Adversarial attacks are an increasingly worrisome threat to the performance of artificial intelligence applications.
If an attacker can introduce nearly invisible alterations to image, video, speech, and other data for the purpose of fooling AI-powered classification tools, it will be difficult to trust this otherwise sophisticated technology to do its job effectively.To read this article in full or to leave a comment, please click here(Insider Story)

Leave a Reply