfy trust continuity constraints and identity structure stability, realizing cross-site identity authentication in an anonymous network environment;

[0133] When the set of paths satisfies the unique path constraint and the structural stability determination result converges, the system will use the corresponding path as the final cross-site user authentication result. Output: If multiple paths have similar scores and all exceed the decision threshold, the system will continue to collect user behavior data by extending the observation time window until the identity authentication result meets the convergence condition.

[0134] The specific steps of S6 are as follows:

[0135] When the path set When the unique path constraint is satisfied and the structural verification results converge and stabilize, the output path is... For the final cross-site user authentication result:

[0136]

[0137] In the above formula, for The overall residual under the identity structure consistency constraint; This is a preset structural stability threshold.

[0138] A cross-site authentication system based on trust continuity and behavioral evolution includes:

[0139] The dual-branch identity feature module is used to perform heterogeneous dual-branch modeling on user identity information from different anonymous websites, achieving a collaborative representation of user identity stability features and behavioral evolution features. The dual-branch module consists of a static identity attribute feature modeling branch and a behavioral feature evolution modeling branch. The static identity attribute feature modeling is used to extract identity attribute feature representations that reflect user identity stability. Behavioral feature evolution modeling is used to extract identity behavioral feature representations that reflect the evolution of user behavior patterns and content styles. ;

[0140] The ISCAM module is used to evaluate cross-site identity features and the reference distribution of user historical identity structure. The degree of consistency deviation between them, and the degree of consistency deviation in generating structural deviation constraint information. and structural deviation direction constraints and in structural deviation constraints The following is a representation of identity attribute features. Representation of identity and behavioral characteristics Alignment is performed to obtain a cross-site identity feature representation modulated by identity structure consistency constraints. ;

[0141] The AHGM module is used to detect deviations in identity structure consistency. When the threshold is exceeded, multiple candidate cross-site identity association hypotheses are generated under multi-task constraints;

[0142] The TCC-BVM module, based on feedback from the consistency deviation output of the identity structure consistency alignment module, combines path-level identity feature evolution and temporal continuity constraints to evaluate the trust continuity of candidate cross-site identity association hypotheses, calculates a path-level trust continuity score, and incorporates subsequent... Adaptive decision threshold The hypothesis of candidate identity association is backtracked for verification and correction, so as to output the cross-site user identity association result that simultaneously satisfies the trust continuity constraint and the identity structure stability.

[0143] Although the present invention has been described above with reference to embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of the invention. In particular, as long as there is no structural conflict, the features in the disclosed embodiments can be combined with each other in any manner. The lack of an exhaustive description of these combinations in this specification is merely for the sake of brevity and resource conservation. Therefore, the present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A cross-site identity authentication method based on trust continuity and behavioral evolution, characterized in that, Includes the following steps: S1: Collect user-related data from multiple anonymous websites, standardize and process user attribute information and user-generated content in a unified manner, and eliminate differences in data format, expression method and time scale among different websites; S2: Construct a static identity attribute feature modeling branch and a behavioral feature evolution modeling branch to perform static-dynamic heterogeneous dual-branch modeling of user identity features on different sites; S3: Construct the Identity Structure Consistency Alignment (ISCAM) module to assess the degree of consistency deviation between cross-site identity features and the user's historical identity structure reference distribution, and generate cross-site identity feature representations; S4: Construct the AHGM module to generate association hypotheses with multi-task constraints, so as to obtain multiple candidate cross-site identity association hypotheses after identity structure conflict occurs; S5: Under the structural deviation constraint feedback of the ISCAM module, construct the Trust Continuity Backtracking Verification TCC-BVM module to realize the backtracking verification and correction of candidate cross-site identity association assumptions; S6: Output cross-site user identity association results that simultaneously satisfy trust continuity constraints and identity structure stability, realizing cross-site identity authentication in an anonymous network environment.

