Singapore Academic Cybersecurity R&D
Harnessing R&D to Secure our Nation
Advanced Anti-Malware Solution Using Deep Learning
- Lead PIs : Ngair Teow-Hin, SecureAge and Ooi Beng Chin, Professor, NUS ( firstname.lastname@example.org )
- Host Institution : SecureAge
- Partner Institution : School of Computing, NUS
The cyber threat landscape is constantly changing, and the cyber security industry’s focus is shifting from prevention to detection. Malicious software remains the most popular and damaging attack vector, costing hundreds of billions in damage. The main existing antivirus solutions can have high accuracy but are ineffective in detecting new malwares in timely manner.
Our aim in this project is to develop an anti-malware solution based on deep learning technologies that is better than existing antivirus engines. More specifically, the solution employs deep learning to achieve the following three goals:
- High accuracy at identifying fresh and stealthy malwares.
- Timely detection of new malwares, while incurring small memory and CPU footprint.
- Adapting quickly to new trends and changes in malware population.
We propose a hybrid model for malware detection with an autoencoder and an attention based CNN. Experimental results demonstrate that the proposed model is effective for malware detection. The test AUC score is 99.80% and the test recall rate is 87.12% when false positive rate is lower than 0.1%, which are significantly better than existing machine learning approaches. We are going to investigate the dynamic malware behavior in the next stage.