abstract |
The present invention discloses a screening method of malicious Android installation package APK based on convolutional neural network, applying deep convolutional neural network to APK modeling, using APK original data as input, using APK's Call Graph, and according to the security sensitive function Or high-frequency call (appearance) function call, design the corresponding deep convolutional neural network for automatic feature extraction, thereby detecting whether the APK is malicious; including the model training process and the APK detection process with the trained model. The method of the invention can reduce the steps of artificially extracting features, thereby reducing analysis time, improving universality and increasing accuracy. |