2. The cross-site identity authentication method based on trust continuity and behavioral evolution according to claim 1, characterized in that, The specific steps of S2 are as follows: S21: Input the username and personal profile into the static identity attribute feature modeling branch to obtain the identity attribute representation that reflects the stability of the user's identity. ; S22: Input user-generated content, behavioral time information, and contextual information into the behavioral feature evolution modeling branch to obtain an identity behavior representation that reflects the evolution of user behavior patterns and content style characteristics. .

3. The cross-site identity authentication method based on trust continuity and behavioral evolution according to claim 2, characterized in that, The specific steps of S21 are as follows: S211: Input the text set consisting of the user's username and personal profile into the cross-language pre-trained semantic encoding model for unified semantic mapping, and obtain a high-dimensional static attribute semantic vector. ; S212: Under the constraint of information bottleneck, for the aforementioned By performing variational inference and information decoupling modeling, we can obtain the core identity components that represent the user's essential personality preferences. and the environmental noise component that characterizes the expressive style and contextual bias of a site ; S213: Regarding the above Discretization mapping, in the global codebook Select the cluster center vector with the smallest distance from it. Quantified static identity representation is obtained At the same time, the aforementioned and Input adversarial learning network, suppress Despite interference with identity discrimination features, the output environment remains unchanged, resulting in a static identity fingerprint representation. ; S214: Static identity fingerprints of the same user across different sites Model as a distribution of latent identity variables Furthermore, a commitment loss constraint is introduced, with minimizing the distribution of cross-site identity latent variables as the primary optimization objective. This achieves cross-site alignment of static identity attributes at the distribution level, ultimately outputting a steady-state identity attribute representation. ; The specific steps of S22 are as follows: S221: Perform content semantic encoding, time period encoding, and context association encoding on the discrete behaviors of users across multiple anonymous websites to obtain user... Behavioral state vector sequence at discrete time points ; S222: The above In the continuous-time behavior evolution model with input structure consistency constraints, the user's cross-site behavior over continuous time intervals... Internal modeling yields a representation of the behavioral evolution trajectory. ; S223: The above Mapping to the public behavior feature space yields the identity and behavior feature representations of the same user on different anonymous sites. Furthermore, cross-site consistency comparison constraints are used to suppress relevant noise in the common behavior feature space.

4. The cross-site identity authentication method based on trust continuity and behavioral evolution according to claim 3, characterized in that, The specific steps of S3 are as follows: S31: Construct the Identity Structure Consistency Alignment (ISCAM) module; S32: Referencing the distribution of historical identity structures and bi-branch feature representation , Inputting the data into the ISCAM module yields the consistency deviation of the structural deviation constraint information. and structural deviation direction constraints ; and based on the above Alignment and The cross-site identity feature representation modulated by identity structure consistency constraints is obtained. ; S32 utilizes the Identity Structure Consistency Alignment (ISCAM) module to perform consistency correction on the current cross-site identity features, which includes the following specific steps: S321: Represent the bi-branch feature and They are all mapped to a reference space defined by the user's historical identity structure, combined with the user's... In the Cross-site identity feature representation of the fusion of historical observation moments Construct a reference distribution of historical identity structures Achieve consistent alignment; S322: Regarding the above The resulting state jump increment Constructing the jump-induced uncertainty measurement matrix , get in Under the constraints of the current observation point Obtained identity status support points With the Support points Consistency deviation between Then through The optimal transmission solution yields the structural deviation direction of the current state pointing towards the historical reference center. ; S323: The aforementioned and and By performing directional consistency alignment and weighted fusion to suppress feature components inconsistent with historical identity structures, a cross-site identity feature representation modulated by identity structure consistency constraints is obtained. .

