Copilot secret scanning for generic passwords is generally available #142208
Unanswered
novium2
asked this question in
Code Security
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Select Topic Area
General
Body
Copilot secret scanning is now generally available. Copilot secret scanning, which detects generic passwords using AI, offers greater precision for unstructured credentials that can cause security breaches if exposed. Over 350,000 repositories have already enabled this password detection.
To enable Copilot secret scanning, select “Scan for generic secrets” within your code security and analysis settings at the repository level, or the code security global settings at the organization level. You can also use the Update a repository API endpoint for enablement at the repository level. Support for enablement through your organization’s code security configurations, as well as enablement for organizations and enterprises with the API, will come in a future release.
Password detection is backed by the Copilot API and is available for all repositories with a GitHub Advanced Security license. You do not need a Copilot license to enable generic secret detection. Passwords found in git content will create a secret scanning alert in the “Experimental” tab, separate from regular alerts.
In effort to reduce false positives and detections of secrets that are used in tests, Copilot secret scanning will not:
– detect more than 100 passwords per push
– detect secrets in media files (.svg, .png, .jpeg)
– detect secrets in language files (.js, .py, .ts, .java, .cs, or .rb) that contain test, mock, or spec in the filepath
– detect additional secrets in files where five or more alerts have been marked as false positive
Read more here
Beta Was this translation helpful? Give feedback.
All reactions