Adaptive Malware Detection using Reinforcement Learning

Abstract

The rapid growth of zero-day malware attacks has resulted in the reduction of the effectiveness of traditional malware detection tools, which normally use signature-based or rule-based malware detection techniques. Traditional malware detection techniques have been ineffective in detecting new malware attacks or zero-day attacks before they occur or even after they occur if they have been encoded or obfuscated. As a result of these challenges, there is a need for an innovative solution for malware detection. In this work, an Adaptive Malware Detection System is proposed using machine learning with reinforcement learning for effective malware detection. The proposed system uses a random forest classifier for intelligent decision-making based on the analysis of data collected from the feature-based data. In addition, reinforcement learning is used for decision-making using a Q-learning algorithm for effective decision-making. The proposed system is capable of adapting to new malware attacks using reinforcement learning, which learns from past experiences. Moreover, an interactive visualization module has been added to the proposed system for better decision-making based on the analysis of decision patterns and reward trends. The proposed solution is effective for use in the latest malware attacks.