Researchers at the University of Washington (UW) have successfully tricked Google's Cloud Video Intelligence application programming interface (API) into identifying a video about gorillas as one dealing with an unrelated topic.
The team did this by injecting subliminal images of a pasta bowl into the video, leading the API to conclude the video's subject matter was spaghetti.
The researchers say the test suggests a significant security issue with Google's artificial intelligence. "Specifically, an adversary can insert an image, periodically and at a very low rate, into the video in a way that all the generated shot labels are about the inserted image," the UW team warns. "Such vulnerability seriously undermines the applicability of the API in adversarial environments."
The researchers also note the issue is compounded because adversaries can circumvent a video-filtering system by inserting a harmless image into a video with illegal content.
From Digital Trends
View Full Article
Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA
No entries found