39results about How to "Improve defense" patented technology

This invention relates to the fields of machine learning and edge computing technology, specifically to a blockchain-based trusted distributed hybrid expert model edge computing system. The system includes a task publisher for publishing learning tasks, which may be training or inference tasks; a blockchain network for receiving learning tasks and evaluating expert weights to obtain sparse activation expert index results; a blockchain network for receiving learning task processing results from the edge layer and achieving consensus; a gating network deployed within the blockchain network; an edge layer for downloading the corresponding expert model from the storage layer based on the sparse activation expert index results to process the learning tasks and uploading the processing results to the blockchain network; and a storage layer for storing all expert networks, each with a CID. This invention significantly improves the defense capability of the distributed MoE network against malicious edge node data tampering attacks, thereby ensuring the learning performance of the MoE system.

The invention discloses a secure transmission method and device, equipment, a medium and a product, and the method comprises the steps: obtaining an access request of a client and a current randomized tag pre-distributed to a web application in a server, and triggering the updating of the current randomized tag through a network attack association feature; scheduling a pre-trained target detection model based on the access request, and determining a security detection result of the access request; if the security detection result is secure, sending the access request to a server, and receiving a response result fed back by the server; and performing behavior analysis on the response result according to the current randomized tag, determining security feedback content and feeding back the security feedback content to the client. By dynamically updating the current randomized tag, the randomness of the defense form is enhanced, the method does not depend on the attack mode of an attacker, novel attacks or attack variants can be dealt with in real time, double attack detection of security detection and behavior analysis is carried out on the access request, and the detection capability and defense capability of the attacks are improved.

The utility model discloses a kind of fireproof light power cable of rat and termite prevention, including conductor, the outside of the conductor is equipped with shielding layer, the outside of the shielding layer is equipped with insulating layer, the outside of the insulating layer is equipped with inner protective layer, the outside of the inner protective layer is equipped with physical protection layer, the outside of the physical protection layer is equipped with protective net layer, in the utility model, the power cable, by inner protective layer inner layer elastic memory resistance layer hardens and releases chlorpyrifos when receiving ≥5N bite pressure, outer layer honeycomb metal-ceramic composite layer improves bite resistance strength;Physical protection layer forms barrier with material hardness;Protective net layer can release medicament, and the defense capability of rat and termite bite is improved compared with traditional cable by multilayer protection;And insulating layer uses high-temperature resistant material to guarantee fireproof basis;Inner protective layer outer layer honeycomb metal-ceramic composite layer can form fireproof barrier when meeting fire.These structures cooperate and improve the fire resistance in fire situation compared with traditional cable.

The invention relates to an acne-removing composition containing natural plant components and a preparation method of the acne-removing composition, and belongs to the technical field of cosmetics. The invention provides an acne-removing composition containing natural plant components. The acne-removing composition is prepared from rhodomyrtus tomentosa fruit extract, hydrolyzed red algae extract, hydrangea crassifolia leaf extract, Lagopsis supina extract and ceramide. Wherein the antibacterial and anti-inflammatory effects of the rhodomyrtus tomentosa fruit extract and the anti-inflammatory and soothing effects of the Lagopsis supina extract directly resist the core etiology and symptoms of acne; the hyperactivity of sebum is reduced by the hydrangea crassifolia leaf extract; the hydrolyzed red algae extract can buffer potential stimulation and maintain skin hydration; ceramide can strengthen the self-defense capability of the skin and prevent relapse during treatment; the five components supplement each other and have a good acne removing effect.

