203 results about "Injection attacks" patented technology

Systems and methods for detecting and mitigating prompt injection attacks on a generative LLM are disclosed. A deployment scenario is considered, in which the generative LLM supports a task automation function. Prompts are received and interpreted by the generative LLM, and outputs from the generative LLM are used to trigger automation actions. The prompts are constructed based on a combination of user input and external data and are, therefore, vulnerable to prompt injection attacks though manipulation of the external data. To mitigate this risk, a separate discriminative classification, decoupled from the generative LLM, engine is configured to identify malicious prompts, and filter out any malicious prompts before they reach the generative LLM.

The invention relates to a method for obtaining a hint injection attack test case for a large language model, which combines a conditional variation auto-encoder cVAE and a Markov chain, and obtains a large language model test case through data generation and context modeling, gradual exposure of malicious instructions and simulation of multiple rounds of dialogue attacks in reality. Realizing multiple rounds of dialogue attacks on the large language model, and challenging the defense capability of the large language model; the design scheme introduces a concealment technology, role play attack, state transition and other technologies, improves the complexity and concealment of attacks, has the core advantages of automation, higher concealment, wide coverage, batch testing and the like, can evaluate the security defense capability of the large language model more truly and comprehensively, and improves the security defense capability of the large language model. The defects of an existing defense mechanism are found, and research on multi-round prompt injection attacks and improvement of a security defense mechanism are promoted.

The invention relates to the field of artificial intelligence and natural language processing, in particular to a method, a model and equipment for identifying a prompt injection attack for a large language model. The method comprises the following steps: acquiring a cue word to be input into an LLM; performing semantic analysis on the cue word to obtain a semantic analysis result; performing prompt injection word detection on the prompt word to obtain a prompt injection word detection result; and according to a semantic analysis result and a prompt injection word detection result, analyzing whether the prompt word relates to a prompt injection attack. The invention provides a fusion recognition mechanism combining semantic analysis and prompt injection word detection, and the fusion recognition mechanism is specially used for recognizing the prompt injection attack aiming at the LLM, so that the defense capability for the LLM prompt injection attack is improved.

The embodiment of the invention relates to the field of artificial intelligence, and provides a model context protocol injection attack protection method and device based on dynamic semantic analysis, computer equipment and a storage medium, and the method comprises the steps: obtaining request data corresponding to a call request initiated by a user through a model context protocol, carrying out the preprocessing of the request data, obtaining the preprocessed request data; performing semantic vectorization on request text and context historical information in the request data through a lightweight bidirectional encoder representation model to output an initial risk score; correcting the initial risk score according to context historical information carried in a model context protocol to obtain a corrected final risk score; and performing hierarchical defense decision according to the final risk score, and determining risk grading information corresponding to each piece of request data so as to execute a protection action corresponding to each piece of risk grading information. By adopting the method, the accuracy of identifying the protocol injection attack can be improved.

A system and method for multiple fraud type detection includes anti-injection attack system that has a layered architectural approach that uses a includes combination of different specific models to detect the attacks in combination with image processing techniques, device signals and liveness checks to detect the variety of different types of fraud attacks or repeat fraud attacks. The anti-injection attack system applies the analysis of the tools used to create deepfake, face morph and face swap attacks to define the elements of its layered architecture that can these various types of attacks.

The invention discloses a multi-modal attack identification method fusing BMama and difference to guide trans-attention, which comprises the following steps: simulating a false data injection attack, a denial of service attack, an address resolution protocol spoofing attack and a domain name system spoofing attack, collecting physical layer sensor data and network layer flow data, and preprocessing multi-modal data; bMama is constructed to perform dynamic time modeling on multi-modal data, a graph neural network is combined to adversariate a variational auto-encoder, features of a power grid system topology and a communication topology structure are fused, and robustness of potential representation is enhanced through adversarial training; the method comprises the following steps of: guiding feature complementary fusion by using modal difference through a difference guide iteration cross-attention fusion mechanism, improving the capability of distinguishing complex attacks, finally carrying out attack detection and classification on fused modals, and executing end-to-end optimization according to a weighted combination of loss of each part. The method can effectively detect and classify the multi-modal attack in the smart power grid, and enhances the safety and reliability of a complex system.

