By leveraging diverse technologies and an automation first approach, the Security Automation Team strives towards improving efficiency, effectiveness, and accuracy within GitLab's Information Security program with a focus on cost savings.
Build security tooling and automation for internal use that enable the Security Department to operate at high speed and wide scale. Additionally, Security Automation will assist other security specialty teams with automation efforts they are leading and developing through the assessment of automation tools and technologies as needed.
The Security Automation team is part of the Security Engineering & Research sub-department. See GitLab's organizational chart and meet our team members.
If you have a need or idea for security-relevant projects at GitLab that require automation, please add the
secauto|workflow::new label directly to your issue as detailed in the
Labels section. You can also create an issue by clicking here. On the other hand, to bring an issue or artifact to our attention without expectation of a deliverable, please add the
secauto|interest label to it. We review and prioritize issues with those labels during our Weekly Meeting, which anyone can attend. In case you want to engage us directly, please reach out on the
#security-automation Slack channel. As usual, you can also tag the SecAuto team within GitLab by using
The SecAuto team continually works on improving the way in which we work and deliver value to the company. This section describes our workflow and how we currently use GitLab to manage our team and projects.
Our workflow is a simplified version of GitLab's Engineering Workflow. Although, we might increasingly adopt conventions and approaches discussed there, our process will most likely perpetually deviate with bias towards simplicity.
The best way to get an overview of what the SecAuto team is working on at any time is to use our epics list and issue boards.
SecAuto's OKRs are team-centric and tackled as a team effort, only issues are assigned directly to DRIs. For all epics tracking SecAuto work and OKRs, we follow the
FY##Q# OKR - NAME => 0%. Finally, OKR-epics must be labeled
Security Management::Security Automation Team and be ideally created at the gl-security level. This must be the case, since Security Leadership relies on the epics listing at that level to follow OKR progress. Although epics do bubble up the hierarchy and they would be visible under https://gitlab.com/groups/gitlab-com/gl-security/-/epics/, linking issues further up an epic's hierarchy is not possible, which can problematic for cross-functional efforts in which SecAuto is often involved.
SecAuto's KRs are also mostly team-centric and to be tackled as a team effort. For all epics or issues tracking SecAuto's KRs we follow the
FY##Q# KR - NAME => 0%. KR-epics and issues must be labeled
Security Management::Security Automation Team and idealy be created in SecAuto's Main Group Hierarchy.
We have three issue boards: the Intake Board to keep track of progress on issues, the Management Board which gives the team an overview of our OKRs as well as matters exclusive to the team's management and the In-Progress Board which tracks what the team is working on during any given (#milestones).
secauto|* labels currently being used by the SecAuto team can be found in the issue label listing for gitlab-com.
SecAuto has a Main Group Hierarchy containing most of our projects and a Main Tracker. However, we also act as maintainers or owners in other namespaces, please refer to the project list section below.
Any issues and associated work concerning the SecAuto team that cannot be clearly assigned to a single, existing SecAuto product and its associated repository.
Issues that are exclusive to a single product or project and for which a repository already exists, these shall be labeled with the appropriate
secauto| labels (workflow, task, priority and customer). If a repository doesn't exist but the issue to be created is relevant only to a single project, component or product, the creator should consider setting up a repository.
Most of the work SecAuto does is externally initiated. Opportunities for contributing originate from the day-to-day operations of teams in the Security Department and accross GitLab. These opportunities usually arise in Slack conversations, as issues in the GitLab groups
gitlab-com, etc. SecAuto's processes rely heavily on labels, they inform both customers and the team about the status of issues and other artifacts as detailed in the
SecAuto's intake process is simple, as detailed in the
Once SecAuto creates a task or is made aware of one relevant to the team, the team must assign the label
secauto|workflow::new. All tasks labeled
secauto|workflow::neware considered part of the SecAuto backlog and must be triaged.
After this has happened, the team will triage the issue, as is done with any issue labeled
Alternatively, the SecAuto team can be made aware of an issue where their indirect contribution or involvement may be desired by applying the
secauto|interest label on an artifact.
secauto|interestcan be used to mark issues and other artifacts across the company's namespaces that might be of interest to secauto. An issue detailing new guidelines and expectations regarding cost management in GCP environments would be one such example.
Tasks that are considered too large for a single iteration will be upgraded to epics in our OKR Epics List or our KR Epics List. From there, additional sub-tasks will be created to define iterative and more manageable work packages in an attempt to better iterate on the epic.