The invention discloses a time synchronization information security identification method and system based on a self-attention mechanism, and relates to a time synchronization and high-precision clock technology, and the system comprises a data collection and processing module which is used for receiving, caching and processing original time information data, converting the original time information data into variable input of a self-attention matrix calculation module, and outputting the variable input of the self-attention matrix calculation module; the variable input is a preprocessed clock error sequence; the self-attention matrix calculation module is used for receiving the preprocessed clock error sequence, calculating context correlation of clock error data and outputting a probability matrix; the security identification decision module is used for averaging all time points represented by the sequence so as to aggregate and obtain global features; global features are mapped into abnormal probabilities through weights and offsets, and sequence judgment results are divided into normal results and abnormal results through binary decision. According to the invention, illegal information embedded in a complex modulation mode can be effectively identified, and the perception and defense capability of the system to hidden attacks is significantly improved.

The present application relates to the technical field of feed premix, especially to a baiting and immunity double-effect premix containing yeast extract and a preparation method thereof, which comprises the following raw materials: yeast extract, compound plant extract, garlic extract, compound vitamin and compound trace element; the yeast extract is prepared by secondary enzymolysis, concentration and spray drying; the compound plant extract is prepared by reflux extraction and spray drying with camellia sinensis shell, Chinese chervil, kaffir lily and dark plum pulp as raw materials; the yeast extract is compounded with the camellia sinensis shell, Chinese chervil and kaffir lily in the compound plant extract, so that the prepared premix has excellent immunity effect; in addition, the yeast extract, dark plum pulp and garlic extract are used to synergistically significantly improve the baiting effect of the feed on aquatic animals and accelerate the feeding speed and feeding amount of the aquatic animals.

This invention relates to a method and system for accessing the trusted space of inland waterway shipping data under zero-trust conditions. It includes defining the shipping trust status and transformation rules of participating entities, constructing a risk scoring model to calculate the risk score of shipping behavior, introducing federated learning to improve the accuracy of the shipping behavior risk score, constructing a trust calculation function to calculate dynamic trust values ​​in real time, designing hierarchical access decision rules for hierarchical authorization to achieve access control, and submitting the hash value of logs to the blockchain to drive the update of shipping trust status, forming a closed loop of credit update and access control. This invention provides a systematic solution for the Jianghuai inland waterway shipping industry that balances strict security, business efficiency, and compliance requirements, achieving the coordinated evolution and proactive improvement of the security defense capabilities of the entire Jianghuai inland waterway shipping ecosystem.

The application discloses a method and device for verifying images, and relates to the technical field of computers. The method comprises the following steps: acquiring a to-be-verified image, and identifying first local features of a plurality of local regions of the to-be-verified image by using a plurality of target local feature recognition models; acquiring a reference image, and identifying second local features of a plurality of local regions of the reference image by using the plurality of target local feature recognition models; for each of the plurality of target local feature recognition models, acquiring a feature similarity between the first local features identified by using the target local feature recognition model and the second local features identified by using the target local feature recognition model; and determining whether the to-be-verified image passes the verification according to the plurality of acquired feature similarities. The method can improve the defense performance of the system against adversarial samples and improve the accuracy of the verified images.

The invention relates to a pedestrian detection method and device with confrontation and defense capabilities, equipment, a storage medium and a program product. The method comprises the following steps: converting a target image into a grey-scale map; adversarial region detection is carried out on the grey-scale map in a frequency domain, adversarial region mask maps of multiple groups of n-level frequency bands in the frequency domain are obtained, and the adversarial region mask maps of any group of any-level frequency bands are used for indicating the adversarial region of the grey-scale map under the level frequency band of the frequency domain; fusing the confrontation area mask graphs of the same group of n-level frequency bands in the plurality of groups of n-level frequency bands to obtain a plurality of target mask graphs, and based on each target mask graph, performing fuzzy processing on the confrontation area in the target image to obtain a plurality of preprocessed images; and respectively carrying out pedestrian detection on the plurality of preprocessed images through a pedestrian detection model, and carrying out non-maximum suppression fusion on a detection result to obtain a target detection result. Therefore, various adversarial attacks can be effectively defended, so that the adversarial defense capability of pedestrian detection is improved.