The invention discloses a power network attack chain dynamic deduction and intelligent response process method, system and device based on deep reinforcement learning and a medium, and belongs to the technical field of network security and power system protection. Multi-modal data is aligned and normalized, an event view cache is constructed, and the generalization detection capability on process camouflage and memory injection attacks is improved through a federated learning collaborative detection mechanism; based on the event view cache and historical threat intelligence, generating a dynamic attack knowledge graph, constructing a deep reinforcement learning model taking the attack knowledge graph as an environment, calculating an attack influence index by using a Bayesian network, and generating a differentiated security response instruction; and realizing attack path backtracking and attack source positioning based on the attack knowledge graph. According to the method, multi-modal data fusion analysis and strategy adaptive updating are realized, and the attack chain identification accuracy and evidence chain construction integrity are remarkably improved.

The invention provides a direct current micro-grid false data injection attack detection method based on deep learning, and aims to improve the safety and stability of a direct current micro-grid when the direct current micro-grid encounters hostile attacks. According to the method, a parallel double-branch model in which Transform and GRU are combined is adopted, the advantages of the Transform and the GRU are brought into full play, and efficient identification of the false data injection attack (FDIA) is achieved. Through multi-level feature fusion and a decision-making mechanism, the model can accurately identify various false data injection attacks in a complex attack scene, and relatively high detection accuracy is kept. Experimental results show that the method shows excellent performance in a direct current micro-grid environment, has a wide application prospect, and particularly has a remarkable application value in engineering practice.

The invention discloses a false data injection attack detection method and system based on elastic energy prediction. The method comprises the steps that a distributed node network is constructed, nodes are deployed to collect real-time energy output data, each node generates a state estimation value according to a Kalman consensus filtering model, an innovation vector is calculated, and a monitoring threshold is adjusted by combining an association weight of a network topology; inputting a state estimation value, an innovation vector and real-time energy output data, calculating an accumulated deviation trend by using a time sequence analysis method, identifying a slow penetration attack in combination with a sliding window, and judging whether a false data injection attack exists or not; when it is judged that the false data injection attack exists, operation parameters of the energy equipment are adjusted, attacked nodes are isolated, non-attacked nodes are redistributed, and state change information is recorded; and according to the attack frequency, the node association weight and the operation cost, the detection parameters are adjusted in the dual-time scale, the resource allocation weight is optimized based on the feedback data, and the detection probability, the response time and the resource allocation strategy are dynamically updated.

In an embodiment, a method includes receiving a prompt provided to a large language model (LLM), generating a semantic fingerprint for the prompt based on semantic and syntactic features associated with the prompt, generating a vector representation incorporating the semantic fingerprint for the prompt, calculating semantic distances between the vector representation and multiple vector representations associated with multiple malicious intents, determining an intent associated with the prompt based on the semantic distances, and determining a defense action on the prompt based on the intent and multiple policies to mitigate a security risk associated with the prompt.

The embodiment of the invention provides a coupling network asynchronous dynamic intermittent safety control method and system, and the method comprises the steps: obtaining multiple attack signals comprising an injection attack signal and a replacement attack signal after building a dynamic model of a following network, a target network and a state error variable; and setting an intermittent controller according to the dynamic network model and the multiple attack signals, and setting a dynamic mechanism based on an exponential function to convert a state error variable into an interval control function. Wherein the interval control function is used for realizing security mean square synchronization control between the following network model and the target network model. According to the method, the asynchronous anti-attack intermittent dynamic cooperative control strategy oriented to the complex coupled network can be constructed based on an average intermittent control rate and dynamic event trigger threshold coupling method by utilizing a node self-adaptive dual spoofing attack probability model, and the problems of difficulty in multi-attack modeling and insufficient control strategy collaboration can be relieved.

A malicious instructions detection model (“detector”) intercepts augmented prompts destined for a large language model (“LLM”). Each augmented prompt was augmented with data from potentially compromised data sources susceptible to indirect prompt injection attacks. The detector tokenizes / preprocesses sentences in the augmented prompts and is invoked on the tokenized / preprocessed sentences to obtain confidence scores that each sentence comprises malicious instructions. If one or more of the confidence scores is above a threshold, the detector blocks the augmented prompt and generates an alert indicating the blocking and the malicious instructions. Otherwise, the detector communicates the augmented prompt to its intended LLM.

Prompt injection attack detection in responses from large language models includes receiving, at a server from a user device, a user prompt segment to an LLM, generating a LLM prompt from the user prompt segment, sending the LLM prompt to the LLM, and receiving a response from the LLM. Prompt injection attack detection further includes comparing the response to a structured data schema for the response to validate the response, and sending, responsive to validating the response, the response to the user device.

