Malware Detection in Android Using Machine Learning on Chip

Abstract

The widespread use and general-purpose computing capabilities of smartphones make them the next big targets of malicious software (malware) and security attacks. Malware detection software in the desktop environment is computationally intensive and the ones in the cloud require good Internet connectivity. But given the battery, computing power, and bandwidth limitations inherent to mobile devices, malware detection on these devices using these conventional methods are not pursued. In our proposed method, to detect malware, it is planned to synthesize the feedforward neural network on a Field-Programmable Gate Array (FPGA). A neural network is trained on a powerful desktop computer and the model obtained is used to generate the Intellectual Property (IP) core. This proof of concept work is done in such a way that the obtained intellectual property core can be embedded later on a mobile system on chip (SoC) as a dedicated malware processing unit. The hardware simulation showed accuracy around 89.59% using a simple Artificial Neural Network (ANN) with one hidden layer on Digilent’s Zybo evaluation board with Xilinx Zynq-7000 family AP SoC (All programmable SoC).