As noted by Bleeping Computer, attackers are rapidly exploiting credentials and access secrets exposed in public Docker Hub images, with some instances occurring in less than eight seconds after publication.
Researchers Billy Tsouvalas and Nick Nikiforakis from Stony Brook University conducted a four-month experiment using 300 honeypot images across eight container registries. By planting AWS, OpenAI and Anthropic API keys, they observed that automated scanners, which they termed ContainerBots, frequently extracted these secrets through static analysis of image layers rather than by running the containers. The study revealed that all 100 AWS keys were compromised within 15 minutes, with the fastest exploitation occurring in 7.86 seconds. The researchers noted that these scanners are not necessarily coordinated but operate continuously to harvest sensitive data. Despite reporting these findings to Docker in May, the researchers observed no significant security improvements on the platform by August. The findings highlight a critical vulnerability in software supply chains, as leaked keys often grant unauthorized access to corporate accounts and cloud resources. Experts recommend that registry operators implement automated secret scanning and rate-limiting to prevent mass-scanning activities that threaten developers and organizations relying on public container images.
Source: Bleeping Computer

