The research paper presents an active learning framework for Android malware detection called ActDroid. With the growing popularity of Android, the need for effective malware detection systems has increased. The authors treat Android malware detection as a streaming data problem and use active online learning to label applications in a timely and cost-effective manner. The framework achieves up to 96% accuracy and requires only 24% of the training data to be labelled. It also compensates for ‘concept drift’, the change in malware over time. The study further explores the practicalities and trade-offs of online learning within Android malware detection.
Publication date: 1 Feb 2024
Project Page: https://arxiv.org/abs/2401.16982
Paper: https://arxiv.org/pdf/2401.16982
