MalPurifier: Adversarial Purification for Android Malware Detection
Researchers propose MalPurifier, an adversarial purification framework to enhance Android malware detection against evasion attacks. The framework integrates three innovations: a diversified adversarial perturbation mechanism, protective noise injection for benign data, and a Denoising AutoEncoder with dual-objective loss. It aims to improve robustness and generalization over existing defensive methods.
Key facts
- MalPurifier is an adversarial purification framework for Android malware detection.
- It addresses evasion attacks on ML-based detection systems.
- Three innovations: diversified adversarial perturbation, protective noise injection, DAE with dual-objective loss.
- Aims to improve robustness and generalization.
- Published on arXiv with ID 2312.06423.
Entities
Institutions
- arXiv