Systems and methods for ransomware protection in advanced injection-based attacks. Events from a driver are analyzed to identify a potentially malicious actor. A confidence level is calculated for the potentially malicious actor identification to weigh the probability that the actor is malware or a source of malware. Behavior associated with the potentially malicious actor can be collected, if it is likely that the potentially malicious actor is associated with malware, such as if the confidence level is over a predetermined threshold. Subsequently, one or more virtual honeypots are generated for the suspicious processes and the response to the virtual honeypots is included in anti-ransomware heuristic analysis.

The invention discloses a false data injection attack detection method based on federal learning and comparative learning, and relates to the technical field of power system security, and the method comprises the steps: collecting voltage, phase, frequency and equipment state power data from a power station, carrying out the denoising through employing a wavelet transformation method, filling a missing value through employing KNN interpolation, converting the data into a JSON format, and carrying out the detection of a false data injection attack. Generating a hash value by using SHA-256, and uploading the data in the JSON format and the hash value to a block chain; according to the method, a false data injection attack detection method is combined with federal learning, each power station can locally train model parameters, so that uploading of original data is avoided, the risk of data leakage is reduced, the model parameters and power data are managed through the block chain technology, the integrity and security of the data are ensured, and the method is suitable for popularization and application. Therefore, the data reliability and the safety protection capability of the power system are improved.

The invention discloses a false data injection attack detection method for an electric power information physical system, relates to the technical field of attack detection, and solves the problem that in the prior art, only the damage of FDIA to a measurement data time mode is considered, and the spatial correlation between measurement data is affected even if the FDIA is not considered. According to the method, an FDIA detection model structure based on correlation difference is designed, and the attention action range is constrained by embedding position information and introducing a correction factor. Besides, in combination with fine granularity and multi-scale characteristics of a measurement data sequence, topological correlation is introduced, correlation differences are defined, a correlation difference-based antagonism discriminant criterion reasoning method is designed, and the distinguishability of normal and attack measurement data is amplified through adversarial training, so that a discriminant criterion is obtained. Complete embodiments show that compared with an existing detection model, the provided model has higher detection accuracy and robustness, and is excellent in performance when facing injection attacks with different amplitudes.

The invention discloses an SQL injection detection method and device, equipment and a storage medium, and relates to the technical field of safety protection, and the method comprises the steps: obtaining an initial SQL statement input by a target user through a target Web application, and carrying out the preprocessing of the initial SQL statement, so as to obtain a corresponding target SQL statement; generating a corresponding token sequence based on the target SQL statement, and constructing a target abstract syntax tree corresponding to the target SQL statement; wherein the token sequence is a sequence generated based on each SQL element in the target SQL statement; and extracting a target feature corresponding to the target SQL statement based on the token sequence and the target abstract syntax tree, and detecting whether the target SQL statement has an injection attack behavior based on the target feature by using a pre-trained target graph neural network model. According to the method, high-accuracy, high-adaptability and low-delay SQL injection real-time detection can be realized.

The invention discloses a three-layer cooperative attack detection and defense method and device oriented to a large language model context protocol. Aiming at the technical bottleneck problems of low detection accuracy and poor real-time performance caused by the fact that an existing large language model is easily subjected to semantic attack when calling an external tool, a three-level protection assembly line is constructed: firstly, an explicit threat is intercepted at a millisecond-level speed by adopting a regular static scanner capable of being subjected to hot updating; secondly, accurately identifying hidden confrontation prompts and complex semantic attacks by utilizing a multi-language text embedding model which is subjected to full-parameter fine adjustment on MCP injection attack statements; and finally, introducing an independent and lightweight LLM to carry out confidence fusion rechecking on a detection result. Experimental results show that the detection accuracy, the recall rate, the false alarm suppression, the response real-time performance, the cross-domain generalization ability to non-attack styles and the like are all remarkably improved, and continuous and reliable safety protection can be provided for a large language model tool chain.

The invention relates to the technical field of network security, in particular to an injection attack detection method based on multi-dimensional feature fusion and time optimization. The detection method comprises the following steps: acquiring character-level features and sub-word-level features of an HTTP request in a two-channel input mode; extracting multi-layer semantic information of the features by using a pre-training model; carrying out fusion processing on the features of different levels through a multi-dimensional feature fusion module; performing weight optimization on the fused features by using an adaptive feature attention module; applying a multi-scale cavity pooling module to extract multi-scale features; and performing time optimization on the model structure based on neural architecture search. According to the method, the detection accuracy of the Web application firewall on attacks such as SQL injection, XSS injection and RCE injection can be effectively improved, meanwhile, the real-time performance of the model is guaranteed, and the method is suitable for security protection in a high-concurrency Web application environment.

