GANG-MAM: GAN based enGine for Modifying Android Malware
G. Renjith, Sonia Laudanna, S. Aji, Corrado Aaron Visaggio, P. Vinod
Abstract
Malware detectors based on machine learning are vulnerable to adversarial attacks. Generative Adversarial Networks (GAN) are architectures based on Neural Networks that may be used to produce successful adversarial samples. The interest towards this technology is quickly growing. This paper proposes a system that first generates feature vectors for making Android malware strongly evasive and then modifies the malicious program accordingly. The system has a twofold contribution: it could be used to build data-sets for validating detectors of GAN-based malware and to enlarge the training and testing data-sets to improve the robustness of malware classifiers.
Topics & Concepts
MalwareComputer scienceAndroid (operating system)Android malwareRobustness (evolution)Adversarial systemCryptovirologyArtificial neural networkMachine learningArtificial intelligenceComputer securityOperating systemGeneBiochemistryChemistryAdvanced Malware Detection TechniquesNetwork Security and Intrusion DetectionSoftware Testing and Debugging Techniques