Database fuzzy testing method and system based on depth feedback and semantic preservation

Abstract

The invention relates to the technical field of computer software security, and provides a database fuzz testing method and system based on depth feedback and semantic preservation, and the method comprises the steps: collecting a standard query template, setting a variation rule base, and generating a variation strategy through a variation strategy generator; sending the variation strategy to a target database for fuzzy testing, and recording response information of the target database to the variation strategy; response information is obtained in real time through a depth feedback mechanism, and a variation strategy is dynamically optimized; generating a test case through a semantic preserving strategy; the priority of the test tasks is dynamically adjusted through an intelligent scheduling module, and resource allocation is carried out; and executing the test case to obtain a test result. According to the method, the effectiveness and the reliability of database fuzzy testing are remarkably improved, the capability of discovering deep security vulnerabilities is greatly enhanced, and powerful support is provided for protecting the database from being threatened by various unknown attack modes.

Description

Technical Field

[0001] The present invention relates to the technical field of computer software security, and in particular to a database fuzz testing method and system based on deep feedback and semantic preservation. Background Art

[0002] With the development of information technology, database systems are increasingly widely used. However, vulnerabilities in database systems may lead to serious data leakage or loss problems. Although traditional fuzz testing methods can detect some basic problems, they perform poorly when dealing with data with complex structures and often ignore semantic information, resulting in problems such as insufficient test coverage, high false positive rates, and low efficiency. Summary of the Invention

[0003] The present invention aims to solve at least one of the technical problems existing in the related art. For this purpose, the present invention provides a database fuzz testing method and system based on deep feedback and semantic preservation, which significantly improves the effectiveness and reliability of database fuzz testing, greatly enhances the ability to discover deep security vulnerabilities, and provides strong support for protecting databases from various unknown attack patterns.

[0004] The present invention provides a database fuzz testing method based on deep feedback and semantic preservation, including: S1: Collect standard query templates, screen and label the standard query templates, set basic mutation rules according to the standard query templates, and construct and regularly update a mutation rule library according to the basic mutation rules; S2: Through a mutation strategy generator, automatically generate a mutation strategy according to the mutation rule library for the query input; send the mutation strategy to the target database for fuzz testing, and record the response information of the target database to the mutation strategy; S3: Real-time obtain the response information of the target database to the mutation strategy through a deep feedback mechanism, and dynamically optimize the mutation strategy by using the response information; S4: Generate test cases with correct syntax and semantics that meet expectations according to the optimized mutation strategy through a semantic preservation strategy; S5: Dynamically adjust the priorities of test tasks and allocate resources through an intelligent scheduling module; execute the test cases to obtain test results.

[0005] Further, in step S1, Collect standard SQL query templates from public resources and internal accumulations, and the standard SQL query templates include CRUD operations; Screen and label the standard SQL query template, where the screening includes reviewing the quality and applicability of the standard SQL query template, and the labeling includes the scope of application, expected output, and precautions.

[0006] Further, the basic mutation rules include random character replacement, inserting or deleting keywords, and changing the numerical size.

[0007] Further, the in-depth feedback mechanism includes: Using a standard database connection interface to obtain the response information of the database to the mutation strategy in real time; Adopting a multi-level caching mechanism, using memory caching to quickly store the response information, persisting important data to disk or a distributed file system, and backing up regularly; Using machine learning algorithms to perform data analysis on the response information to identify normal behaviors, abnormal behaviors, and potential security threats; Optimizing the mutation strategy according to abnormal behaviors and potential security threats through a combination of reinforcement learning algorithms and genetic algorithms.

[0008] Further, the use of machine learning algorithms to perform data analysis on the response information to identify normal behaviors, abnormal behaviors, and potential security threats includes; Using unsupervised learning algorithms to identify normal behaviors; Using supervised learning algorithms to train a model to identify abnormal behaviors and potential security threats.

[0009] Further, the response information includes the returned data set, execution time, error code, and status information; The machine learning algorithms include any one of K-means clustering, ARIMA model, support vector machine, random forest, convolutional neural network, and long short-term memory network.

[0010] Further, the semantic preservation strategy includes: Designing an SQL language parser using the recursive descent parsing method, where the SQL language parser supports nested queries and conditional expressions; Parsing the SQL query and its mutated version through the SQL language parser; Taking the parsed SQL query and its mutated version as training data, and evaluating the impact of the mutated query on semantics by training a deep learning model; Performing semantic consistency checks after each mutated query, and verifying semantic consistency by comparing the query result sets or expected invariants before and after mutation.

[0011] Further, the deep learning model is a recurrent neural network or a long short-term memory network.

