32results about How to "Low detection false positive rate" patented technology

The invention relates to a multi-feature fusion phishing webpage detection method, which comprises two parts such as a training process and a detection process. The multi-feature fusion phishing webpage detection method integrates three views of phishing webpage characteristics by combining a semi-supervised learning tri-training method, and mainly solves a problem that the existing phishing webpage detection methods mostly need to perform classification model training by using supervised learning through a large amount of annotation data. The method provided by the invention mainly combines a coordinated training algorithm, starts from webpage URL characteristics, webpage information characteristics and webpage search information characteristics, applies the idea of multiple views and multiple classifiers to phishing webpage detection, and achieves the purposes of reducing the total numbers of manual annotation training samples and timely recognizing a phishing webpage through coordinated training and learning of different classifiers.

The invention relates to a financial fishing webpage detection method based on Web page characteristics. The method is based on pre-established financial first Title keyword library, second Title keyword library, sensitive keyword library and webpage Logo icon characteristic point rule library, and comprises the steps of obtaining an HTML of a to-be-detected webpage by employing a crawler, extracting text information of a Title label, calculating a matching degree with the first and second Title keyword libraries, and if the matching degree is greater than a threshold value, judging as a fishing webpage, or otherwise, going to the next step of detecting; extracting the text information of a special label of the to-be-detected webpage, making a statistics on a matching number with the sensitive keyword library, calculating a sensitive characteristic value, and if the characteristic value is greater than the threshold value, judging as the fishing webpage, or otherwise, going to the next step of detecting; and carrying out fixed point interception on the to-be-detected webpage, obtaining a Logo icon of the to-be-detected webpage, extracting characteristic points of the Logo icon, comparing with the icon characteristic point rule library, calculating a similar degree according to the matching number of characteristic points, and if the similar degree is greater than the threshold value, judging as the fishing webpage, or otherwise, as a normal webpage. According to the financial fishing webpage detection method based on the Web page characteristics related by the invention, whether the to-be-detected Web page is the financial fishing webpage can be judged accurately and quickly.

The invention discloses a subway fare evasion behavior detection method and system based on infrared thermal imaging. The method comprises the following steps: detecting whether a pedestrian enters agate image pedestrian detection range or not; performing infrared thermal imaging image background difference by utilizing the automatically updated background, and extracting a pedestrian infrared thermal imaging top view image; carrying out morphological processing on the extracted pedestrian infrared thermal imaging top view image to obtain a binarized pedestrian top view image passing througha gate opening based on automatic updating of an appropriate threshold value; performing parallel ROI (Region Of Interest) extraction on the binarized pedestrian top view image, and setting the ROI tobe the same as gate ports in number (N) to obtain N mutually independent binarized gate passage pedestrian top views; carrying out connected region marking on the N mutually independent binary gate passage pedestrian top view binary images to acquire pedestrian parameters; and judging fare evasion behaviors of the N gate ports. According to the invention, fare evasion behaviors can be effectivelyidentified and avoided.

The invention provides a network intrusion detection method based on multilayer immunization. The method includes: an innate immunization layer performs autologous clustering on a network normal data packet to generate a pattern detector set so as to realize quick response to a known attack; a self-adaptive immunization layer generates a mature detector set used for network data packet abnormality detection through secondary negative selection so as to eliminate network attacks and intrusion that the innate immunization layer cannot eliminate; and a detection response layer uses the pattern detector set generated by the innate immunization layer and the mature detector set generated by the self-adaptive immunization set to perform parallel and dynamic detection on network data to be detected, and feeds back to the innate immunization layer and the self-adaptive immunization layer according detection results so as to improve system detection efficiency and quality.

The invention relates to an early fire smoke detection method based on a semi transparent feature. The method includes 1, acquiring an image as an original one; 2, calculating a recover image of the original image, if the original image is a first frame image being acquired, storing the image and the recover image thereof as a background image and the recover image thereof, repeating the step 1, if not, executing a step 3; 3, acquiring the similarity of the corresponded small regions of the two recover regions; 4, establishing a dissimilar block marking image IO1 different to the recover image in size; 5, calculating the similarity of each image block of the original image and background image; 6, establishing a dissimilar block marking image IO2 different to the original image in size; 7, acquiring image semi transparent regions; 8, searching for a communication region in the semi transparent regions, and calculating the area; 9, searching for a complete smoke region; 10, calculating a dynamic image; 11, outputting alarm signals, and marking the smoke region location in the original image. The method can be applied to the fields of petrochemical factories, large warehouses, large-scale public places and so on.

