Speed Kills: Exploring Confused Deputy Attacks Through Edge AI Accelerators
The paper titled 'Speed Kills: Exploring Confused Deputy Attacks Through Edge AI Accelerators' investigates security vulnerabilities in AI Accelerators (AIA). It reveals that confused deputy attacks are feasible on six out of seven popular AIAs, potentially affecting over 100 million devices. The authors propose a validation defense mechanism that incurs minimal runtime overhead.
- ▪The study focuses on confused deputy attacks using AI Accelerators (AIA).
- ▪Six out of seven AIAs were found to be vulnerable, impacting over 128 System On Chips (SOCs).
- ▪The proposed defense mechanism shows minimal runtime overhead of approximately 15%.
- ▪The findings have been acknowledged by the corresponding vendors and assigned the CVE-2025-66425.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Cryptography and Security arXiv:2605.17707 (cs) [Submitted on 18 May 2026] Title:Speed Kills: Exploring Confused Deputy Attacks Through Edge AI Accelerators Authors:Datta Manikanta Sri Hari Danduri, Aravind Kumar Machiry View a PDF of the paper titled Speed Kills: Exploring Confused Deputy Attacks Through Edge AI Accelerators, by Datta Manikanta Sri Hari Danduri and 1 other authors View PDF Abstract:AI Accelerator (AIA) are specialized hardware e.g., Tensor Processing Unit (TPU), that enable optimal and efficient execution of AI applications and on-device inference. The growing demand for AI applications has led to the widespread adoption of AIAs on Edge or embedded devices on Edge or embedded devices.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv.org.
Discussion
0 commentsMore from arXiv.org
