Red team exercises are meant to provide comprehensive assessments that reflect real-world conditions and demonstrate tangible risks posed by a malicious actor during a legitimate attack. The results of these exercises are intended to be used to improve security awareness and training and to assess levels of security control effectiveness. Please see more about the principles and vision of our red team and the security team as a whole.
We choose to operate under the following rules of engagement to provide the most benefit to the organization with as little ambiguity as possible for each engagement. In short, the intent is to give guidance to GitLab organization as a whole as to how engagements are defined, communicated, and executed, as well as how the resulting information is disseminated to aid the organization.
In general, a scope definition for an engagement should be exclusive vs. inclusive in nature. It should be scenario driven, with specific goals, based on real security threats. Those scenarios can be elaborated with to the input of other security teams, such as the Threat Intelligence team. They should be recorded ahead of the engagement, validated by the Director of Security, and should include the following details:
A tangible artifact of authorization from someone with the authority to give it should exist prior to the opening of the engagement. This means getting permission in writing from GitLab representatives when systems or teams they manage may be impacted. We obtain the proper permission from any third party providers such as Amazon Web Services, Google Cloud, etc. should those systems be relevant. These artifacts can be either persistent or transitive and should be stored securely should retrieval of evidence become necessary.
The process for obtaining the necessary permission can be different per engagement and service provider and will require due diligence. For example, AWS has its own unique set of guidelines for performing assessments that can be found here: https://aws.amazon.com/security/penetration-testing/. In contrast, Google Cloud's policy is completely different from the evidence illustrated here: https://support.google.com/cloud/answer/6262505?hl=en.
As security professionals, we aim to be ethical in every engagement while maintaining the spirit of mimicking real-world attacks in the wild. We respect the privacy of the employees at GitLab and follow the guidelines mentioned in the Employee Information Privacy section of the Code of Business Conduct & Ethics page of the handbook during our engagements. As a GitLab employee, it is expressly important to review the statements in this documentation regarding privacy of personal information and personal devices used for GitLab business.
The Red Team is an extension of the Security Team as a whole. We actively collaborate with other members of the Security Team to improve the security posture of the organization. We take every reasonable opportunity to share knowledge before, during, and after an engagement. Some engagements, specifically those intended to test the organization's response to a security event without prior warning, are of course communicated differently. It is expected that the red team will need to maintain a certain separation from the organization it is protecting in order to maintain the proper scope, independence, and authority to act as an actual attacker and produce effective results and aid the organization.
As we are a results driven organization, engagement results are properly and securely documented, communicated, and actionable with proper respect given to the sensitivity of information. In order to make results actionable, we collaborate with our teammates in ours and other departments to track remediation through existing organizational processes, making issues confidential where necessary. Specifically, we follow the functional and operational escalation points mentioned in the previous link for tracking remediation.
Critical vulnerabilities and exploits found during an engagement that are currently or have immediate propensity to impact normal GitLab business operations should engage the security on-call and follow our incident response guide. Sensitive information and communicated by creating a confidential security issue using existing Security Issue Triage process in the appropriate project.
All red team testing and engagements are first carefully planned in advance and will not intentionally impact production system or customers. However, accidents do happen and in the case of a production system being impacted during an engagement, either directly or indirectly, we escalate the incident by notifying the red team manager. We also engage our security operations team using our incident response guide at the earliest reasonable time giving full disclosure of what caused the issue. Depending on the circumstances, the infrastructure team may need to be made aware through an incident report to negate or reduce the impact to our customers. Proper root cause analysis is recorded following resolution of the incident.