Research On Malicious Android Application Detection Method Based On Multi-class Image Features

At this stage,Android malware problems are emerging,and how to detect malware quickly and effectively is an important problem facing the current security field.In order to improve the detection effect of Android malware,this paper carries out research from three perspectives of feature selection,feature fusion and model improvement,and proposes the Android malicious application detection method based on the mixed feature image combining Markov image and grayscale image,and the Android malware detection method based on the residual network fusion attention mechanism,and a large number of experimental results prove the The effectiveness of the method is proved by a large number of experimental results.The main work of this paper is as follows:For the traditional Android malware detection based on static analysis method and dynamic analysis method,there are problems of malicious application confusion variants,this paper proposes an Android malware detection method based on Markov images combined with grayscale images.The method first uses Markov and B2 M algorithms to generate Markov and grayscale images respectively,and then uses transfer probability to fuse the two different images to construct a new feature texture space,which is mapped into a two-dimensional space to obtain a hybrid feature image.Since different kinds of feature images can yield different feature descriptions,constructing a fused feature space can more clearly represent and improve the description of malware.Finally,it is validated on the Drebin dataset,and the experimental results show that the proposed method in this paper is more effective and accurate when performing malware detection compared with a single feature image,with an accuracy rate of 97.54%.To address the problems that traditional malware detection techniques have relatively single models and are prone to low accuracy due to the influence of obfuscation techniques,this paper proposes an Android malware detection method based on a fused attention mechanism of residual networks.In order to avoid the problem of low accuracy caused by gradient degradation of the network during training,this paper introduces the attention mechanism in the residual network to deepen the depth of the network and thus improve the generalization ability of the network.Finally,the validation is carried out on the CICMal Droid 2020 dataset.The experimental results show that the method can efficiently perform uniform representation of bytecode sequence files and extract and classify feature sequences with an experimental accuracy of 97.97% compared with traditional malware detection models.The intelligent malicious application detection method based on multi-class image features proposed in this paper can effectively solve the problems of confusing variants and improving malware accuracy to a certain extent,and there is room for application in the development of mobile security.