The paper presents a methodology for evading malware detection models using active opcode insertion and reinforcement learning based MalAOI. They develop a function to classify a malware sample as benign software by modifying it to generate a new sample, while retaining functionality and introducing minimal additional load. A reinforcement learning environment enables automatic selection of suitable insertion positions in malware samples and corresponding benign code sequences to generate adversarial malware avoiding detection.
How To Spot And Steer Clear Of It
A malvertising campaign named Storm-0408 impacted nearly a million devices, with users downloading malware from GitHub that gave criminals access to login details, OneDrive accounts,
