As part of AI Sweden’s Edge Learning Consortium, and together with Dakota State University, the AI Security Lab will aim to provide developers and users of Edge Learning Infrastructure the understanding to develop robust and secure AI solutions.

By utilizing edge devices for training AI, edge learning opens up a number of new possible surfaces for adversarial attacks. To reduce the risks and impact of attacks on edge learning systems, it is important to understand what vulnerabilities exist.

The AI Security Lab will involve security researchers in the development of edge learning solutions. They can help provide the understanding to develop robust solutions from the start for developers and users of edge learning infrastructure.

Challenges

• What new vulnerabilities could exist in an edge learning environment?
• What would be preferred attack surfaces in an edge learning environment?
• What types of threats should we be aware of? (e.g., stealing information, sabotaging the model, hijacking the system, etc.)
• What are strategies and technology to detect attacks?
• What are the strategies and technology to prevent attacks?
• What are the strategies and technology to thwart an ongoing attack?

The AI Security Lab is part of the Edge Learning Consortium.

Madison Cyber Labs at Dakota State University (DSU) and AI Sweden are the key sponsors of this lab. The first activity will be a student exchange program that will start this summer where six students (3 American, and 3 Swedish) will spend one month at AI Sweden and one month at DSU to explore cyber security issues in an edge learning environment. Read more about the exchange program.

Want to know more?

Please contact Beatrice Comoli 

beatrice.comoli@ai.se

+46 (0)70 146 09 64