[0012] Further, the intelligent scheduling module includes: Dynamically adjust the priority of tasks according to the current test progress and resource usage Allocate tasks among multiple test nodes using a load balancing algorithm; Provide a graphical interface to support users to customize and configure the test scope and mutation strategy.

[0013] The present invention also provides a database fuzz testing system based on deep feedback and semantic preservation for executing any one of the above-mentioned database fuzz testing methods based on deep feedback and semantic preservation, including: A mutation rule library acquisition module, which collects standard query templates, screens and annotates the standard query templates, sets basic mutation rules according to the standard query templates, and constructs and regularly updates a mutation rule library according to the basic mutation rules; A fuzz testing module, which automatically generates a mutation strategy according to the mutation rule library through a mutation strategy generator; sends the mutation strategy to the target database for fuzz testing, and records the response information of the target database to the mutation strategy; A deep feedback module, which obtains the response information of the target database to the mutation strategy in real time through a deep feedback mechanism, and dynamically optimizes the mutation strategy by using the response information; A semantic preservation module, which generates test cases with correct syntax and expected semantics according to the optimized mutation strategy through a semantic preservation strategy; An intelligent scheduling module, which dynamically adjusts the priority of test tasks and performs resource allocation through the intelligent scheduling module; executes test cases to obtain test results.

[0014] One or more of the above technical solutions in the embodiments of the present invention have at least one of the following technical effects: Significantly improves the effectiveness and reliability of database fuzz testing, greatly enhances the ability to discover deep security vulnerabilities, and provides strong support for protecting databases from various unknown attack patterns.

[0015] The additional aspects and advantages of the present invention will be partially given in the following description, partially become obvious from the following description, or be understood through the practice of the present invention. Description of the Drawings

[0016] To more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0017] Figure 1 It is a schematic flowchart of a database fuzz testing method based on deep feedback and semantic preservation provided by the present invention.

[0018] Figure 2 It is a schematic structural diagram of a database fuzz testing system based on deep feedback and semantic preservation provided by the present invention.

[0019] Reference numerals: 101, mutation rule library acquisition module; 102, fuzz testing module; 103, deep feedback module; 104, semantic preservation module; 105, intelligent scheduling module. Detailed implementation manners

[0020] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention. The following embodiments are used to illustrate the present invention, but cannot be used to limit the scope of the present invention.

[0021] In the description of this specification, the descriptions referring to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the embodiments of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, without conflict, those skilled in the art can combine the different embodiments or examples described in this specification and the features of the different embodiments or examples.

[0022] The following will describe Figures 1 to 2 a database fuzz testing method and system based on deep feedback and semantic preservation of the present invention.

[0023] As Figure 1 shown, a database fuzz testing method based on deep feedback and semantic preservation includes: S1: Collect standard query templates, screen and annotate the standard query templates, set basic mutation rules according to the standard query templates, and construct and regularly update a mutation rule library according to the basic mutation rules; Collect standard SQL query templates from public resources and internal accumulations. The standard SQL query templates include CRUD operations and other complex query scenarios; CRUD operations include Create, Read, Update, and Delete.

[0024] Screen and annotate the standard SQL query templates. Screening includes reviewing the quality and applicability of the standard SQL query templates, and annotation includes scope of application, expected output, and precautions.

[0025] Each template needs to be strictly screened to ensure its representativeness. By setting up a special review committee, review the quality and applicability of the templates.

[0026] Make detailed annotations for each template to explain its scope of application, expected output, and any special precautions. Provide specific example inputs and outputs to help users understand the specific uses of the templates.

[0027] The basic mutation rules include random character replacement, inserting / deleting keywords, changing the value size, etc.

[0028] Regularly update the mutation rule library and add new mutation rules to cope with the ever-changing security threats.

[0029] S2: Through a mutation strategy generator, automatically generate a mutation strategy from the query input according to the mutation rule library; send the mutation strategy to the target database for fuzz testing, and record the response information of the target database to the mutation strategy; The mutation strategy generator can automatically generate a suitable mutation strategy according to the user's input.

[0030] S3: Real-time obtain the response information of the target database to the mutation strategy through a deep feedback mechanism, and dynamically optimize the mutation strategy by using the response information; The deep feedback mechanism includes: Use a standard database connection interface to real-time obtain the response information of the database to the mutation strategy; The standard database connection interface is JDBC or ODBC. The response information includes the returned data set, execution time, error code, and status information.

[0031] Adopt a multi-level caching mechanism, use in-memory caching to quickly store the response information, persist important data to disk or a distributed file system, and perform regular backups; To handle data streams in high-concurrency scenarios, a multi-level caching mechanism is adopted to store response information.

