Research And Implementation Based On Neural Network For SQL Vulnerability Detcting Problem

Web technology has been widely used in people's work and life. With the rapid development of Web technology, all kinds of security problems have become more and more serious, and SQL vulnerability injection attacks have become one of the most serious problems in the Web security issues. The problems of detection on SQL vulnerability injection are researched in this paper. An analysis and detection method of SQL vulnerability injection attacks based on neural network is proposed,aimed to improve the speed of vulnerability detection and improve the comprehensive nature of vulnerability detection type. The main research contents of this paper are as follows:(1) A kind of SQL injection vulnerability analysis model based on the artificial neural network is put forward. First of all,feature vectors are formed by extracting the keywords based on the the characteristics of the SQL keyword injection attacks. Then a large number of feature vectors of known SQL injection datas and artificial neural network algorithm are used to train analysis model, judging whether SQL injection statement.(2) In view of the complexity and variety of SQL statement types, the method generating SQL sample document is proposed based on method of input variation, to solve the prolems of time consuming and low efficiency caused by the SQL injection sample statements which is directly online collection. Attack SQL statements are obtained by performing input variation on the normal SQL statement through various variation rules. efficiently generating a more comprehensive sample of attack statements.(3) Based on the above principles, a neural network detection model based on neural network is proposed. Through analysing the test result of the system, the SQL injection vulnerability detection method based on neural network model makes contibution to reducing the cost of false alarm rate and time overhead.