Securing machine learning models to be hacker-proof will be the next frontier in cybersecurity.https://www.wired.com/story/researcher-fooled-a-google-ai-into-thinking-a-rifle-was-a-helicopter/ …
-
To your point this problem applies broadly. Are all ML vulnerabilities equally concerning to you, or are there some that concern you the most? (Within current foreseeable anyway.)
In general, I think cybersecurity is increasingly among the greatest threats facing society. The risk with ML and AI is that we don’t even know how to detect or defend against vulnerabilities in ML models. I’m most worried about cyber hacking impact in the medical domain.
