New Framework Improves Android Malware Detection Stability
/ 1 min read
🌀 New framework enhances Android malware detection through temporal invariance. Researchers have developed TIF, the first temporal invariant training framework designed to improve the stability of malware detectors against distribution shifts caused by evolving malware variants. Traditional classifiers struggle with these shifts due to their reliance on empirical risk minimization, which fails to capture stable features. TIF addresses this by organizing environments based on application observation dates and employing multi-proxy contrastive learning to generate high-quality representations. Experiments over a decade-long dataset demonstrate TIF’s effectiveness, particularly during early deployment stages, outperforming existing methods and meeting real-world detection needs. This framework can be integrated into any learning-based malware detection system.