The invention discloses an ACLY-targeting oligopeptide and application thereof in bacterial infection, and belongs to the technical field of biological medicines. Three candidate peptides are obtained through molecular docking, virtual screening and the like, and it is found through experiments that the candidate peptide PVP can destroy ACLY-NLRP3 interaction and promote NLRP3 inflammasome activation, and the defense capacity of a host to bacterial infection is enhanced. The invention discloses a self-limiting loop of immune-metabolism cross dialogue, the ACLY-NLRP3 axis is positioned as a new therapeutic target of inflammasome driven diseases, and meanwhile, a new short peptide is provided for treating bacterial infection.

The application belongs to the technical field of security of deep learning neural network model, and in particular to a text deep learning adversarial training method and system based on malicious disturbance. The method constructs a malicious disturbance sample set based on original samples for training a deep learning neural network model, then randomly selects a malicious disturbance sample from the malicious disturbance sample set as a training input, replaces the original sample, and introduces a tiny benign disturbance sample into an embedding layer of the deep learning neural network model to generate a final adversarial instance containing the malicious disturbance sample and the benign disturbance sample, thereby improving the robustness of the deep learning neural network model, greatly improving the ability of the deep learning neural network model to resist malicious disturbance attacks, and reducing the risk of the model being misled by the adversarial instance. In addition, the method maintains the original task performance of the deep learning neural network model and maintains or even improves the performance of the model on the original task.

This invention provides a text processing method, a text processing device, and an electronic device, relating to the field of artificial intelligence technology. The method includes: acquiring multiple original samples, the original samples including text and labels; generating adversarial samples for each of the original samples; determining positive and negative samples for each of the original samples based on the adversarial samples; performing supervised contrastive learning and adversarial training based on the positive and negative samples to obtain a text classification model; and using the text classification model to classify the text to be processed. This invention improves the robustness of the text classification model.

The utility model discloses an anti-riot shield, which is suitable for the technical field of defense equipment, and comprises a shield main body for providing protection for resisting explosion, impact, violent confrontation and the like and a protruding area assembly arranged on the outer side surface of the shield main body, and the protruding area assembly comprises a protruding fixing part and a protruding body. Psychological deterrence is formed for violence and terrorism molecules through the protrusions on the surface of the shield, the intention of front impact of the violence and terrorism molecules is reduced, the defense effect is improved, the violence and terrorism molecules are forced to keep a distance, direct collision is avoided, and the near-body bucket winding risk is reduced; the state that the protruding bodies are completely embedded in the surface of the shield body and the state that the protruding bodies stretch out of the surface of the shield body can be switched, the shield is suitable for different scenes, the use range is expanded, and the flexibility of combat is improved.

The application discloses a kind of unmanned ship cluster specified time position decision-making method and device based on game theory, by converting unmanned ship cluster seeking optimal confrontation position problem into non-cooperative game problem, constructs water unmanned ship cluster game model, and designs distributed nash equilibrium solution algorithm under specified time;The method can solve the nash equilibrium of non-cooperative game problem in any specified time, quickly solve the optimal confrontation position problem of unmanned ship cluster seeking, enhance the attack and defense capability of our unmanned ship, improve the efficiency of decision-making;In addition, the initial time-varying gain item can be zero by using TBG function, which avoids the problem of excessive input at the initial time. Since the design and calculation of time-varying gain θ (t) are relatively simple, the application is relatively simple in structure and easy to implement.

This invention discloses an adversarial example generation method, system, storage medium, and terminal, belonging to the field of adversarial attack technology. The method includes: determining the context region of a target object and dividing it into multiple sub-regions; performing importance analysis on each sub-region; mapping the influence of each sub-region on the detection results of the target detection model to the context region, generating a context attribution graph; selecting one or more influential sub-regions exceeding an influence threshold; and fine-tuning the influential sub-regions under a total loss function (including a loss function and a perceptual loss function) to generate adversarial examples. This invention accurately guides the attack direction through the context attribution graph and introduces a dual loss function, suppressing position, direction, and confidence prediction while ensuring the visual imperceptibility of adversarial perturbations, thus preserving the integrity of the target object and enhancing the stealth of the attack. Attacking only some influential sub-regions improves attack efficiency.