A tainting engine can work in conjunction with a syntax attack detection template to identify when a threat actor attempts a malicious attack in a cloud application scenario. Non-intrusive instrumentation can be used to provide detection of an attempted attack regardless of whether the cloud application is vulnerable to such attacks. Detection of attempted attacks can be an important part of maintaining network security, even in cases where an application itself is not vulnerable to such attacks. Further details about the attempted attack can be assembled, and a variety of actions can be taken in response to detection.

The invention relates to the technical field of artificial intelligence and information security, provides a cue word injection attack detection method and device of a large language model, equipment, a storage medium and a program product, and has a better cue word injection attack detection effect. The method comprises the following steps: generating cue words from preset test corpora, and inputting the cue words into a first large language model; the test corpus generation cue word is used for indicating the first large language model to generate a cue word injection test corpus in combination with multiple cue word injection modes for each cue word injection attack type; the cue word injection test corpus is input into the to-be-tested large language model, and an input and output data pair is obtained according to response content output by the to-be-tested large language model; and calling the second large language model to detect whether the to-be-detected large language model is successfully attacked according to the input and output data to obtain a target detection result aiming at the cue word injection attack.

The invention relates to a scene self-adaptive white-box prompt injection attack system. The system comprises a self-adaptive context prompt generation module, an enhanced GCG algorithm-based confrontation suffix generation module and a simulator-based intelligent agent test module with a body. The adaptive context prompt generation module constructs specific prompt templates for different scenes, and automatically identifies and encodes unique semantic features of different application scenes. And the adversarial suffix generation module based on the enhanced GCG algorithm calculates the coordinate gradient of each token position, provides direction guidance for subsequent token replacement based on a double-component loss function, and adopts a weighted combination loss function to realize system optimization. The simulator-based intelligent agent test module carries out semantic understanding and intention recognition through a target large language model, converts a natural language into a control instruction, recognizes different types of commands, and maps the analyzed commands to a specific API for calling.

The embodiment of the present application provides a method and system for asynchronous dynamic intermittent safety control of a coupled network. The method can obtain multiple attack signals including injection attack signals and replacement attack signals after establishing a dynamic model of a follower network, a target network, and a state error variable, and set an intermittent controller according to the dynamic network model and the multiple attack signals, and then set a dynamic mechanism based on an exponential function to convert the state error variable into an interval control function. The interval control function is used to achieve safe mean square synchronization control between the follower network model and the target network model. The method can utilize a node-adaptive double deception attack probability model, based on the average intermittent control rate and the dynamic event trigger threshold coupling method, to construct an asynchronous anti-attack intermittent dynamic collaborative control strategy for complex coupled networks, which can alleviate the difficulties in modeling multiple attacks and the lack of coordination of control strategies.

The invention relates to an elastic control method of a high-order nonlinear system under pulse false information injection attack. The method comprises the following steps: acquiring a system output signal and a control input signal of a current control period, and a state estimation value vector and a fuzzy weight vector estimation value of a previous control period; inputting a system output signal and a control input signal of a current control period, and a state estimation value vector and a fuzzy weight vector estimation value of a previous control period into a nonlinear state observer based on a fuzzy logic system to obtain a state estimation value vector of the current control period; the controller based on the fuzzy logic system carries out backstepping recursion through the state estimation value vector of the current control period and the fuzzy weight vector estimation value of the previous control period to obtain an elastic control input signal; and obtaining a fuzzy weight vector estimator of the current control period through an adaptive law based on the state estimation value vector of the current control period. By adopting the method, state jump and performance damage of attacks can be resisted.

The invention discloses an active defense security control method for a hidden false data injection attack, and relates to the field of information physical system security control, in particular to an active defense security control method. The invention aims to solve the problem that the existing detection method can encrypt a transmitted signal, but is limited to detection of a specific type of attack, or dynamic coupling changing the configuration of a control system can increase the burden of a state estimator, and even causes the performance reduction of the system. The method comprises the following steps of: 1, acquiring output feedback gains of the information physical system under different switching signals; and 2, based on the output feedback control gain obtained in the step 1, performing active attack detection and positioning on the FDI attack, and when the attack is detected and positioned, performing active switching by a controller of the information physical system so as to construct an active security controller aiming at the FDI attack.

