Definition of DevSecOps: security written into the DevOps approach
DevSecOps adds security to the development and operations pairing of DevOps. The term refers to security controls built into the pipeline that builds, tests and deploys software, and handled by the teams that deliver. As with DevOps, the definition comes from usage.
Freely available publications describe how to implement it. Publication NIST SP 800-204C (March 2022) presents DevSecOps as a paradigm suited to cloud-native applications, based on continuous integration, delivery and deployment pipelines. The OWASP DevSecOps Guideline project sets it a goal: to detect security issues, design flaws as well as vulnerabilities, as early and as cheaply as possible, and then to keep detecting them.
A final check cannot keep up with the delivery pace
When an application is delivered a few times a year, a security audit before each production release remains practicable. When it is delivered every week, this final check risks becoming a bottleneck, or a formality that people get around.
DevSecOps moves verification to each change and automates what can be automated. OWASP puts it this way: to keep pace with continuous integration and deployment, security must be introduced early, from the writing and testing of code.
The controls DevSecOps adds to the pipeline
The OWASP guide lists the steps to introduce into a basic pipeline, among which:
- scanning for secrets (credentials, keys) left in code repositories;
- software composition analysis (SCA) and static code analysis (SAST);
- infrastructure code scanning, to spot configuration errors;
- software supply chain security: component inventory (SBOM), signing and provenance of artefacts;
- dynamic analysis of the running application (DAST);
- compliance checks.
The same guide points out that integration and deployment tools often widen the attack surface themselves: the pipeline is protected like an asset. What each analysis detects and lets through is detailed in our article on application security.
Tooling does not do everything. The approach requires rules written in advance (which severity blocks a delivery, within what time a vulnerability is fixed, who grants an exception) and development teams that handle the results themselves. To position its practices, an organisation has two public frameworks: the OWASP SAMM maturity model, and the NIST SSDF framework (SP 800-218), which groups its practices into four families: prepare the organisation, protect the software, produce well-secured software, respond to vulnerabilities.
DevSecOps, application security, shift-left: not to be confused
Application security is the discipline: the set of practices that protect software. DevSecOps is a way of organising it, within the delivery flow rather than alongside it.
Shift-left consists of moving controls towards the start of the cycle. It is a principle of DevSecOps, not a synonym: OWASP also speaks of shift-everywhere, for detection that continues throughout the life cycle.
DevSecOps is neither a tool purchase nor the end of human checks. Design review and penetration testing keep their place, for what an automated analysis does not see.
Illustration: a vulnerable dependency stopped before the merge
Take a team developing an appointment booking portal. A developer adds a document generation library. At the merge request, software composition analysis reports that an indirect dependency carries a published vulnerability, of a severity that the organisation’s rule classes as blocking.
The merge is refused, with an indication of the fixed version. The developer updates the dependency, the tests pass again, and the change is integrated. Without this control, the vulnerability might only have been spotted after the production release, with a fix to carry out as a matter of urgency.
These controls are set up in the delivery pipeline described in our page DevOps and platform engineering. How they fit with other tests is covered by our page software quality.