A Hybrid Reinforcement Learning Framework for Dynamic Resource Allocation in Malware Analysis Systems

A Hybrid Reinforcement Learning Framework for Dynamic Resource Allocation in Malware Analysis Systems

연구 분야: Analysis

학회: 2024 International Conference on Intelligent Algorithms for Computational Intelligence Systems (IACIS)

The emerging strains of malware demand adaptive cybersecurity systems with the ability to shift resourcefully to performance versus good detection – with optimal resource use, especially as threats quickly evolve. This paper presents a policy using reinforcement learning (RL) for dynamic resource allocation in malware analysis: an RL agent observes files and assigns computer resources based on their potential level of threat. The agent learns by trial and error to give precedence to sample examination while maximizing the usage of resources; the study also describes RL environment design, the structure of the agent, and the reward system. The model has shown good accuracy at different stages which include 96.1% for training, 92.2% for testing, and 90.5% for validation data against previous methods in fast-changing threats circumstances. This continuous learning ability of the model allows it to be effective in the light of new threats, which has helped develop resource-aware, scalable, and adaptable malware analysis systems against changing risks in cybersecurity. This model provides benefits and some of the main challenges in this domain for dynamic malware analysis and resource allocation.