Cancer-Spotting AI and Human Experts Can Be Fooled by Image-Tampering Attacks

Congratulations to Shandong Wu and his team on this great high impact publication.

UPMC published a news release here: https://www.upmc.com/media/news/121421-wu-ai-safety


AI safety paper in Nature Communications PDF link: https://rdcu.be/cDbTM.

And here are several media stories about the study, and there may be more coming up:

https://www.theverge.com/2021/12/14/22831966/hackers-mammogram-ai-cancer-fake-images

https://www.healthcareitnews.com/news/diagnostic-ai-may-be-vulnerable-cyberattacks-upmc-study-shows

https://medicalxpress.com/news/2021-12-cancer-spotting-ai-human-experts-image-tampering.html?deviceType=desktop

https://www.auntminnie.com/index.aspx?sec=ser&sub=def&pag=dis&ItemID=134543

https://www.eurekalert.org/news-releases/937655

https://www.sciencedaily.com/releases/2021/12/211214084541.htm

https://www.beckershospitalreview.com/cybersecurity/diagnostic-algorithms-can-be-fooled-by-cyberattacks-upmc-research-finds.html

https://californianewstimes.com/cancer-spotting-ai-and-human-experts-can-be-fooled-by-image-tampering-attacks/623144/

https://techiai.com/cancer-spotting-ai-and-human-experts-can-be-fooled-by-image-tampering-attacks/