The invention relates to the technical field of microorganisms, in particular to a multifunctional lactobacillus paracasei strain capable of promoting growth and development and application of the multifunctional lactobacillus paracasei strain. The Latin name of the multifunctional Lactobacillus paracasei strain is Lacticaseibacillus paracasei ZG1, the preservation number of the multifunctional Lactobacillus paracasei strain is CGMCC No.36863, the preservation date of the multifunctional Lactobacillus paracasei strain is December 5, 2025, and the preservation unit of the multifunctional Lactobacillus paracasei strain is China General Microbiological Culture Collection Center. Experiments prove that the multifunctional lactobacillus paracasei strain has the functions of promoting neurocognitive development, immune system maturation and bone growth, and can be added into an infant probiotic preparation to play a role in promoting growth and development. Compared with existing probiotics, the lactobacillus paracasei ZG1 has a multi-target and multi-system synergistic promotion function on growth and development, is expected to replace a traditional mixed probiotic scheme, and fundamentally solves the core problems of single function, mixed antagonism and the like of most of existing probiotics.

This invention discloses a steganography detection and attack method based on data format deviation, specifically including: acquiring a BinaryPlist file to be detected and reading the file as a binary byte stream; parsing the file header of the BinaryPlist file to determine whether the file conforms to a preset BinaryPlist file identifier format; parsing the offset table information in the BinaryPlist file to obtain the file offset address corresponding to each object in the object table; based on the offset address, parsing each object in the object table one by one to obtain the object type, object length, and object content; determining whether the offset relationship, object boundary, and object content of the object deviate from the format specification of the BinaryPlist file; if a format deviation is detected, determining the position of the corresponding steganography data block; and performing an attack operation on the steganography data block, including steganography data replacement or steganography data erasure, to destroy or block the steganography information.

The application discloses a kind of neural network model encryption and function fine-grained authorization method, based on initial training dataset, corresponding automatic trigger key generator or artificial trigger key dataset is constructed;Initial training dataset is mixed with the trigger key dataset or artificial trigger key dataset generated respectively, and mixed training dataset is obtained by construction: the mode is encrypted using the way of mixed training dataset training neural network model, so that no effective license can use model, more safe and reliable;On the basis of the encryption method, the application proposes a deep neural network dual verification method, which perfectly avoids existing adversarial attacks;In addition, the application also provides a fine-grained authorization method for using neural network function, according to the different functions of neural network, the corresponding trigger key is bound to realize the fine-grained authorization use of neural network, with the characteristics of only permitting users to use part of the function of the model.

The invention discloses a virus defense method, system and device and a computer readable storage medium, relates to the technical field of information security and storage protection, and is applied to defense software to detect whether a target host is attacked by a virus or not. In response to the fact that the target host is attacked by the virus, generating a hardware-level blocking signal; the hardware-level blocking signal is sent to a preset hardware device, so that the hardware device modifies firmware area parameters of the target host hard disk after responding to the hardware-level blocking signal, and then the target host hard disk refuses write-in operation on the physical level; wherein the communication link between the defense software and the hardware device is a link independent of the operating system of the target host, and the communication link between the hardware device and the hard disk of the target host is a link independent of the operating system of the target host. According to the method, independent defense which is not influenced by the authority of the operating system can be realized, the defense mechanism can still run normally even if the system is crashed or controlled, the defense capability of the hard disk to viruses is improved, and the security of the hard disk is ensured.