Everyone in the team is responsible for maintaining consistency in workflow and labeling. We're a small team and the more organized we are the better we'll be able to collaborate and communicate our results. New issues in the Main SecAuto tracker will be collectively reviewed and labeled before or during the weekly SecAuto meeting. New issues in project-specific trackers will be reviewed and properly labeled by each project's DRI(s) upon creation, at a minimum they should label them
secauto|workflow::new as they exist outside of the Main SecAuto hierarchy and would otherwise not be visible to the week's triager.
secauto|task::management is meant for all issues and conversations related to the strategic, tactical and managerial aspects affecting the Security Automation team and its projects.
secauto|interest can be used to mark issues and other artifacts across the company's namespaces that might be of interest to SecAuto. An issue detailing new guidelines and expectations regarding cost management in GCP environments would be one such example.
Tasks, issues and other artifacts on which SecAuto is expected to deliver can be in the
Once SecAuto or another team creates a task for the team to triage, the label
secauto|workflow::new must be applied to it. All tasks labeled
secauto|workflow::new are considered part of the SecAuto backlog and must be triaged.
The next step is for tasks to be triaged and thus be labeled
secauto|workflow::ready so they can be worked on by the team. A task will only be labeled
secauto|workflow::ready once the following holds:
secauto|customerlabel identifiying the customer(s) benefiting from the deliverables.
Once a task is labeled
secauto|workflow::ready, work on it can begin. After work has begun, it will be labeled
secauto|workflow::in-progress and once it's concluded
Such is the ideal, more common path for a task to take.
secauto|workflow::cancelled labels can be applied at any time in the life-cycle of a task and are mostly used to track how often we are blocked, how many times we triage, prioritize and work on something only for us or our customers to halt the task, etc.
Every task must have at least one
secauto|customer* label, these labels signal what teams are requesting and will benefit from the delivery on a given task. Furthermore, it can be assumed that the author of the issue, the members of their team participating in the issue, and if needed their manager, can act as stable counterparts during the time period the SecAuto team works on said task.
If there are multiple customer teams, each team should have at least one stable counterpart explicitly listed on the issue.
All work performed by SecAuto falls within the following areas:
secauto|task::managementrelates to managerial activities affecting the SecAuto team
secauto|task::incidentreactively mitigates a critical malfunction on product by SecAuto
secauto|task::bugreactively addresses a non-critical, undesired condition in a product by SecAuto
secauto|task::improvementproactively delivers a non-critical related to a product by SecAuto
secauto|task::maintenanceproactively delivers an operative related to a product by SecAuto
secauto|task::infrastructureproactively delivers value related to a non-product, non-user-facing infrastructure component used by SecAuto, this includes scripts for cleanup, GCP infrastructure.
secauto|task::analysisdelivers information after performing evaluations, analyses, investigations, brainstorming and the like.
secauto|task::documentationformalizes and documents the results of other task types whenever this is required.
Infrastructure, maintenance and improvement labels exist since the nature of the work is different. As SecAuto, we work with more than features, in our day-to-day we act as developers, architects, consultants, at times even SREs. These labels help capture the nature of that work for strategic and tactical purposes such as metrics gathering.
secauto|priority::high are our priority markers. We use this to determine what needs to be done and when.
We assign priority labels based on a simplified version of the RICE framework in use by the Engineering Department and define Impact and Effort as follows:
|Low||a deliverable will somewhat improve the ability of the customer to act efficiently and effectively|
|Moderate||a deliverable will improve the ability of the customer to act efficiently and effectively|
|High||a deliverable will considerably improve the ability of the customer to act efficiently and effectively|
|Low||producing a deliverable will take one SecAuto team member less than a week|
|Moderate||producing a deliverable will take one SecAuto team member less than a month|
|High||producing a deliverable will take one SecAuto team member less than a quarter|
Taking this into consideration, we calculate priority as follows:
These priority calculations assume a good understading of the customers needs, also known as confidence. In cases where a task is not clearly defined or its purpose unclear, a
secauto|task::investigation issue should be created instead.
secauto|needsReviewTeam must be looked at by the respective party mentioned in the label in order for work to proceed. For example, the creation of a new, expensive GCP environment should be labeled
secauto|needsReviewManager anytime a manager's input is needed. Furthermore, work potentially leading to down-time on a product used by a SecAuto customer should be labeled
secauto|needsReviewCustomer. Finally, for non-code-managed changes where approval cannot happen in an MR or where peer-review is desired,
secauto|needsReviewTeam should be used.
The SecAuto team's weekly meeting is a 50-minute meeting that can also be found on the Security Department Team Meetings calendar.
This process relies on a rotating role of "SecAuto triager" to ensure that issues that do not correspond to project with an existing DRI are triaged asynchronously (see below).
