The authors of the paper from North Carolina State University explain that the theft of AI models poses a serious threat because such models require significant computing resources to create and their leakage makes them vulnerable to attacks. This not only violates intellectual property rights, but can also reveal sensitive data embedded in the model’s behavior.
The study used a commercially available Google Edge TPU chip designed to run AI models on end-user devices. The scientists used a method to monitor electromagnetic signals by embedding a sensor in a TPU chip that provides real-time data about the AI model’s behavior.
Using this data, the researchers were able to reproduce the architecture of the model and even reconstruct it with 99.91% accuracy. This highlights the vulnerabilities of modern AI systems running across devices and forces developers to create effective security measures.
Now, after demonstrating the vulnerability, scientists plan to develop countermeasures to protect AI models from such attacks.
Source: Ferra
I am a professional journalist and content creator with extensive experience writing for news websites. I currently work as an author at Gadget Onus, where I specialize in covering hot news topics. My written pieces have been published on some of the biggest media outlets around the world, including The Guardian and BBC News.
