Intrusion Detection System Based On Genetic Neural Networks Applied Research

With the development at full speed of scientific and technological progress and network technology of the computer, information industry and using getting enormous development, all come with it, the network attacks and develops constantly too, hacker's means and tool are more and more advanced, or attack the system, or steal personal information, people have already realized that guarantee the importance of the online security.Network safe practice have encryption, identity distinguish technology, visit control technology and fire wall technology, etc. mainly at present, but they are greater defect and deficiency. Invade the development of the detection technique, can remedy some from a certain angle insufficiently, for example invade detection system have initiative on one hand. Involve a great deal of, the combination application study of reaching from research of invading the detection system itself with other relevant disciplines in research of invading the detection technique at present, have promoted the development of the detection system of invading in various degree.Invade developing history, classification, standardized process and data collection mechanism and analytical method of the detection system in thesis survey, have proposed a detection method of invasion based on neural network of heredity, because of the limitations of inherent learning method of the neural network, improve neural feedforward of network go against, propagate algorithm to be very difficult fundamentally. Utilize hereditary extensive search of suitability of algorithm and optimize method, etc. advantage come, overcome feedforward against propagate algorithm disappear slowly and exchanging the defect that the part disappear most, go against, propagate combination of algorithm solve, utilize hereditary algorithm can find, dose the optimum problem that solves within short time often alone too with feedforward at the same time. Utilize the feedforward to go against the gradient information which propagates the algorithm and avoid this kind of phenomenon. Measure result indicate method this result good, study fast, categorised rate of accuracy high, rate of failing to report and rate low wrong report.