Abstract:

Aiming at the low detection accuracy of traditional Android malware detection technology, an Android malware detection model based on the Dual-channel Convolutional Neural Network (DCNN) is proposed. First, it extracts the software original opcode sequence and generates the command function sequence. Then, it uses these two sequences as the input to the two channels of the convolutional neural network to iteratively train and adjust the neurons weights in each layer. Finally, the trained detection model implements the Android malware detection. Experimental results demonstrate that our detection model has a good detection accuracy and detection precision for malware.

Key words: malware, classification detection, opcode sequence, command function sequence, convolutional neural network