Completing the following asynchronous tasks prior to the meeting will keep the synchronous meeting on schedule:
secauto|workflow::newissues they come accross (the triager, mainly those in the Main SecAuto Tracker, and DRIs, those relevant to the respective projects in which they are involved):
secauto|customerlabels identifying the customers requesting the deliverables.
secauto|priority::high. If so, they are expected to engage the most-likely DRI or raise with the team to find a DRI that will work on the task. This should be done immediately or, the latest, at the upcoming SecAuto Team Meeting. For other issues, these will remain in the
secauto|workflow::newstage as a backlog items. The triager should note this in a comment to provide feedback for the requester on expected time frames.
secauto|priority::high, the Triager should make an initial attempt at scoping the work for an iteration. They or the DRI should, if this can be done without the teams consideration, break down the task in smaller issues prior to the meeting that can be scheduled and assigned for delivery during the next and following weeks.
secauto|interest, the SecAuto Triager should skim such issues. They then, if relevant, raise these with the team during refinement part of the weekly meeting or list them as interesting reading in the meeting document. Once reviewed, the
secauto|interestqueue should be emptied after the meeting.
The "SecAuto triager" is responsible for facilitating the refinement portion of the meeting:
secauto|workflow:blockedas well as raise blockers and review requests for any unfinished work.
secauto|priority::hightasks that are still marked
secauto|workflow::readyand sort them with the most pressing at the top of the queue.
secauto|workflow:newissues with high, moderate and low priority, in that order, are discussed and assigned to team members for them to triage them and bring them to a
secauto|workflow:done, determine which of the outstanding issues will be carried-over to the next milestone or dropped from it and move them accordingly.
The nature of our workflow is iterative and incremental as planning, analysis, design, implementation, deployment and maintenance happen repeatedly and progress is tracked over time.
This is an informal stage that usually happens in 1:1s, team meetings or over Slack. We strive towards making this process productive by focusing on four key aspects:
We expect most, if not all, new deliverables related to a new product by SecAuto to start with a
These issues can then be closed or promoted to an epic or have other tasks branch out from them.
It is important to ensure that intake issues are sized and scoped properly to avoid having a scenario where an issue stays open stagnant for a long time or eventually stagnates. This helps offer value, prevent over-engineering and make it easier to review and provide feedback.
These are the main tenets of design that we consider to be best practices:
In order to ensure overall efficiency and effectiveness in creation and building of any SecAuto project, it is necessary to have a well-defined process that illustrates and provides guidelines on how to get from ideation to implementation the quickest.
As part of our collective project management process it's important to continuously steer the team's efforts, thereby increasing our chances of attaining a timely, high-quality delivery. Continuous collaboration and iteration is hence encouraged amongst stakeholders so as to ensure that customer expectations are met.
A good project management process also allows the SecAuto team to have a strong change control system where documentation (through MRs, issues, 1:1s and meetings) on changes to a project is updated and approved accordingly while keeping stakeholders informed.
We've come to understand that a roughly shared structure across the different SecAuto projects is a net positive and allows for efforts to cross-pollinate, and reduces the effort required for team members to get started or involved. Having a scaffold also guarantees compliance to architectural and procedural requirements for developing tools. Thus, strive to include the following items within a project:
CODEOWNERS(by default the SecAuto team)
Feel free to clone this basic project scaffold template to get you started on building out a SecAuto project.
For any given project it is essential to standardise its git and branch management workflow. In the initial stages of an MVP, when there is a single developer working on the project and not much is in production, it's fine to commit and deploy from master. However, as soon as a project has it's first production release or the project team grows to more than 1, the team should transition to branch and MR-based or stable-master workflows and protect master, only allowing deployment from said branch.
For projects that will be versioned, we will create a release tag or branch when the project is to be deployed to production.
We should strive towards ease of deployment, portability and reproducibility regardless of environment by adhering to these guidelines:
Documentation for projects will depend on their eventual use and ownership. If a team other than SecAuto is going to own the project after it's released, more documentation is usually encouraged.
If the project is going to be internal to SecAuto, a more iterative approach to documentation will suffice.
In general, the following documentation should be provided in either case:
GitLab CI and Kubernetes/Knative are the most common environments for running our automation, with Python and Go being the preferred languages.
Kubernetes/Knative services are built as Docker images by GitLab CI (naturally) in their code project. Deployment to GKE clusters is performed by separate deployment projects containing the service configuration. Monitoring and alerting is configured using Terraform.
|Google Workspace Group Members Reporter|
|Suricata Runner Metrics|