CN-115937912-B - Behavior fingerprint digital feature extraction method, system, medium and equipment

CN115937912BCN 115937912 BCN115937912 BCN 115937912BCN-115937912-B

Abstract

The invention discloses a behavior fingerprint digital characteristic extraction method, a system, a medium and equipment, wherein the method comprises the following steps of acquiring a behavior fingerprint sequence of a user, extracting digital characteristic information of the user, respectively constructing a digital characteristic vector matrix for a user set to be matched and a known user set, and extracting a characteristic vector set of the user for the acquired behavior fingerprint sequence of the user; calculating the similarity distance between the known user digital feature vector and the user digital feature vector to be matched based on a similarity distance formula; and judging the identity of the user according to a preset matching strategy. The invention expands the behavior fingerprint of the user, solves the sparsity problem of the behavior fingerprint of the user to a certain extent, and achieves the effect of improving the recognition accuracy.

Inventors

YANG CAN
LI JIAHAO
ZHU YINGYING

Assignees

华南理工大学

Dates

Publication Date: 20260508
Application Date: 20221229

Claims (9)

1. The behavioral fingerprint digital characteristic extraction method is characterized by comprising the following steps: acquiring a behavior fingerprint sequence of a user, extracting digital characteristic information of the user, respectively constructing a digital characteristic vector matrix for a user set to be matched and a known user set, and extracting a characteristic vector set of the user for the acquired behavior fingerprint sequence of the user; the specific steps of acquiring the behavior fingerprint sequence of the user and extracting the digital characteristic information of the user include: Extracting an n-order feature vector set of the user for the acquired behavior fingerprint sequence S u of the user u The n-order feature vectors of all users are collected and combined, and the occurrence frequency of each digital feature vector is counted and combined to generate a tuple Each element in the tuple is composed of a digital feature vector and its frequency of occurrence in the tuple; According to tuples The frequency of each digital characteristic vector in the array is ordered from big to small to form ordered tuples, and the ordered tuples are recorded as From ordered tuples Extracting feature component tuples of set proportions Fusing n-order feature vector sets of all users to obtain a behavior frequency matrix comprising all users and digital feature vectors thereof; Calculating the similarity distance between the known user digital feature vector and the user digital feature vector to be matched based on a similarity distance formula; And judging the identity of the user according to a preset matching strategy.
2. The behavioral fingerprint digital characteristic extraction method according to claim 1, wherein the behavioral fingerprint sequence of the user is derived from log records of various systems, part of log records are extracted to form data, a time point is selected, and the data set is cut into a user set to be matched and a known user set according to the time point.
3. The behavioral fingerprint digital characteristic extraction method according to claim 1, wherein the n-order feature vector set of the user is extracted based on FaG method The method comprises the following specific steps: Creating a time sliding window S n , and sliding on a behavior fingerprint sequence S u of the user u to obtain a behavior state; Constructing a directed graph Wherein, the Vertex set contained in window S n representing the ith slide of user u on behavioural fingerprint sequence S u , e is a directed graph Is a set of edges; extracting directed graphs Two-dimensional digital feature vectors in (a); Sliding window S n slides over behavioral fingerprint sequence S u of length l u , producing m graphs, constituting the nth order feature atlas of user u Acquiring an n-order feature vector set formed by feature vectors of a user u from the 1 st order to the n-th order The concrete steps are as follows: Where m=l-n+1.
4. The behavioral fingerprint digital feature extraction method according to claim 1, wherein calculating a similarity distance between a known user digital feature vector and a user digital feature vector to be matched based on a similarity distance formula specifically comprises: The similarity distance formula is specifically expressed as: where U x denotes the user to be matched, U y denotes the known user, Representing a matrix of feature vectors of known users, A behavior probability vector representing an unknown user, Representing the probability matrices of behavior of all known users, D (||·) represents the KL divergence between the two vector feature vectors.
5. The behavioral fingerprint digital characteristic extraction method according to claim 1, wherein the matching policy includes a maximum similarity matching policy and a bipartite graph matching policy.
6. The behavioral fingerprint digital characteristic extraction method according to claim 5, wherein the maximum similarity matching strategy is specifically expressed as: Wherein, the Representing predicted user identity if Then it indicates that the predicted user identity is the same as its actual identity, the prediction result is correct, w ij represents the similarity distance between the two user feature vectors, and U y represents the set of known users; The bipartite graph matching strategy is specifically expressed as: The setting conditions are as follows: Where U x is the unknown user set, E represents the edge of the bipartite graph match, Representing the identity of the predicted set of users of the unknown user.
7. The behavior fingerprint digital characteristic extraction system is characterized by being used for realizing the behavior fingerprint digital characteristic extraction method according to any one of claims 1-6, and comprises a behavior fingerprint sequence acquisition module, a digital characteristic information extraction module, a similarity distance calculation module and a user judgment module; the behavior fingerprint sequence acquisition module is used for acquiring a behavior fingerprint sequence of a user; The digital characteristic information extraction module is used for extracting digital characteristic information of a user, respectively constructing a digital characteristic vector matrix for a user set to be matched and a known user set, and extracting a characteristic vector set of the user for the acquired behavior fingerprint sequence of the user; the similarity distance calculation module is used for calculating the similarity distance between the known user digital feature vector and the user digital feature vector to be matched based on a similarity distance formula; The user judging module is used for judging the identity of the user according to a preset matching strategy.
8. A computer-readable storage medium comprising a stored program, characterized in that the program when executed implements the behavioral fingerprint digital feature extraction method according to any one of claims 1-6.
9. A computing device comprising a processor and a memory for storing a processor-executable program, wherein the processor, when executing the program stored in the memory, implements the behavioral fingerprint digital feature extraction method of any one of claims 1-6.

