基于串行自编码器的Web服务注入攻击检测OA

Web services generate a large amount of data every day.Injection attacks bypass normal security checks by construc-ting special inputs that are highly consistent with normal data.The current data exist in plaintext and lack a hidden process,which increases the difficulty of detection.Therefore,a Web service injection attack detection method based on serial autoen-coder is proposed.This method uses support vector machine to train Web service page features to achieve automatic extraction of Web service data information.A serial autoencoder is employed to map the input sequence into latent variables,and a decod-er reconstructs them,automatically deriving the anomaly score of the data to be detected from the original data,thereby extrac-ting the injection attack location in Web services.An injection attack detection alarm threshold is set,and a template matrix and a detection matrix are formed to calculate their similarity,distinguishing hidden attack information from massive Web service data.If the similarity exceeds the alarm threshold,it is sent as a message event to an expert system,which integrates artificial intelligence for verification to obtain attack detection results.Experimental results show that the proposed method achieves a true positive rate of 100％,an F1-score of around 0.9,and a performance loss of less than 10％,indicating high detection accu-racy,fast speed,and the ability to perform adaptive injection attack detection without modifying the Web service execution en-gine.