Sign up to save tools and stay up to date with the latest in AI
bg
bg
1

When Machine Learning Tells the Wrong Story

Nov 09, 2024 - jackcook.com
The article discusses the author's research on hardware security, specifically a machine-learning-assisted side-channel attack, which he presented at the ISCA conference in 2022. The research, which won several awards, demonstrated for the first time that system interrupts can leak information about user activity. The author also discusses his attempts to write a blog post about the research, but struggled due to the complexity of the topic and the personal significance of the research.

The author provides a detailed explanation of side-channel attacks and website fingerprinting, using examples and interactive demos. He also discusses his final project at MIT, where he replicated a study by Shusterman et al. on website fingerprinting, and found that he could reliably classify up to 100 websites. The author concludes by reflecting on the impact of the research on his academic and professional trajectory.

Key takeaways:

  • The author presented a research paper at ISCA on a machine-learning-assisted side-channel attack, which won awards and recognition, including first place in Intel’s 2024 Hardware Security Academic Award.
  • The paper demonstrated for the first time that system interrupts, a low-level mechanism that operating systems use to interact with hardware devices, can leak information about user activity.
  • The research paper also highlighted the dangers of applying machine learning toward hardware security research, a lesson that the author believes is widely applicable, even outside of hardware security research.
  • The author also discussed the concept of website fingerprinting, where one website attempts to identify the website open in another tab by taking advantage of a widely-studied side channel: the CPU cache.
View Full Article

Comments (0)

Be the first to comment!