Exploiting Neural Network Model for Hiding and Triggering Malware

In the recent years, security issues have been increasing with the rapid development of artificial intelligence technology. Malware delivery is important for exploitation of the target and must be delivered covertly and evasively. Existing work has given possibility of embedding malware into a neural network model with limited impact on the performance. With the wide application of artificial intelligence, utilizing neural networks for attacks becomes emerging trend. In this paper, we propose attack methods for target exploitation using neural networks models. This work can provide a reference scenario for the defense on neural network assisted attacks.