The application provides a kind of smart alarm method and system for baton, comprising: respectively collecting the current use state corresponding to each preset baton, respectively matching the management scheme corresponding to each of the preset baton, detecting the current action of the user using the corresponding preset baton according to the management scheme, establishing the action points of the user, establishing the use risk degree of the user according to the action points, when the use risk degree is greater than the preset degree threshold, positioning the current use position of the user and obtaining the monitoring video stream within the preset range of the current use position, analyzing the monitoring video stream to generate remote alarm help information, transmitting the remote alarm help information to the preset remote terminal for display and alarm, when the user uses the baton, the risk degree of the user when using the baton can be judged according to the action of the user, and when the risk degree is too high, it can reflect that the user is defending an emergency event, and remote alarm help is performed at this time.

The application provides a face living body detection method and device based on artificial intelligence, an electronic device and a storage medium. The face living body detection method based on artificial intelligence comprises the following steps: adding multiple filter colors to a collected face image to obtain a face test image set; performing feature extraction on images in the face test image set to obtain face test features; constructing a living body detection loss function based on the face test image set and the face test features; training a neural network based on the face test image set and the living body detection loss function to obtain a face living body detection model; and detecting a face image to be detected based on the face living body detection model to obtain a face living body detection result. The application trains the neural network in all directions by constructing multiple loss functions, so that the defense capability of the obtained face living body detection model against filter color attacks is improved, and the face fraud risk is reduced.

The invention discloses an adversarial sample defense method, system and device and a storage medium, and the method comprises the steps: matching an input sample of a target model with each adversarial sample stored on a block chain, and obtaining a sample matching result; determining a risk type of the input sample based on the sample matching result; based on the risk type of the input sample, an execution strategy about the input sample is determined for the target model, and the execution strategy includes indicating the target model to perform normal reasoning or defense processing by using the input sample. By means of the mode, the accuracy and generalization ability of target model confrontation sample recognition can be improved, and therefore the response speed and the defense effect of a model defense system are improved.

The invention discloses an automatic driving behavior decision-making method based on an intention-driven risk field. The method comprises the following steps: 1, eliminating geometric distortion by constructing a generalized bounding box; 2, introducing a momentum correction and intention coupling mechanism, and dynamically remodeling a risk field topology to simulate defensive driving psychology; and 3, scoring by using a space-time most disable principle, carrying out decision by combining with a dual risk threshold state machine, and introducing a decision reversal mechanism to break deadlock. According to the method, the problem of driving behavior decision oscillation is effectively solved, and the traffic efficiency and safety in a complex interaction scene are remarkably improved.

The application discloses an image data processing method and an image processing device, and belongs to the technical field of image processing. The image data processing method comprises the following steps: receiving a global model sent by a server end; training the global model based on local training images and actual labels corresponding to the local training images, so as to obtain a new local model, and the new local model is used for outputting a predicted label corresponding to the local training image and a loss value corresponding to a target loss function being not greater than a loss threshold; the target loss function is a loss function with an added regularization term; and the new local model is sent to the server end, and the new local model is used for being aggregated by the server end to generate a new global model, and the new global model is used for being sent to a client end. The image processing method improves output accuracy, reduces the dependence of the client end on the defense mechanism of the server end, improves the defense effect, and thus improves the reliability and safety of the model.

The application discloses a backdoor defense method based on a generative adversarial purification immune system, and belongs to the technical field of wireless communication and network security. In view of the problem that a deep reinforcement learning positioning model is vulnerable to data poisoning backdoor attacks, a generative adversarial purification immune system (GANPIS) defense framework is constructed. Firstly, the attacked model is taken as a discriminator to play a game with a generator, and a backdoor trigger feature is reversely reconstructed. Secondly, a positioning anomaly index is constructed based on the backdoor feature, and the poisoning samples in the training data are accurately detected and removed. Finally, a hybrid adversarial training strategy is adopted, and a dynamic trigger penalty term and a composite loss function are combined to purify and retrain the model. The method does not require attack prior information, effectively blocks the backdoor attack, significantly reduces the attack success rate, and completely maintains the high positioning accuracy of the model in the normal scene, and can be used in safety-sensitive scenes such as indoor navigation and intelligent security.