The invention discloses a host intrusion detection method and device, and relates to the technical field of computers. One specific embodiment of the method comprises the following steps: acquiring signal data of a host, wherein the signal data comprises static signal data n and dynamic signal data s; searching a static detection rule rets matched with the static signal data from a preset featurelibrary; obtaining a dynamic prediction result retd according to the dynamic signal data and an SVM model; wherein the training parameters of the SVM model are obtained by adopting a particle swarm optimization algorithm; and judging whether the current process of the host is an attack behavior or not according to the static detection rule rets and the dynamic prediction result retd. According tothe embodiment, the detection rate of unknown attack behaviors can be improved; the detection false alarm rate of normal operation is reduced, and the stability of normal service is ensured; the detection rule is automatically generated and the detection feature library is updated when the attack behavior is detected, the manual intervention cost is reduced, and the response time to the unknown attack behavior is shortened.

The invention relates to the technical field of vehicle-mounted network intrusion detection, in particular to a vehicle-mounted controller local area network intrusion detection method based on a sliding window and a convolutional encoder neural network. The method comprises the steps: carrying out the data cleaning of received original data, and selecting a CAN ID and a CAN Data as features; converting the characteristics of each dimension of the CAN ID and the CAN Data into binary data; calculating a feature size according to the digit of the converted binary data; setting a square window according to the obtained feature size, and performing sliding window processing on the data by using the window to obtain processed features; inputting the processed features into a convolutional encoder neural network, wherein the convolutional encoder neural network obtains a detection result according to the input data. According to the invention, the abnormal flow and the normal flow can be discriminated with a low false alarm rate and a high detection rate, and the purpose of CAN network intrusion detection is achieved.

The invention discloses an agent user detection method based on a communication behavior portrait. The method comprises the following steps: constructing a portrait model based on the communication behavior; preprocessing the input real traffic data, and aggregating the network traffic according to organized user IP attributes and time attributes in a specified time window to obtain traffic data of each user; extracting communication object stability characteristics and communication object quantity characteristics of a user in a specified time window based on the flow data, and then filteringand screening the two characteristics based on a threshold value to obtain IP information of a suspected proxy user; performing machine learning detection classification on the network flow of each suspected proxy user; and marking the IP of which the flow judgment result satisfies the threshold setting as the IP of the proxy user. The detection method provided by the invention is high in universality, can meet the data volume in the current big data network environment, adopts the portrait model based on the communication behavior, and is low in detection false alarm rate.

An embodiment of the invention provides an API-based Android malicious program detection method, includes decompiling APK file to get smali file, obtaining sensitive API set by Pscout tool, matching API in smali file with API in sensitive API set, composing matching sensitive API set by successful matching API. LibD tool is used to eliminate the API called by the third party library in the matching sensitive API set, and the purified matching sensitive API set is obtained. According to the purified matching sensitive API set, the first eigenvector is constructed, and TF-IDF algorithm reconstructs the first feature vector to obtain a second feature vector;taking the second feature vector as training set, the training depth learning model is used to get the effective classification model, and the effective classification model is used to detect the Android program and obtain the Android malicious program. The embodiment of the invention also provides an active interactive device and a non-transient readable storage medium for realizing the method. The invention can accurately detect the Android malicious program.

The invention provides a webpage feature tampering detection method and device, electronic equipment and a storage medium, and the method comprises the steps of obtaining webpage sample data, and building a webpage sample data set document; extracting text features, structural features and network features from the data set document; and mining a nonlinear relationship among the text feature, thestructure feature and the network feature, and further detecting the tampering behavior category of the webpage feature. The method is advantaged in that through a heterogeneous feature classificationalgorithm based on deep learning, text features, structural features and network features are effectively fused, different types of webpage tampering behaviors are recognized more accurately, and thefalse alarm rate of detection is reduced.

The invention discloses a block chain transaction monitoring method and system based on static characteristics and dynamic instrumentation, and the method and system reduce the detection false alarm rate of malicious transactions to the greatest extent through transaction data static semantic analysis and transaction replay dynamic execution detection, distinguish malicious transaction and normal transaction behaviors more accurately. The transmission behavior of sensitive data in the transaction is detected by intelligent contract virtual machine instrumentation and dynamic taint analysis methods; and meanwhile, a machine learning method is utilized, return results of malicious transaction behaviors and normal transaction behaviors detected by a system dynamic and static method are analyzed, and a final study and judgment stage is added, so that whether the transaction is the malicious transaction or not can be effectively judged. The invention can detect the transaction of the corresponding smart contract of the block chain, assists a machine learning algorithm to effectively improve the detection accuracy in a mode of combining static feature analysis and dynamic execution detection and analysis, and has a significant effect on improving the detection accuracy.