[0032] A machine learning algorithm is used to perform data analysis on the response information to identify normal behaviors, abnormal behaviors, and potential security threats, including: An unsupervised learning algorithm is used to identify normal behaviors; Clustering is performed by calculating the distance between samples (such as Euclidean distance), and the calculation expression of its objective function is: Where, is the objective function of clustering, is the number of clusters, is the sample set of the i-th cluster, is the center of the i-th cluster, is the sample point, ||·|| is the Euclidean distance, is the function to find the minimum value.

[0033] A supervised learning algorithm is used to train a model to identify abnormal behaviors and potential security threats; The autoregressive integrated moving average model (ARIMA) is used to predict time series to monitor long-term trends and periodic changes, helping to identify potential security threats. Given time series data , the calculation expression of the ARIMA model is: Where, is the backshift operator, , is the time series data at time is the number of backshift steps, is the polynomial of the autoregressive (AR) part, with order p, is the differencing operation used to make the time series stationary, is the differencing order, is the polynomial of the moving average (MA) part, with order , is the white noise sequence, following a normal distribution , is the variance.

[0034] The machine learning algorithm includes any one of K-means clustering, ARIMA model, support vector machine, random forest, convolutional neural network, and long short-term memory network.

[0035] Once an abnormal situation is detected, immediately trigger the feedback mechanism, adjust the subsequent test strategy, and judge whether the test result is abnormal through the preset success / failure criteria. If an abnormality is found, save the input as a valuable test case and adjust the subsequent mutation strategy accordingly.

[0036] Optimize the mutation strategy according to abnormal behaviors and potential security threats through the combination of reinforcement learning algorithm and genetic algorithm. The reinforcement learning algorithm (such as Q-learning) selects the best action according to the environmental state and combines with the genetic algorithm to simulate the natural selection process, gradually optimizing the mutation strategy, including: Initialize the expected return, and the calculation formula is: where S is the state space, A is the action space, is the expected return for taking action in state , is the assignment symbol; Select the action at time by combining the greedy strategy and random noise, and the calculation formula is: where is the state at time, is the expected return for taking action in state , is the exploration rate at time, and gradually decreases over time, is the random noise, and is used to simulate the exploration behavior, is the value of that makes the function reach the maximum value.

[0037] Update the expected return according to the Bellman equation, and the calculation formula is: where is the expected return for taking action in the updated state , is the expected return for taking action in state , is the learning rate, and controls the update step size. is the discount factor, , measuring the importance of future rewards, is the immediate reward at the current time step, is the state at time is the state at time is the action at time is the action that obtains the maximum expected return in state ; is the maximum expected return for taking action in state ; When the environment returns a termination signal, end the current episode and repeat the above process until all episodes are completed.

[0038] In the initial stage of testing, adopt a more conservative mutation strategy, such as simple operations like character replacement and word swapping, to ensure coverage of as many basic scenarios as possible. Build a library containing various mutation rules, and users can select appropriate rule combinations according to their needs.

[0039] Continuously collect and analyze the data generated during the testing process, and use incremental learning methods to continuously update the model parameters, enabling more intelligent selection of mutation paths.

[0040] S4: Generate syntactically correct and semantically expected test cases according to the optimized mutation strategy through a semantic preservation strategy; Design an SQL language parser using the recursive descent parsing method, and the SQL language parser supports nested queries and conditional expressions; The SQL language parser has a large number of built-in grammar rules, supports user-defined rules, supports multiple SQL languages, and has good scalability.

[0041] Parse the SQL query and its mutated versions through the SQL language parser; Use the parsed SQL query and its mutated versions as training data to evaluate the impact of the mutated query on semantics through training a deep learning model; the deep learning model is a recurrent neural network or a long short-term memory network.

[0042] Perform a semantic consistency check after each mutated query, and verify the semantic consistency by comparing the query result sets before and after the mutation or the expected invariants.

[0043] A semantic consistency check is performed after each mutation to ensure that the mutated query still conforms to the original intention. The consistency check is achieved by comparing the query result sets before and after the mutation or verifying certain expected invariants.

[0044] Through the semantic preservation strategy, when performing SQL query mutation, it is ensured that the mutated query still conforms to the original intention, avoiding false positives caused by mutation operations. This not only improves the accuracy of test results but also reduces the impact on normal business logic.

[0045] By using machine learning algorithms to analyze historical data, it is possible to predict which types of inputs may lead to abnormal behaviors or security vulnerabilities and take preventive measures in advance.

[0046] S5: Dynamically adjust the priorities of test tasks and allocate resources through the intelligent scheduling module; execute test cases to obtain test results; Dynamically adjust the priorities of tasks according to the current test progress and resource usage, and dynamically calculate task priorities based on factors such as the historical performance and potential value of tasks.