The invention relates to a transferable adversarial sample generation method and device based on wavelet transform and multi-scale filling, and a storage medium, and the method comprises the following steps: obtaining an input image, converting the input image from a spatial domain to a frequency domain through wavelet transform, and carrying out the random block disruption of a high-frequency component block, and obtaining a preprocessed image; mI-FGSM-based sample updating is performed on the preprocessed image, and in the updating process, a global random multi-scale filling method is used, MI-FGSM input is transformed from multiple angles of a space domain and a frequency domain, a plurality of substitution models are combined, combined gradient loss is calculated, and training is performed to obtain an adversarial sample. Compared with the prior art, the method has the advantages of remarkably improving the migration capability of the adversarial sample among the black box models, improving the attack success rate and the like.

This invention discloses a deep learning-based intelligent door lock face recognition encryption method and system, comprising: Step 1: acquiring face images and performing face detection, key point localization, and affine alignment processing to obtain standardized face images; Step 2: inputting the standardized face images into an improved MagFace network and introducing a fractal dimension adaptive embedding mechanism to generate fractal modulated face feature vectors; Step 3: performing fractal spectrum encoding processing on the fractal modulated face feature vectors to generate fractal structure feature vectors; Step 4: calculating feature stability indices and dividing the feature set into stable and sensitive feature sets; Step 5: generating binary identity codes and environmental disturbance verification codes; Step 6: generating authentication encoding sequences based on challenge parameters; Step 7: performing consistency matching and generating door lock control commands. This invention improves the stability and authentication security of face recognition.

This application discloses a method, apparatus, and computer program product for large-scale model security protection. The method includes: acquiring user input information, including original prompt words; filtering the user input information using a security filtering strategy; if the security filtering result is compliant, strengthening the original prompt words using a prompt word strengthening strategy to obtain strengthened prompt words; performing reasoning based on the strengthened prompt words using the target large-scale model to be protected to generate original output content; performing risk assessment using a fine-tuned security large-scale model deployed via a bypass based on the strengthened prompt words and the original output content; and making a decision output based on the security filtering result and the risk assessment result. This application significantly improves the security and reliability of large-scale model applications by constructing a multi-layered protection architecture driven by "rules and models"; and effectively balances assessment accuracy and performance loss by decoupling and parallelizing the fine-tuned security large-scale model deployed via a bypass with the main process.

This invention discloses a method and system for detecting covert poisoning attacks on large-scale RAG models, relating to the field of information security technology. This invention detects risky data injected by collaborative backdoor attacks by globally embedding and vectorizing knowledge base data, combined with clustering and topological feature recognition. The defense scheme directly detects structural anomalies in the embedding space. Specifically, the LDCC algorithm is used to perform clustering in the embedding space based on core point determination and neighborhood expansion mechanisms. This eliminates the need for a preset number of clusters and can automatically identify small and densely packed poisoned clusters. It effectively identifies the "independent and compact" topological fingerprints left by novel attack methods in high-dimensional embedding spaces, thereby improving the defense against these new attack methods and enhancing system security.

The application discloses a method for processing IPv4 network source routing attack behavior based on GENEVE, comprising the following steps: acquiring real-time network state information, calculating the maximum receiving distance threshold of the link node according to the network state information; setting the neighbor node parameter according to the maximum receiving distance threshold, preventing the attack node from establishing malicious false neighbors, and acquiring abnormal monitoring data; constructing a routing space monitoring model for the abnormal monitoring data, generating a jump space of routing jump through a backtracking method, and limiting the corresponding communication link capacity through the non-overlapping of the jump space and the communication delay; acquiring a path state matrix according to the communication link capacity, adjusting the weight coefficient according to the current network security state through the path state matrix, generating a network security state flow table, responding to the real-time routing jump defense strategy, and performing a deceptive response to the malicious probe packet, so that the attacker cannot obtain the real network topology information, and better defense effect is achieved.