@ant_av7: Take a look at our latest research on DockerHub secrets exposure! This time, we focused on high-value targets and we fo…
Summary
Binarly's research on DockerHub uncovers widespread exposure of secrets and credentials that could compromise major enterprises, and shows how fine-tuned LLMs can be used for cost-effective automated triage.
View Cached Full Text
Cached at: 09/04/26, 06:39 AM
Take a look at our latest research on DockerHub secrets exposure!
This time, we focused on high-value targets and we found credentials and API keys that could be used to compromise major enterprises and popular open-source projects. We also identified over 200,000 internet-exposed endpoints that used encryption with leaked private keys.
Finally, we demonstrated how fine-tuned LLMs can be used to perform automated triage of generic secret findings in a cost-effective and secure manner.
Blog: https://binarly.io/blog/docker-hub-secrets… Advisories: https://github.com/binarly-io/Vulnerability-REsearch/tree/main/SecretsResearch…
Docker Hub Secrets Expose Critical Credentials | Binarly
Source: https://www.binarly.io/blog/docker-hub-secrets Skip to main content
BINARLY🔬 (@binarly_io): In the era of software supply chain compromises, secrets hiding in public container images can create a direct path to source code, CI/CD, cloud infrastructure, private registries, and internal communications.
We scanned 90,000+ DockerHub namespaces and found 98,816 secret
Similar Articles
@Docker: The malware didn't bring its own credential scanner. It borrowed yours. In this AI Coding Agent Horror Stories issue: @…
A malicious npm package (s1ngularity) exploited post-install hooks to repurpose installed AI coding agents as credential scanners, stealing secrets from developers. Docker's blog discusses how Docker Sandboxes can mitigate such attacks by isolating credentials from agent reach.
@github: Protect your secrets. Join us July 28 for a GitHub Security webinar exploring secret leak trends, public exposure monit…
GitHub is hosting a security webinar on July 28 covering secret leak trends, public exposure monitoring, and practical steps for organizations to reduce risk and respond quickly.
Why 38% of AI Agent container escapes didn't need kernel 0-days: Analysis of 109 empirical incidents (Open Dataset + Defense Harness)
An empirical post-mortem of 109 AI agent security incidents finds that 38% of container breakouts exploited trivial configuration mistakes (e.g., mounted docker.sock, leaked env vars, missing taint tracking) rather than kernel 0-days. The authors release an open dataset and an open-source Multi-Agent Supervisor Security Harness with taint propagation and execution boundary controls.
@charles_irl: Responding to the increase in offensive cybersecurity means improving cybersecurity defense(-in-depth). Before this sto…
Independent security researcher Ben Zimmermann, working with MongoDB's Kingfisher team, scanned 7.25 million public repositories on Gitee, China's largest Git platform, uncovering 26,400 unique live credentials including an active organization-admin key for a major enterprise AI platform and thousands of cloud provider access keys.
@github: Devs and agents are moving faster than ever. Secret protection needs to keep pace. GitHub’s new context-aware classifie…
GitHub introduces a fine-tuned, context-aware secret classifier built with Microsoft Applied Sciences that evaluates candidate secrets in under 2 milliseconds and could more than double the secrets caught by push protection. The essay shares nine quarters of data showing developers aren't becoming more careless — they're being outpaced by AI agents that now appear in one in three pull requests.