The invention provides an attack and defense integrated method for large model prompt injection attacks, which comprises the following steps of: performing large model prompt injection attacks on user instructions by using an attack strategy based on emotion enhancement and control output to form malicious contents; the method comprises the following steps of: firstly, carrying out defense on malicious content, then fusing prevention strategies Sandwich and Instruction with a detection method Knowanswing, constructing a multi-level defense strategy UnD for cooperative work to carry out prompt injection defense on the malicious content to obtain an output result, and finally, carrying out effect evaluation by using an attack interference rate, an attack misleading rate, a METEOR score, a defense success rate and a known answer matching rate. According to the method, the prompt injection attack with higher attack efficiency is realized in the aspect of attack, the cooperation of detection and prevention is realized in the aspect of defense, the security of the large model facing the prompt injection attack is improved, and the large language model is ensured to safely and reliably run in various application scenes.

The invention belongs to the technical field of computer information security, particularly discloses a content security protection method and device based on semantic consistency, and aims to solve the problem that high-level cue word injection attacks are difficult to effectively recognize in the prior art. Carrying out real-time interception on dominant illegal texts through a content filtering module; quantizing generation rationality differences of prompt words between attack and normal language models by using a word vector confusion degree calculation module; evaluating the logic coherence of the sentence structure of the prompt word through a statement semantic consistency judgment module; integrating the multi-dimensional feature data and carrying out risk classification by adopting a machine learning algorithm; and routing the suspected attack request to a value fine tuning model for processing according to a judgment result. According to the technical scheme, high-precision recognition and response to complex cue word injection attacks are achieved, and the content safety protection capacity and compliance guarantee level of a large model in an interaction scene are remarkably improved.

The invention relates to a node injection attack method based on adaptive target selection, and the method comprises the following steps: S1, target node selection: calculating a comprehensive score of a node based on uncertainty and topology centrality, and dynamically selecting a target node set of a current attack round; s2, feature generation: using an adaptive feature generator to generate node features which are similar to target node distribution and have strong aggressiveness; and S3, disturbance edge construction: selecting an optimal disturbance edge connection mode for the injection node according to strategy network output in reinforcement learning. According to the method, the attack flexibility can be improved through dynamic target selection, the attack performance can be remarkably enhanced through combination of disturbance characteristics and structures, and the method has good concealment, expandability and generalization ability. The method is widely applied to security evaluation and defense research fields related to graph neural networks, such as social network analysis, recommendation systems, knowledge graphs and the like.

The application discloses a large language model prompt word injection attack detection method and device based on context learning, and belongs to the technical field of artificial intelligence, and comprises the following steps: based on a Bert pre-training model, inherent features and dependency relationships between different levels and different labels are learned by introducing a label-based attention module, a multi-level, multi-label and fine-grained classification model for prompt word injection attack is designed, and accurate identification of the prompt word injection attack is realized.Meanwhile, according to the context learning, the prediction ability of the classification model is combined with the ability of the large language model, and the defense ability of the large language model to the prompt word injection attack is improved.The application can automatically detect the prompt word injection attack, improve the effectiveness and comprehensiveness of detection, and can be effectively applied to the field of large model security detection.

The invention discloses a multi-agent control method for resisting data injection attack communication range constraints, and relates to the technical field of control. The method comprises the steps that S1, a security control strategy based on historical data is obtained, the security control strategy comprises the steps that a sender agent correspondingly sends # imgabs3 #, # imgabs4 # and # imgabs5 # data at three continuous moments # imgabs0 #, # imgabs1 # and # imgabs2 #, at the moment, corresponding information capable of being received by a receiver agent is # imgabs6 #, # imgabs7 # and # imgabs8 #, the receiver agent calculates the amount of approximate external attacker injection information based on the received data, and the amount of the external attacker injection information is calculated according to the amount of the external attacker injection information; obtaining data in an approximate neighbor state, wherein the data is discrete data; s2, carrying out serialization on the discrete data obtained in the step S1 to obtain smoothed data; s3, obtaining a distance-based communication weight strategy with a penalty factor to avoid an outlier phenomenon; s4, in combination with the communication weight strategy obtained in the step S3, a predetermined time control scheme is constructed and obtained based on backstepping control; under the attack of data injection, the controller of the intelligent agent can work normally.