Research And Implementation Of Malicious JavaScript Code Detection System For Applet Based On Deep Learning

In recent years,the Applets have developed rapidly because they can be used without installation.However,the existing detection methods of Applets are relatively backward.The traditional detection methods based on signature or pattern represented by various anti-virus software are susceptible to code confusion,while the current detection models based on deep learning are mostly general detection models.In the Applets detection scenario,some features such as unbalance of data set and multiple code repeats between versions cannot be adapted accordingly,resulting in loss of detection accuracy and efficiency.Based on the above background,this paper proposes a new detection model based on the tree convolutional network.Aiming at the problem of unbalanced data set in the Applets scenario,the oversampling method is used to expand the unbalanced data set in the Applets scenario,and the attention mechanism is used to make the model focus on malicious features.Aiming at the problem of multiple code repeats between versions in the Applets scene,the maximum impact tree of code is obtained by using the markup method,and the maximum impact surface of code is modeled,thus speeding up the speed of model detection.In this paper,a series of experiments have confirmed the excellent performance of the designed model in the Applets code detection scenario.In a series of evaluation indicators,the designed model in this paper has improved compared with the baseline model.At the same time,this paper also carries out the engineering realization of the Applets detection system.This paper ensures the feasibility,reliability and stability of the whole system through rigorous architecture design and detailed test,and finally realizes the smooth landing of the whole detection system.