"This Tiny Chip Can Safeguard User Data While Enabling Efficient Computing on a Smartphone"
Researchers from the Massachusetts Institute of Technology (MIT) and the MIT-IBM Watson AI Lab developed a new chip that can efficiently accelerate Machine Learning (ML) workloads on edge devices such as smartphones while securing sensitive user data against two common types of attacks: side-channel attacks and bus-probing attacks. Health-monitoring apps can be slow and energy-inefficient as the ML models behind them must be shuttled between a smartphone and a central memory server. Engineers try to increase speed using hardware that reduces the need to move a lot of data back and forth. However, these ML accelerators are vulnerable to attacks involving the theft of sensitive information. To address this vulnerability, the team developed an ML accelerator that is resistant to the two most common types of attacks. Their chip can protect a user's health records, financial information, and other sensitive data while allowing large Artificial Intelligence (AI) models to run efficiently. This article continues to discuss the security solution developed for power-hungry AI models.
Submitted by grigby1