Sorry, not available in this language yet

Software Security Testing

The Security Testing practice is concerned with prerelease testing, including integrating security into standard quality assurance processes. The practice includes use of black-box security tools (including fuzz testing) as a smoke test in QA, risk-driven white-box testing, application of the attack model, and code coverage analysis. Security testing focuses on vulnerabilities in construction.

Security Testing Level 1

[ST1.1: 108] Perform edge/boundary value condition testing during QA.

QA efforts go beyond functional testing to perform basic adversarial tests and probe simple edge cases and boundary conditions, with no particular attacker skills required. When QA pushes past standard functional testing that uses expected input, it begins to move toward thinking like an adversary. Boundary value testing, whether automated or manual, can lead naturally to the notion of an attacker probing the edges on purpose (for example, determining what happens when someone enters the wrong password over and over).

[ST1.3: 97] Drive tests with security requirements and security features.

QA targets declarative security mechanisms with tests derived from security requirements and security features. A test could try to access administrative functionality as an unprivileged user, for example, or verify that a user account becomes locked after some number of failed authentication attempts. For the most part, security features can be tested in a fashion similar to other software features—security mechanisms such as account lockout, transaction limitations, entitlements, and so on are tested with both expected and unexpected input as derived from security requirements. Software security isn’t security software, but testing security features is an easy way to get started. New software architectures and deployment automation, such as with container and cloud infrastructure orchestration, might require novel test approaches.

[ST1.4: 56] Integrate opaque-box security tools into the QA process.

The organization uses one or more opaque-box security testing tools as part of the QA process. Such tools are valuable because they encapsulate an attacker’s perspective, albeit generically. Traditional dynamic analysis scanners are relevant for web applications, while similar tools exist for cloud environments, containers, mobile applications, embedded systems, and so on. In some situations, other groups might collaborate with the SSG to apply the tools. For example, a testing team could run the tool but come to the SSG for help with interpreting the results. When testing is integrated into Agile development approaches, opaque-box tools might be hooked into internal toolchains, provided by cloud-based toolchains, or used directly by engineering. Regardless of who runs the opaque-box tool, the testing should be properly integrated into a QA cycle of the SSDL and will often include both authenticated and unauthenticated reviews.

Security Testing Level 2

[ST2.4: 25] Drive QA tests with AST results.

Share results from application security testing, such as penetration testing, threat modeling, composition analysis, code reviews, etc., with QA teams to evangelize the security mindset. Using security defects as the basis for a conversation about common attack patterns or the underlying causes allows QA teams to generalize this information into new test approaches. Organizations that leverage software pipeline platforms such as GitHub, or CI/CD platforms such as the Atlassian stack, can benefit from teams receiving various testing results automatically, which should then facilitate timely stakeholder conversations—emailing security reports to QA teams will not generate the desired results. Over time, QA teams learn the security mindset, and the organization benefits from an improved ability to create security tests tailored to the organization’s code.

[ST2.5: 31] Include security tests in QA automation.

Security tests are included in an automation framework and run alongside functional, performance, and other QA test suites. Executing this automation framework can be triggered manually or through additional automation (e.g., as part of pipeline tooling). When test creators who understand the software create security tests, they can uncover more specialized or more relevant defects than commercial tools might (see [ST1.4]). Security tests might be derived from typical failures of security features (see [SFD1.1]), from creative tweaks of functional and developer tests, or even from guidance provided by penetration testers on how to reproduce an issue. Tests that are performed manually or out-of-band likely will not provide timely feedback.

[ST2.6: 21] Perform fuzz testing customized to application APIs.

QA efforts include running a customized fuzzing framework against APIs critical to the organization. An API might be software that allows two applications to communicate or even software that allows a human to interact with an application (e.g., a webform). Testers could begin from scratch or use an existing fuzzing toolkit, but the necessary customization often goes beyond creating custom protocol descriptions or file format templates to giving the fuzzing framework a built-in understanding of the application interfaces and business logic. Test harnesses developed explicitly for specific applications make good places to integrate fuzz testing.

Security Testing Level 3

[ST3.3: 12] Drive tests with design review results.

Use design review or architecture analysis results to direct QA test creation. For example, if the results of attempting to break a design determine that “the security of the system hinges on the transactions being atomic and not being interrupted partway through” then torn transactions will become a primary target in adversarial testing. Adversarial tests like these can be developed according to a risk profile, with high-risk flaws at the top of the list. Security defect data shared with QA (see [ST2.4]) can help focus test creation on areas of potential vulnerability that can, in turn, help prove the existence of identified high-risk flaws.

[ST3.4: 4] Leverage code coverage analysis.

Testers measure the code coverage of their application security testing to identify code that isn’t being exercised and then adjust test cases to incrementally improve coverage. AST can include automated testing (see [ST2.5], [ST2.6]) and manual testing (see [ST1.1], [ST1.3]). In turn, code coverage analysis drives increased security testing depth. Coverage analysis is easier when using standard measurements such as function coverage, line coverage, or multiple condition coverage. Measuring how broadly the test cases cover security requirements is not the same as measuring how broadly the test cases exercise the code. Note that standard-issue opaque-box testing tools (e.g., web application scanners, see [ST1.4]) provide exceptionally low coverage, leaving much of the software under test unexplored.

[ST3.5: 4] Begin to build and apply adversarial security tests (abuse cases).

QA teams incorporate test cases based on abuse cases (see [AM2.1]) as testers move beyond verifying functionality and take on the attacker’s perspective. One way to do this is to systematically attempt to replicate incidents from the organization’s history. Abuse and misuse cases based on the attacker’s perspective can also be derived from security policies, attack intelligence, standards, and the organization’s top N attacks list (see [AM2.5]). This effort turns the corner in QA from testing features to attempting to break the software under test.


[ST3.6: 3] Implement event-driven security testing in automation.

The SSG guides implementation of automation for continuous, event-driven application security testing. An event here is simply a noteworthy occurrence, such as dropping new code in a repository, a pull request, a build request, a push to deployment, or a Tuesday at noon. Event-driven testing implemented in pipeline automation typically moves the testing closer to the conditions driving the testing requirement (whether shift left toward design or shift right toward operations), repeats the testing as often as the event is triggered, and helps ensure that the right testing is executed for a given set of conditions. Success with this approach depends on the broad use of sensors (e.g., agents, bots) that monitor engineering processes, execute contextual rules, and provide telemetry to automation that initiates the specified testing whenever event conditions are met. More mature configurations proceed to including risk-driven conditions (e.g., size of change, provenance, function, team).