[0047] Continuously monitor multiple key metrics such as CPU utilization, memory occupancy, and disk I / O rate to ensure the smooth operation of the test process.

[0048] Provide an intuitive graphical interface that allows users to set areas of concern (such as specific types of SQL injection attacks) or exclude certain parts that are not of interest (such as problems specific to old version databases) according to their own needs. Configuration options: Provide rich configuration options covering multiple aspects such as test scope and mutation strategy.

[0049] Use a load balancing algorithm to distribute tasks among multiple test nodes; It can dynamically adjust the priorities of tasks according to the current test progress and resource usage, reasonably allocate computing resources, prevent resource waste or over-occupation, and improve overall efficiency.

[0050] Provide a graphical interface to support users in customizing the test scope and mutation strategy; Provide an intuitive graphical interface, simplify the operation process, and even non-technical personnel can easily use this tool to perform database security detection. Support internationalization requirements, the interface supports multi-language switching, and provides online document viewing and downloading functions to facilitate users to consult relevant materials and technical support services at any time.

[0051] Classify the test results into three categories: normal, warning, and abnormal, corresponding to no impact, minor impact, and serious problems respectively. Users can view the test progress and results in real time, and intuitively display the test results in the form of charts to help users quickly understand the test overview.

[0052] Such as Figure 2As shown in the figure, a database fuzz testing system based on deep feedback and semantic preservation is used to execute the above-mentioned database fuzz testing method based on deep feedback and semantic preservation, including: The mutation rule library acquisition module 101 collects standard query templates, screens and annotates the standard query templates, sets basic mutation rules according to the standard query templates, and updates the mutation rule library regularly; The fuzz testing module 102 generates mutation strategies automatically according to the mutation rule library through the mutation strategy generator; sends the mutation strategies to the target database for fuzz testing, and records the response information of the target database to the mutation strategies; The deep feedback module 103 obtains the response information of the target database to the mutation strategies in real time through the deep feedback mechanism, and dynamically optimizes the mutation strategies by using the response information; The semantic preservation module 104 generates test cases with correct syntax and expected semantics according to the optimized mutation strategies through the semantic preservation strategy; The intelligent scheduling module 105 dynamically adjusts the priorities of test tasks and allocates resources through the intelligent scheduling module; executes the test cases and obtains the test results.

[0053] Through the collaborative work of the above-mentioned modules, real-time monitoring and analysis of database responses are carried out through a deep feedback mechanism, which can instantly identify potential security threats or performance bottlenecks, and dynamically adjust the test strategy accordingly. This adaptive mutation strategy adjustment mechanism makes the test process more intelligent and effectively improves the probability of discovering deep security vulnerabilities. It realizes the batch sending of mutated queries to the target database by automated scripts and records all response details, greatly improving the test efficiency; through the semantic preservation strategy, it ensures that the mutated query still conforms to the original intention when performing SQL query mutation, avoiding false alarm problems caused by mutation operations. It not only improves the accuracy of test results but also reduces the impact on normal business logic. By using machine learning algorithms to analyze historical data, it can predict which types of inputs may lead to abnormal behaviors or security vulnerabilities and take preventive measures in advance. It has a built-in intelligent scheduling module that can dynamically adjust task priorities according to the current test progress and resource usage, reasonably allocate computing resources, prevent resource waste or over-occupation, and uses a load balancing algorithm to reasonably allocate tasks among multiple test nodes, improving the overall efficiency. It introduces a multi-level caching mechanism to effectively handle the large amount of data streams generated under high concurrency, ensuring the stability and response speed of the system. It provides an intuitive graphical interface to simplify the operation process, enabling even non-technical personnel to easily use this tool for database security detection. It supports internationalization requirements, the interface supports multi-language switching, and provides online document viewing and downloading functions, facilitating users to consult relevant materials and technical support services at any time. Through a continuous learning mechanism, the model parameters are continuously updated, enabling the system to more intelligently select mutation paths and adapt to new threat patterns. By using an incremental learning method to gradually optimize the mutation strategy, it ensures that the system is always in the best state. The model is retrained regularly to adapt to new threat patterns, and combined with transfer learning methods, the generalization ability and accuracy of the model are continuously improved. The present invention not only significantly improves the effectiveness and reliability of database fuzz testing but also greatly enhances the ability to discover deep security vulnerabilities, providing strong support for protecting the database from various unknown attack patterns.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A database fuzz testing method based on deep feedback and semantic preservation, characterized in that Including: S1: Collect standard query templates, screen and label the standard query templates, set basic mutation rules according to the standard query templates, and construct and regularly update a mutation rule library according to the basic mutation rules; S2: Through a mutation strategy generator, automatically generate a mutation strategy from the query input according to the mutation rule library; send the mutation strategy to the target database for fuzz testing, and record the response information of the target database to the mutation strategy; S3: Real-time obtain the response information of the target database to the mutation strategy through a deep feedback mechanism, and dynamically optimize the mutation strategy by using the response information; S4: Generate test cases with correct syntax and expected semantics according to the optimized mutation strategy through a semantic preservation strategy; S5: Dynamically adjust the priority of test tasks and allocate resources through an intelligent scheduling module; execute the test cases to obtain test results.

