How I discovered leaked Snowflake credentials for a Fortune 500 Manufacturing Company using GitHound
In late 2025, I used GitHound to uncover a high severity security issue involving leaked Snowflake tokens and sensitive information on a public GitHub repository. The affected organization, a Fortune 500 manufacturing company, handled the incident quickly and professionally, and later granted us permission to publish an anonymized case study to help educate the broader security community.
By searching GitHub for Snowflake related keywords, I came across a useful one that revealed a large number of Jupter Notebooks with credentials: snowflake "session:role-any". Digging deeper, I discovered that the session:role-any keyword corresponds with Snowflake’s OAuth API , meaning that its occurrence in GitHub repos indicates that those repos are associated with companies using OAuth with Snowflake. This is an example of breadth-first secret hunting, a topic I outlined way back in 2020.
After I found this keyword, I loaded it into GitHound Explore, which monitors GitHub for leaks in real time:
Within minutes, I had a hit.
A data scientist working for a Fortune 500 manufacturing company had accidentally posted 20+ Jupyter Notebooks to a public GitHub repository. One notebook included the search keyword, then GitHound found numerous secrets in the file.
For example:
But that’s not all!
Sifting through the repo’s other files, there were other juicy leaks!
The data scientist had created a separate Jupyter Notebook that accessed the company’s internal Sharepoint file storage system, and printed a snippet of real order information right in the Jupyter Notebook.
I reported this to the company responsible on HackerOne, who quickly validated the report and ensured that any leaked tokens had been rotated. They approved our anonymized public disclosure of this report (for which I am very thankful to the team!).
Data scientists rely on ephemeral cloud tools and quick scripts, which bypasses traditional security controls deployed in software development pipelines. Jupyter Notebooks contain both code and data, making them a valuable place to hunt for leaks. And all of this leads to more leaks, more accidental exposures, and more opportunities for bug bounty hunters and secret monitoring tools.