The invention discloses a fluorescent biosensor for simultaneously detecting two kinds of RNA as well as a preparation method and a use method thereof. When a target detection object miRNA appears in a reaction system, the capture chains cDNA1 and cDNA2 are respectively combined with the target microRNA, and the fluorescence of the CQDs and the Cy5 is recovered. Meanwhile, double-strand specific nuclease (DSN) is used to act on the double strands formed by the pDNA and the target MicroRNA, so that the microRNA participates in the next round of reaction, and the signal is amplified. The construction method provided by the invention is simple, the detection sensitivity is high, the expression quantity of two miRNAs in a sample can be quantitatively detected at the same time, the reaction condition is mild, the sensing system is stable and effective, and compared with expensive and tedious clinical gene sequencing, the method has great practical significance and development potential and is beneficial to popularization and application.

The invention relates to an anti-dry-burning gas stove which comprises an inner fire cover, a temperature detection piece and an air guide assembly. The temperature detection piece and the inner firecover are correspondingly arranged, the temperature detection piece is used for detecting the temperature of a target position on a pot placed above the inner fire cover, a through hole is formed in the inner fire cover, and the through hole penetrates through the bottom face and the top face of the inner fire cover; an air guide duct is arranged in the air guide assembly, an air outlet of the airguide duct is located near the target position, and the air guide duct is in butt joint with the through hole, so that airflow in the through hole can only flow to the air outlet end along the air guide duct, so that the air circulation of the target position is improved, heat generated by combustion flames is taken away, and the detection misjudgment probability of the temperature detection piece is reduced. In the whole process, guiding is conducted based on physical structures such as the air guide assembly, the purpose of reducing the misjudgment probability is achieved, the situation that the physical structures lose the guiding function or the guiding function is reduced due to the fact that the temperature in the using environment is high is avoided, and therefore the service lifecan be guaranteed.

The invention provides a template reconstruction method based on a self-attention mechanism. The method comprises the steps of firstly, performing layer-by-layer down-sampling on an input image to obtain a down-sampled image; performing feature extraction on the down-sampled image; then respectively performing table look-up mapping on features of each layer by adopting a self-attention mechanism to obtain a residual error corresponding to each layer, finally, performing layer-by-layer up-sampling on the downsampled image, and meanwhile, fusing residual error information of each layer to generate an up-sampled image corresponding to each layer, thereby finally generating an up-sampled image, namely a reconstructed template image.

The invention relates to the fields of gene engineering and molecule biology, and provides a KRAS gene detection kit and a detection method. The kit has the advantages that a specific amplification primer sequence is designed according to the specific sites of No.2 exon and No.3 exon regions of KRAS, so that the false positive rate of the to-be-detected sites is low in detection, the sensitivity is high, and the gene mutation with mutation ratio of 5% or above can be detected; library molecules are respectively obtained from an F terminal and a R terminal at both ends of a KRAS nucleic acid segment of the to-be-detected site, and the F terminal and the R terminal respectively contain amplification primer binding sites; when the KRAS nucleic acid segment containing the multiple to-be-detected sites is sequenced, the amplification primers for amplifying the library molecules are unified, the consistency in amplification efficiency is guaranteed, and the difficulty in design of the primeris decreased; the R terminal is of a hairclip structure, and can be used for sealing the tail end of the corresponding amplification product, so that the F terminal is effectively prevented from being connected with the R terminal, and the connecting efficiency is improved.

The present invention relates to the technical field of vehicle network intrusion detection, in particular to a vehicle controller local area network intrusion detection method based on a sliding window and a convolutional encoder neural network, including: performing data cleaning on received raw data, and selecting CAN ID and CAN Data as a feature; convert the features of each dimension of CAN ID and CAN Data into binary data; calculate the feature size according to the number of digits of the converted binary data; set a square window according to the obtained feature size, and use this window to perform a sliding window on the data processing to obtain the processed features; the processed features are input into the convolutional encoder neural network, and the convolutional encoder neural network obtains detection results according to the input data; The detection rate distinguishes abnormal and normal traffic to achieve the purpose of CAN network intrusion detection.

