Deep Learning for Mobile Malware Detection

Queen’s University Belfast, ECIT, Centre for Secure Information Technologies

Dr. Niall McLaughlin

Lecturer, School of Electronics, Electrical Engineering and Computer Science

The number of mobile malware apps is growing every year and it is impossible for human analysts to keep up. At Queen’s University Belfast, we are using artificial intelligence and deep learning to hunt for malware in mobile apps.

We designed a custom deep neural network that learned to identify malware by scanning the code of tens of thousands of malware apps – far more than any person could ever examine by hand. The network learned the subtle cues in the code for itself, without human experts to tell what to look for. It combines these cues with permissions, API calls and manifest information to reach a more accurate decision. Our approach is platform agnostic, so it can be used on any mobile platform where training data is available. We have shown that our approach is resilient against counter-measures such as encryption and is effective against zero-day attacks.

Using the GPU in all mobile phones, we can efficiently scan thousands of apps per-second. Our neural network runs stand-alone on the end user’s mobile device, without needing a network connection.

Our solution will support the continued rapid growth in Industrial IoT for the smart city, and significantly improve security and data protection. It enables enterprise-users to secure IoT devices across their network, but also enables individuals to scan the apps on their device without the need to upload their personal data to the cloud. Thus, we empower users to take control of the security of their devices.

If you would like to learn more about how your University can participate in a future Mobile World Scholar Challenge, send us a message at [email protected]. Please include your name, title and university in the email.

Download the official #MWC19 event app