2. The database fuzz testing method based on depth feedback and semantic preservation according to claim 1, wherein In step S1, Collect standard SQL query templates from public resources and internal accumulations, and the standard SQL query templates include CRUD operations; Screen and label the standard SQL query templates, where the screening includes reviewing the quality and applicability of the standard SQL query templates, and the labeling includes the scope of application, expected output, and precautions.

3. A database fuzz testing method based on depth feedback and semantic preservation according to claim 1, characterized in that The basic mutation rules include random character replacement, inserting or deleting keywords, and changing the value size.

4. A database fuzz testing method based on depth feedback and semantic preservation according to claim 1, characterized in that The deep feedback mechanism includes: Use a standard database connection interface to real-time obtain the response information of the database to the mutation strategy; Adopt a multi-level caching mechanism, use in-memory caching to quickly store the response information, persist important data to disk or a distributed file system, and perform regular backups; Use machine learning algorithms to perform data analysis on the response information to identify normal behaviors, abnormal behaviors, and potential security threats; Optimize the mutation strategy according to abnormal behaviors and potential security threats through a reinforcement learning algorithm combined with a genetic algorithm.

5. A database fuzz testing method based on depth feedback and semantic preservation according to claim 4, characterized in that, The use of machine learning algorithms to perform data analysis on the response information to identify normal behaviors, abnormal behaviors, and potential security threats includes; Use unsupervised learning algorithms to identify normal behaviors; Use supervised learning algorithms to train models to identify abnormal behaviors and potential security threats.

6. According to the method for database fuzz testing based on deep feedback and semantic preservation as claimed in claim 4, characterized in that The response information includes the returned data set, execution time, error code, and status information; The machine learning algorithms include any one of K-means clustering, ARIMA model, support vector machine, random forest, convolutional neural network, and long short-term memory network.

7. A database fuzz testing method based on deep feedback and semantic preservation according to claim 1, characterized in that The semantic preservation strategy includes: Design an SQL language parser using the recursive descent parsing method, and the SQL language parser supports nested queries and conditional expressions; Parse the SQL query and its mutated version through the SQL language parser; Use the parsed SQL query and its mutated version as training data, and evaluate the impact of the mutated query on semantics by training a deep learning model; Perform semantic consistency checks after each mutation query, and verify semantic consistency by comparing the query result sets or expected invariants before and after mutation.

8. A database fuzz testing method based on deep feedback and semantic preservation according to claim 7, characterized in that, The deep learning model is a recurrent neural network or a long short-term memory network.

9. A database fuzz testing method based on depth feedback and semantic preservation according to claim 1, characterized in that The intelligent scheduling module includes: Dynamically adjust the priorities of tasks according to the current test progress and resource usage, Use a load balancing algorithm to distribute tasks among multiple test nodes; Provide a graphical interface to support users to customize and configure the test scope and mutation strategy.

10. A database fuzz testing system based on deep feedback and semantic preservation, characterized in that, To execute a database fuzz testing method based on deep feedback and semantic preservation as described in any one of claims 1 to 9, including: A mutation rule library acquisition module, which collects standard query templates, screens and annotates the standard query templates, sets basic mutation rules according to the standard query templates, and constructs and periodically updates a mutation rule library according to the basic mutation rules; A fuzz testing module, which automatically generates a mutation strategy from the query input according to the mutation rule library through a mutation strategy generator; sends the mutation strategy to the target database for fuzz testing, and records the response information of the target database to the mutation strategy; A deep feedback module, which obtains the response information of the target database to the mutation strategy in real time through a deep feedback mechanism, and dynamically optimizes the mutation strategy by using the response information; A semantic preservation module, which generates test cases with correct syntax and semantics that meet expectations according to the optimized mutation strategy through a semantic preservation strategy; An intelligent scheduling module, which dynamically adjusts the priorities of test tasks and allocates resources through the intelligent scheduling module; executes the test cases to obtain test results.

Images