The invention provides a data detection method, a baseline model construction method and electronic equipment. The data detection method comprises the following steps: acquiring to-be-detected HTTP data; comparing the feature parameters of the to-be-tested HTTP data with feature parameters of a preset baseline model, and determining a suspicious score of the to-be-tested HTTP data; and outputting abnormal data; wherein the abnormal data are to-be-tested HTTP data of which the suspicious score is greater than a preset threshold value. Compared with data detection through a simple character string comparison mode in the prior art, the mode has the advantages that the detection reliability and the detection rate of abnormal data can be improved, and the false alarm rate of HTTP data detection can be reduced.

The invention discloses a method and device for image registration of industrial prints based on deep learning. Specifically, according to corresponding cutting rules, slices of the same size are respectively extracted from a template image and an image to be registered to obtain a group of slices. Image slice pairs are obtained after deep fusion, and the image slice pairs are input into the registration network model for training to obtain the registered image slices; the edge parts of the registered image slices are discarded, and then the slices are aligned and spliced ​​to obtain a complete Registered print image of . The framework of the registration network model of the present invention is a UNet-like network, and the spatial transformation layer is used to perform non-rigid registration on feature maps of different scales, and the registration feature map is fused with the feature maps of adjacent scales in the decoder, and at the same time, the adjacent scales are aligned The quasi-field is also fused, which comprehensively improves the registration ability of the model for images of larger deformed prints. The invention can solve the problems existing in image registration of some industrial paper printed matter at present, and the registration effect is good.

The invention relates to a multi-feature fusion phishing webpage detection method, which comprises two parts such as a training process and a detection process. The multi-feature fusion phishing webpage detection method integrates three views of phishing webpage characteristics by combining a semi-supervised learning tri-training method, and mainly solves a problem that the existing phishing webpage detection methods mostly need to perform classification model training by using supervised learning through a large amount of annotation data. The method provided by the invention mainly combines a coordinated training algorithm, starts from webpage URL characteristics, webpage information characteristics and webpage search information characteristics, applies the idea of multiple views and multiple classifiers to phishing webpage detection, and achieves the purposes of reducing the total numbers of manual annotation training samples and timely recognizing a phishing webpage through coordinated training and learning of different classifiers.

The invention discloses a high-precision leaked liquid detection and recovery early warning device, which comprises a leaked liquid confluence port, a leaked liquid confluence channel, a leaked liquidcollection tank and a detection sensor. The leaked liquid confluence port is communicated with the leaked liquid confluence channel, the tail end of the leaked liquid confluence channel is provided with the leaked liquid collection tank, and the detection sensor is fixed above the leaked liquid collection tank. The high-precision leaked liquid detection and recovery early warning device can accurately judge and detect whether liquid permeates into the liquid cooling modules in the case and the cabinet or not, the misjudgment rate of traditional liquid leakage detection is reduced, and liquidleakage reminding is carried out in time.

The invention relates to a laser Raman spectrum analysis processing method, comprising the following steps: 1) performing baseline correction processing based on an adaptive iterative reweighted penalty least squares method and smoothing processing based on a penalty least squares method on the Raman spectrum original signal ; 2) performing peak detection based on continuous wavelet transform to the signal processed in step 1). Compared with the prior art, the invention has the advantages of fast speed and high peak detection precision.

The invention discloses an industrial printed matter image registration method based on deep learning and a device thereof, and the method specifically comprises the steps: extracting slices with the same size from a template image and a to-be-registered image according to a corresponding cutting rule, obtaining a group of slices, carrying out the deep fusion of the group of slices, and obtaining an image slice pair, inputting the image slice pair into a registration network model for training to obtain a registered image slice; and rejecting edge parts of the registration image slices, and then performing slice alignment splicing to obtain a complete registration printed matter image. The framework of the registration network model is a UNet-like network, non-rigid registration is carried out on feature maps of different scales by using a spatial transformation layer, the registration feature maps and feature maps of adjacent scales in a decoder are fused, and registration fields of the adjacent scales are also fused, so that the registration capability of the model on large-deformation printed matter images is comprehensively improved. According to the method, the problems existing in image registration of part of industrial paper printed matters at present can be solved, and the registration effect is good.

