Research And Implementation Of Network Sensitive Information Filtering

With the popularization of internet and the fast increase of network users, network information security becomes an important problem which affects nation security and society steady. The sensitive information of network, especially erotic information, disorder society method and destroy society genius, brings infinitely negative infection to people especially young children. Under this background, this paper deploys the research on techniques of network information filtering, main point to erotic information filtering. And then designs a set of networking information filtering system.The main work of this paper is as follows:Pattern match algorithms are broadly used in text content filtering,at the same time pattern match algorithms in the use of packet filtering system are effectively.But now pattern match algorithms cannot satisfy with mass text match,under the analysis of classical pattern match algorithms,A fast algorithm to perform multiple patterns match in a string was described. The proposed match algorithm is based on the concept of deterministic finite state automata. The actual experiments show that in the case of common instance, the time it takes to search a string is a half of the time DFSA method takes, efficiently improved the matching speed.The main character of erotic images is contain a large size of exposed skin,so skin detection can be used to erotic image detection.Additional skin detection method only use Pixel-level color information which may mistakenly classified things resemble human skin in color as human skin.a skin detection method based on Bayes decision in YCgCr color space is proposed. Experimental results have demonstrated that this method can provide lower false dismissal rate and lower false detection rate, compared with simple color space model and statistical model in YCbCr color space.Nude image identifying is a classify problem, our task is to find the decision rule that can optimally separates erotic images from benign images. Though most nude images contain large Patches of skin, relying solely on color information to identify nude image is insufficient.we extract skin-related features from the output of the skin detector,extract texture features by use of wavelet packet analysis, extract shape features from Hu invariant moment.These features compose a joint feature vector fed to SVM classifier for training and testing, Experimental results have demonstrated that this method can provide higher correct detection rate and lower false detection rate.Based on the classify of images,the paper designs a set of network information filtering system. The filtering system synthesizes the image filtering techniques and traditional filtering techniques, such as keyword filtering, packet filtering and URL blocking, so that it can filter text and image networking information fairly well. The filtering system is composed of 4 modules, i.e., data acquisition, data analysis, access controls and content identify. Because of the filtering system's adoption of Winsock2 SPI, it can capture and filter networking data packets conveniently.