AI-Powered Malware Deobfuscator

This paper presents the AI-Powered Malware Deobfuscator which is a highly sophisticated and fully containerized security platform that is capable of detecting hybrid threats. The three key security processes are operating in coordination to make the system work, with the first being a static machine learning (ML) pipeline, which is a regression - based Scikit-learn random forest model that classifies Windows PE files based on extracted features using the pefile library; the first is distinguished as having helpful and understandable feature priority rankings; the third is a specific browser extension which blocks web threats dynamically by enforcing a real-time domain blocklist; and the second is non-ML, deterministic signature path, which critically uses the EICAR test string to test the operational integrity The architecture of the whole solution is highly reproducible, with each component coordinated easily by Docker Compose, and an application written in React so the user interface and scan status display are modern, and a back-end written with Fast API so that it can provide every necessary service in a structured, self-sufficient, and reliable way. Notably, zero-vector fallback mechanism to gracefully deal with non-PE and other artifacts out-of-domain ensures that the system is robust. The research provides a comprehensive pedagogical framework including the technical blueprint, engineering approaches, and threat model. It is supported by a rigorous examination procedure comprising of end-to-end functional assessment and offline measures. The conclusion of the work provides a future development roadmap where more advanced functions such as dynamic behavioral analysis and multi-modal content inspection are taken into consideration.