5. The cross-site identity authentication method based on trust continuity and behavioral evolution according to claim 4, characterized in that, The specific steps of S4 are as follows: S41: Design the AHGM module by generating the associated assumptions of multi-task constraints; S42: Deviate from the consistency of S32 Input into the AHGM module, when When the threshold is exceeded, multiple candidate cross-site identity association hypotheses are obtained; S42 generates the AHGM module using the association assumption of multi-task constraints, realizing the cross-site identity association assumption under the joint constraints of identity structure consistency and behavioral evolution. The specific steps are as follows: S421: The above Cross-site identity features exceeding a preset threshold are represented as a set of conflict features, and the set of conflict features is mapped to a dynamic evolutionary manifold space defined by the behavioral evolution model to uniformly characterize the evolutionary constraints of identity features in the continuous time dimension. S422: For any candidate cross-site identity association path Calculate the consistency cost of static identity attributes respectively. The cost of consistency between behavioral content and style evolution Consistency cost of interest evolution distribution ; S423: Encode each search entity in the optimized search algorithm as a candidate cross-site identity association path. and make the random walk step size Deviation from consistency Adaptive association; when a user's behavior sequence is interrupted due to prolonged offline time, random walks are used to resolve the issue. Achieve probabilistic jumps across unobserved time intervals and establish potential identity association paths between discontinuous time nodes; S424: Regarding the above The consistency score is calculated by inputting it into the multi-task consistency cost, and then filtered and ranked by the evolution span of candidate paths over time to obtain a set of multi-path cross-site identity association hypotheses. .

6. The cross-site identity authentication method based on trust continuity and behavioral evolution according to claim 5, characterized in that, The specific steps of S5 are as follows: S51: Design a Trust Continuity Backtracking Verification (TCC-BVM) module; S52: Under the constraints of path-level identity feature evolution and time continuity, trust continuity assessment, adaptive threshold decision and backtracking correction are performed on candidate identity association paths; S52 utilizes the Trust Continuity Backtracking Verification TCC-BVM module to perform path-level verification of candidate cross-site identity association paths, which includes the following specific steps: S521: For any candidate cross-site identity association path Reconstructing the path-level identity feature sequence in chronological order In the Construct the corresponding covariance matrix under the induction of ; Regarding the Constructing the directional projection operator Local trust continuity deviation in the distribution of identity features of adjacent stations along the calculation path ; and combine the perceived weights based on the time span between adjacent stations in the path. Regarding the Weighted aggregation is used, and a path-level decision threshold that adaptively varies with the deviation from the consistency of the identity structure is introduced. The path-level trust continuity scoring function is obtained. ; S522: According to the above For candidate paths Perform path-level structural consistency backtracking verification to obtain the set of paths that satisfy the trust continuity constraint. .

7. A cross-site identity authentication system based on trust continuity and behavioral evolution, characterized in that, include: The dual-branch identity feature module is used to perform heterogeneous dual-branch modeling of user identity information from different anonymous websites, so as to achieve collaborative representation of user identity stability features and behavioral evolution features. The ISCAM module is used to evaluate cross-site identity features and the reference distribution of user historical identity structure. The degree of consistency deviation between them, and the degree of consistency deviation in generating structural deviation constraint information. and structural deviation direction constraints and in structural deviation constraints The following is a representation of identity attribute features. Representation of identity and behavioral characteristics Alignment is performed to obtain a cross-site identity feature representation modulated by identity structure consistency constraints. ; The AHGM module is used to detect deviations in identity structure consistency. When the threshold is exceeded, multiple candidate cross-site identity association hypotheses are generated under multi-task constraints; The TCC-BVM module, based on feedback from the consistency deviation output of the identity structure consistency alignment module, combines path-level identity feature evolution and temporal continuity constraints to evaluate the trust continuity of candidate cross-site identity association hypotheses, calculates a path-level trust continuity score, and incorporates subsequent... Adaptive decision threshold The hypothesis of candidate identity association is backtracked for verification and correction, so as to output the cross-site user identity association result that simultaneously satisfies the trust continuity constraint and the identity structure stability.

Images