The aim of this document is to define the key steps required as part of responding to a security incident, in line with Sourcegraph’s Security Incident Response Policy.
When triaging a security event, follow the documented security event response process.
If you believe that the event report is a valid incident, then formally raise an incident by opening an incident tracking document in the ‘Security Incidents’ Google Drive, based off the template provided here. Access to this document should initially be tightly limited to those who are involved in investigating the incident, until the full scope of the intrusion is known. Assume that the attacker has visibility of your actions in public channels. Keep in mind that some incidents may also involve sensitive information about employees, and treat this data carefully.
You will need to choose a severity for the incident – use the guidelines from the Security Incident Response Policy to guide you. The severity of the incident will determine escalation and reporting requirements.
Once the incident has been declared, the incident manager must coordinate the work required to:
- Identify exactly what systems and data have been compromised as part of the incident, using available audit data
- Identify the steps required to contain the intruder (such as isolating the network that the attacker has gained access to) and carry them out
- Remove the intruder’s access to Sourcegraph’s system and data
The incident manager is not expected to carry out all of this work themselves – they should bring in relevant team members as required.
For communication, a Zoom call should be created with the members of the team involved in dealing with the incident. All tracking of progress should be done via the incident tracking document.
Aside from the team working on investigation and containment, other members of Sourcegraph’s team might also need to be involved. In particular, these are:
- For all S2 incidents and above, the Security Lead must immediately be notified.
- For all S2 incidents and above, the Legal team must immediately be notified.
- For all S1 incidents, the VP of Engineering must be notified.
Clients must be notified if Sourcegraph has any reason to believe that their data has been impacted during an incident. Clients should be notified:
- Within 24 hours of discovering that their data has been impacted to tell them as much
- Within 72 hours of the first notification with a detailed report on exactly what data has been compromised according to known information at the time.
When the full impact of the incident is known, clients should also be provided with a full report on exactly what data of theirs has been compromised.
In the process of investigation, the team responding to the incident must collect any evidence that they think will be necessary for future investigation. Any information that helps to trace the course which an attacker has taken through Sourcegraph’s systems is potentially useful.
When convenient, this information should be stored in the Security Incidents Google Drive, in a folder titled with the incident ID. Google Drive provides version history, ensuring that any changes to evidence can be tracked and attributed to their authors.
Once the attacker has been removed from the system, the next priority should be to recover the systems affected. Work with the Engineering and Operations teams as required to get systems back up and running.
For systems that have been compromised, ensure that you restore systems to the last known good state – that is, the last point at which we are confident the systems had not been compromised.
For secrets such as service account keys and passwords, ensure that these are rotated wherever there is reason to believe that they might have been compromised.
It is the security team’s job at this point to provide the other Engineering and Operations teams with the information that they need to return the business to its normal operating posture.
Once the attacker has been removed from the system and the immediate recovery and remediation steps have taken place, the incident manager is expected to coordinate the completion and publication of a postmortem report for all S1 and S2 severity incidents.
A template for the report can be found here. The postmortem must cover:
- The root cause of the incident
- A summary of all data and systems impacted
- Information on the attacker, where available
- Steps that could be taken by the security team to reduce the chance of this incident reoccuring
- Steps that could be taken by the wider organization to reduce the change of this incident reoccurring
All action items identified must be tagged with an owner, and the owner must be made aware of the action item.
If it is deemed safe, feel free to publish and circulate the incident tracking document and postmortem across the organisation.
In some cases it may be necessary to conduct detailed analysis of the evidence collected during the incident. If there are the necessary skills to do this within the team, then this analysis can proceed internally. Otherwise, the team should engage external consultants to carry out this work. Consult with the Security Lead if you are unclear about what steps are required here.