Description

Behavior fingerprint digital feature extraction method, system, medium and equipment Technical Field The invention relates to the technical field of user identification, in particular to a behavior fingerprint digital feature extraction method, a system, a medium and equipment. Background At present, as the binding of human life and various scientific and technological products is deeper and deeper, the safety problem becomes the problem which needs to be solved urgently at present, and one important field is the identity recognition technology. The conventional identity recognition technology is mainly based on biological physiological characteristics, including fingerprints, faces, voiceprints and the like, the authentication mode is usually performed at a certain time point of program operation and needs biological active participation, meanwhile, with the continuous development of the anti-network technology, the one-time authentication mode also faces greater challenges, while the authentication mode based on the user behavior fingerprints, such as the network television watching behavior, keyboard knocking behavior, web browsing behavior and the like of the user, does not need the active participation of the user, and the implicit acquisition of the user behavior by the system, and can continuously perform identity authentication as a supplement to the existing identity recognition technology to enhance the security of the system. At present, a basic scheme for carrying out identity recognition based on user behavior fingerprints is to calculate the frequency of each behavior in the user behavior fingerprints to form a digital feature vector of a user, and judge the identity of the user by calculating the similarity of the digital feature vectors between the known user and the unknown user, but the behavior fingerprints are often sparse, and a researcher can achieve the effect of enhancing the recognition accuracy by splicing a plurality of continuous behaviors to form a new behavior. Disclosure of Invention In order to overcome the defects and shortcomings in the prior art, the invention provides the behavioral fingerprint digital characteristic extraction method, which can obtain better identity recognition effect and remarkably reduce the types of the extracted digital characteristics, thereby greatly reducing the time required for recognition and improving the accuracy rate in the process of identity recognition based on similarity distance calculation. A second object of the present invention is to provide a behavioral fingerprint digital feature extraction system; A third object of the present invention is to provide a computer-readable storage medium; It is a fourth object of the present invention to provide a computing device. In order to achieve the above purpose, the present invention adopts the following technical scheme: the invention provides a behavior fingerprint digital feature extraction method, which comprises the following steps: acquiring a behavior fingerprint sequence of a user, extracting digital characteristic information of the user, respectively constructing a digital characteristic vector matrix for a user set to be matched and a known user set, and extracting a characteristic vector set of the user for the acquired behavior fingerprint sequence of the user; Calculating the similarity distance between the known user digital feature vector and the user digital feature vector to be matched based on a similarity distance formula; And judging the identity of the user according to a preset matching strategy. As an optimal technical scheme, the behavior fingerprint sequence of the user is derived from log records of various systems, partial log records are extracted to form data, a time point is selected, and the data set is cut into a user set to be matched and a known user set according to the time point. As a preferred technical solution, the steps of obtaining the fingerprint sequence of the user and extracting the digital feature information of the user include: Extracting an n-order feature vector set of the user for the acquired behavior fingerprint sequence S u of the user u The n-order feature vectors of all users are collected and combined, and the frequency of each digital feature vector occurrence is counted and combined to generate a tupleEach element in the tuple is composed of a digital feature vector and its frequency of occurrence in the tuple; According to tuple x The frequency of each digital feature vector is ordered from big to small to form ordered tuple which is recorded as: from ordered tuples Extracting feature component tuples of set proportions And fusing the n-order feature vector sets of all the users to obtain a behavior frequency matrix comprising all the users and the digital feature vectors thereof. As a preferable technical scheme, extracting the n-order feature vector set of the user based on FaG methodThe method comprises the following specific