Neural networks have become the driving force behind pioneering advancements in a myriad of real-world applications, including computer vision, autonomous driving, and healthcare and they are vulnerable to integrity attacks, risking consequences such as identity theft, financial losses, and even human safety.

Lakehead University’s Dr. Ehsan Atoofian and his team of researchers are forging a robust defense strategy. By scrutinizing the specific feature-detecting neurons within deep neural networks (DNNs), the team of researchers can identify malicious inputs. In essence, they are fortifying machine learning applications against attacks targeting energy and latency, thus ensuring the continued security and reliability of these technologies.