Fast, Automatic, and Accurate Code-based Attack Attribution through Deep Learning
Dr. Steve Wang
KEY INTERESTS
Network security; Applied cryptography; Computer security; Software engineering
APPOINTMENTS/AFFILIATIONS
Professor, Department of Computer Science, James Madison University
ACADEMIC DEGREES
BS, Information Science, University of Science and Technology of China
MS, Computer Engineering, University of Science and Technology of China
PhD, Information Technology, George Mason University
FAST, AUTOMATIC, AND ACCURATE CODE-BASED ATTACK ATTRIBUTION THROUGH DEEP LEARNING
This project explores efficient process and algorithms to build neural network models to automatically classify, and hence predict, programmer styles of malicious software for quick attack attribution. Human artifacts like computer programs often carry the individual styles of their creators. If retrieved properly and effectively, such style information can be used to categorize the artifacts, compare the relative "distances" among multiple artifacts, and may even be used to trace the authorship of an artifact, including malicious software for attack attribution. For source code-based authorship identification, this research has produced a new deep contrastive learning algorithm, which achieves high accuracy compareable to reported research and has been published in a peer-reviewed AI conference. The methodologies from this research have also been used to trace the authorship of Bitcoin. Our research results debunk popular speculations on Bitcoin authorship and have been published in a security conference.