The invention relates to a torsion spring assembly detection mechanism suitable for a vehicle door locking safety device. The torsion spring assembly detection mechanism comprises a base, a clamping tool, a torsion test function part and a force measurement assembly. The clamping tool is used for clamping the vehicle door locking safety device arranged in the clamping tool. The torsion test function part comprises a driving arm and a power unit, wherein the driving arm moves in the front-back direction, deflects in the circumferential direction under the driving force of the power unit and isused for driving the swing plate to rotate by a preset angle. The force measurement assembly comprises a mounting plate, a front force transfer rod and a front pressure sensor. The mounting plate is fixed on the base, and a front insertion hole extends downwards from the upper plane of the mounting plate. The front force transfer rod and the front pressure sensor are sequentially inserted into thefront insertion hole and abut against each other. When the swing plate rotates to the preset angle, the swing arm abuts against the front force transfer rod under the combined action of the torsionalmoment of the first torsional spring and the torsional moment of the second torsional spring. The front pressure sensor is used for sensing the jacking force borne by the front force transfer rod.

The invention discloses high-voltage resistance circle detection equipment which comprises a rack. The rack is provided with a rotation assembly, a lift assembly and a pushing assembly. The rotation assembly is connected with a square rotation base. The four side surfaces of the square rotation base are respectively provided a hanger assembly. Each hanger assembly comprises magnetic hangers in horizontal / vertical arrangement. The rotation assembly can drive the square rotation base and the hanger assemblies to realize intermittent rotation, and cooperatively drive products adsorbed on the magnetic hangers to rotate among a loading station, a test station, a draining station and an unloading station. The lift assembly and the pushing assembly are arranged on the test station and the unloading station respectively. The lift assembly is movably docked to the hanger assembly rotated to the test station, and the magnetic hangers of the hanger assembly are electrically connected with a high-voltage power supply. The lift assembly can drive the hanger assembly to carry products to be vertically and lively immersed into or removed out of water tanks to realize high-voltage resistance testcooperatively. The pushing assembly can reject products carried by the hanger assembly rotated to the unloading station in a classified manner, and cooperates with a material stop assembly to finish unloading of qualified products / defective products.

The invention relates to the technical field of power transmission and transformation equipment, in particular to a power transmission and transformation equipment Internet-of-things network access consistency detection method and device. The method comprises the following steps of: firstly simulating system modules necessary for achieving protocol consistency detection, then acquiring the state data of power transmission and transformation equipment through a data collection module, storing the acquired data in a data storage module, and generating a unique input / output sequence corresponding to all states of the communication protocol standard of the Internet-of-things terminal equipment. Compared with a traditional manual detection means, the method has the advantages that the detection efficiency can be obviously improved, the detection misjudgment probability is reduced, and a more complete and detailed detection result report can be provided. The problems that the power transmission and transformation equipment Internet-of-things network access consistency is generally detected manually at present, the labor intensity of workers is high, the protocol consistency cannot be comprehensively detected, and the probability of misjudgment during detection is high, so that the detection efficiency of the power transmission and transformation equipment is low are solved.

The invention discloses an agent user detection method based on a communication behavior portrait. The method comprises the following steps: constructing a portrait model based on the communication behavior; preprocessing the input real traffic data, and aggregating the network traffic according to organized user IP attributes and time attributes in a specified time window to obtain traffic data of each user; extracting communication object stability characteristics and communication object quantity characteristics of a user in a specified time window based on the flow data, and then filteringand screening the two characteristics based on a threshold value to obtain IP information of a suspected proxy user; performing machine learning detection classification on the network flow of each suspected proxy user; and marking the IP of which the flow judgment result satisfies the threshold setting as the IP of the proxy user. The detection method provided by the invention is high in universality, can meet the data volume in the current big data network environment, adopts the portrait model based on the communication behavior, and is low in detection false alarm rate.

The invention discloses a method for reducing the false positive rate of breast ultrasonic focus detection. The method comprises the following steps: tracking a suspected focus in a video by using a KCF tracking algorithm; extracting sub-videos according to a tracking result and preprocessing the sub-videos; inputting the preprocessed sub-videos into a CSN network for prediction; if the probability of the genuine yang category output by the CSN is greater than a preset threshold value, judging that the suspected focus is genuine yang, and if the probability of the genuine yang category output by the CSN is less than or equal to the preset threshold value, judging that the suspected focus is false yang. According to the method for reducing the false positive rate of breast ultrasonic focus detection provided by the embodiment of the invention, the inter-frame information is utilized to screen the results of the single-frame detection algorithm, so that the false positive rate is greatly reduced.