Unprecedented Cybersecurity Breakthrough: The Pathfinder Attack
Kazem Taram, an assistant professor of computer science at Purdue University and a UC San Diego computer science PhD graduate, stated that the Pathfinder microarchitectural control-flow extraction attack is the most accurate and powerful tool of its kind. The research involved collaboration with UC San Diego coauthors Dean Tullsen, Hosein Yavarzadeh, Archit Agarwal, and Deian Stefan. Additional coauthors included Christina Garman and Daniel Moghimi from Purdue University, Daniel Genkin from Georgia Tech, and Max Christman and Andrew Kwong from the University of North Carolina Chapel Hill.
This research was supported by several organizations including the Air Force Office of Scientific Research, the Defense Advanced Research Projects Agency, the National Science Foundation, the Alfred P. Sloan Research Fellowship, as well as donations from Intel, Qualcomm, and Cisco. The funding helped to advance the exploration and understanding of microarchitectural control-flow extraction.
The researchers followed responsible disclosure practices by notifying Intel and AMD of their findings in November 2023. Intel subsequently alerted other affected hardware/software vendors. Both Intel and AMD committed to addressing the concerns raised in the paper by issuing a Security Announcement (Intel-SA-2024) and a Security Bulletin (AMD-SB-7015), respectively. The findings were also shared with Vulnerability Information and Coordination Environment (VINCE) under Case VU#157097 which pertains to a class of attack primitives that enable data exposure on high-end Intel CPUs. This collaborative effort highlights the importance of ethical and transparent communication in addressing cybersecurity vulnerabilities.
The Pathfinder attack is unique because it provides an accurate way to extract sensitive information such as passwords or encryption keys from software running on a target machine’s processor without relying on any external input or communication channels. This makes it difficult for defenders to detect or prevent attacks using traditional security measures like firewalls or intrusion detection systems.
The researchers believe that their work will have significant implications for both industry and academia. It could help developers create more secure software by identifying potential vulnerabilities early in the development process
