Responsible detection of generic secrets with Copilot secret scanning

About detecção de segredo genérico with Verificação de segredos do Copilot

Verificação de segredos do Copilot's detecção de segredo genérico is an AI-powered expansion of secret scanning that identifies unstructured secrets (passwords) in your source code and then generates an alert.

GitHub Secret Protection users can already receive alertas de verificação de segredo for partner or custom patterns found in their source code, but unstructured secrets are not easily discoverable. Verificação de segredos do Copilot uses large language models (LLMs) to identify this type of secret.

When a password is detected, an alert is displayed in the "Generic" list of secret scanning alerts (under the Security tab of the repository, organization, or enterprise), so that maintainers and security managers can review the alert and, where necessary, remove the credential or implement a fix.

Para usuários com GitHub Enterprise Cloud, um proprietário da empresa deve primeiro definir uma política no nível empresarial que controla se o detecção de segredo genérico pode ser habilitado e desabilitado em repositórios de uma organização. Por padrão, essa política é definida como "permitida". The feature must then be enabled for repositories and organizations.

Input processing

Input is limited to text (typically code) that a user has checked into a repository. The system provides this text to the LLM along with a meta prompt asking the LLM to find passwords within the scope of the input. The user does not interact with the LLM directly.

The system scans for passwords using the LLM. No additional data is collected by the system, other than what is already collected by the existing secret scanning feature.

Output and display

The LLM scans for strings that resemble passwords and verifies that the identified strings included in the response actually exist in the input.

These detected strings are surfaced as alerts on the secret scanning alerts page, but they are displayed in an additional list that is separate from regular alertas de verificação de segredo. The intent is that this separate list is triaged with more scrutiny to verify the validity of the findings. Each alert notes that it was detected using AI. For information on how to view alerts for generic secrets, see Exibindo e filtrando alertas da varredura de segredos.

Improving the performance of detecção de segredo genérico

To improve the performance of detecção de segredo genérico, we recommend closing false positive alerts appropriately.

Verify the accuracy of alerts and close as appropriate

Since Verificação de segredos do Copilot's detecção de segredo genérico may generate more false positives than the existing secret scanning feature for partner patterns, it's important that you review the accuracy of these alerts. When you verify an alert to be a false positive, be sure to close the alert and mark the reason as "False positive" in the GitHub UI. The GitHub development team will use information on false positive volume and detection locations to improve the model. GitHub does not have access to the secret literals themselves.

Limitations of detecção de segredo genérico

When using Verificação de segredos do Copilot's detecção de segredo genérico, you should consider the following limitations.

Limited scope

Detecção de segredo genérico currently only looks for instances of passwords in git content. The feature does not look for other types of generic secrets, and it does not look for secrets in non-git content, such as GitHub Issues.

Potential for false positive alerts

Detecção de segredo genérico may generate more false positive alerts when compared to the existing secret scanning feature (which detects partner patterns, and which has a very low false positive rate). To mitigate this excess noise, alerts are grouped in a separate list from partner pattern alerts, and security managers and maintainers should triage each alert to verify its accuracy.

Potential for incomplete reporting

Detecção de segredo genérico may miss instances of credentials checked into a repository. The LLM will improve over time. You retain ultimate responsibility for ensuring the security of your code.

Limitations by design

Detecção de segredo genérico has the following limitations by design:

• Verificação de segredos do Copilot will not detect secrets that are obviously fake or test passwords, or passwords with low entropy.
• Verificação de segredos do Copilot will only detect a maximum of 100 passwords per push.
• If five or more detected secrets within a single file are marked as false positive, Verificação de segredos do Copilot will stop generating new alerts for that file.
• Verificação de segredos do Copilot does not detect secrets in generated or vendored files.
• Verificação de segredos do Copilot does not detect secrets in encrypted files.
• Verificação de segredos do Copilot does not detect secrets in file types: SVG, PNG, JPEG, CSV, TXT, SQL, or ITEM.
• Verificação de segredos do Copilot does not detect secrets in test code. Verificação de segredos do Copilot skips detections when both conditions are met:
  • The file path contains "test", "mock", or "spec", AND
  • The file extension is .cs, .go, .java, .js, .kt, .php, .py, .rb, .scala, .swift, or .ts.

Evaluation of detecção de segredo genérico

Detecção de segredo genérico has been subject to Responsible AI Red Teaming and GitHub will continue to monitor the efficacy and safety of the feature over time.

Next steps

• Enabling Copilot secret scanning's generic secret detection
• Gerenciar alertas da verificação de segredo

Further reading

• Sobre a verificação de segredo