Директива CRA требует от компаний перехода от поиска уязвимостей к непрерывной верификации безопасности

Why Vulnerability Detection Isn’t Enough for the Cyber Resilience Act The Cyber Resilience Act (CRA) is raising the bar for software cybersecurity. Identifying vulnerabilities is only the beginning. Organizations increasingly need a repeatable process for assessing risk, fixing problems, verifying remediation, and producing evidence that the software can be trusted. For years, software security programs have focused heavily on finding vulnerabilities: run security scans, identify weaknesses, open tickets, correct problems, and close those tickets. The CRA makes that model increasingly insufficient because it establishes cybersecurity requirements across the life cycle of products with digital elements, from secure design and development to vulnerability handling, security updates, and post-release support. Finding a vulnerability is therefore no longer the end of the cybersecurity workflow. It is the beginning of an evidence chain showing what was affected, how the problem was addressed, whether the remediation worked, and how that conclusion was verified. From Vulnerability Detection to Verification When a vulnerability is discovered, whether through traditional security tools, vulnerability disclosures, automated analysis, or increasingly AI-assisted techniques, organizations need to determine which products and versions are affected, whether the vulnerability is exploitable within the product’s configuration, which source code or third-party component is involved, and what needs to change. AI can help accelerate that investigation and assist developers with remediation. But once a change is implemented, another question becomes equally important: How do we know the fix actually worked? A security fix changes the software, and like any software modification, it can have unintended consequences. Updating a component or changing source code might affect interfaces, memory behavior, timing, functionality, or dependent software. The change could even introduce another weakness. This is where vulnerability remediation becomes a software verification problem. Verification might therefore include static analysis, unit and regression testing, security testing, structural code coverage, or verification against security requirements. A closed vulnerability ticket demonstrates that a workflow was completed. It does not necessarily demonstrate that the resulting software is secure. This distinction is becoming urgent. CRA vulnerability-reporting requirements begin applying on Sept. 11, ahead of the broader requirements in December 2027. Organizations need operational processes capable of moving quickly from detection to assessment, remediation, verification, and documentation rather than reconstructing information manually after an incident. An SBOM Is Only Part of the Answer Software bills of materials (SBOMs) have become an important part of software supply chain security. Modern products can incorporate open-source libraries, operating systems, middleware, communication stacks, commercial components, and software from numerous suppliers. An SBOM provides visibility into those components. When a new vulnerability is disclosed, it can help determine whether potentially affected software appears in a product and which versions require investigation. But an SBOM cannot answer every cybersecurity question. The presence of a component associated with a known vulnerability does not automatically establish whether that vulnerability is exploitable in a particular product. And replacing an affected component does not prove that the resulting software has been successfully verified. The SBOM establishes what may be present. Engineering verification establishes what the organization did about the risk and whether the resulting change behaves as intended. That requires evidence connecting the vulnerability, affected software, remediation, verification activities, and results. Static analysis can identify weaknesses introduced by a change. Regression tests can determine whether existing functionality still behaves correctly. Security tests can verify the remediation, while requirements traceability can connect security requirements with the tests used to verify them. Secure by Design and Verified Continuously The CRA also reinforces a broader change in software engineering: cybersecurity cannot remain concentrated at the end of development. Penetration testing, vulnerability scanning, and security reviews remain important, but they should not be the first time an organization discovers a problem. Finding a weakness immediately before release can be expensive and disruptive, particularly when other software already depends on the affected code. Secure-by-design development moves security upstream. Security requirements can be considered alongside functional requirements, secure coding practices applied as software is written, and static analysis used to detect weaknesses before integration. Unit testing can verify security-relevant behavior at the component level. CI/CD can extend those controls throughout development. A source code change can automatically trigger static analysis, tests, and coverage analysis. Requirements traceability can connect security requirements to verification results. The pipeline then becomes more than a mechanism for delivering software faster. It becomes a continuous cybersecurity evidence engine, generating and retaining information about what was tested, what was found, what changed, and how the change was verified. Why This Matters Across the Software Ecosystem Modern products increasingly depend on complex software ecosystems that include proprietary code, open-source components, operating systems, middleware, communication technologies, cloud services, and software from numerous suppliers. This complexity creates cybersecurity challenges that extend across industries and throughout the product life cycle. Vulnerabilities can emerge long after software is released. A newly disclosed vulnerability in an open-source or third-party component, for example, may require organizations to determine which products are affected, assess whether the vulnerability is exploitable, implement a remediation, and verify that the resulting change does not introduce new problems. The challenge becomes even greater as connected products receive software updates throughout their operational lives. Fixing a vulnerability may require changing source code, updating a library or component, or modifying a software configuration. Each change can affect other parts of the system and therefore needs to be verified before it is released. This makes cybersecurity resilience an ongoing engineering responsibility rather than a one-time activity performed before a product reaches the market. Organizations need processes that can identify issues in deployed software, trace them to affected products and components, implement changes, verify those changes, and retain evidence of what was done. Every security-related software update therefore raises a fundamental engineering question: How do we know this change is ready to release? Verification Matters for C, C++, and AI-Generated Code This challenge is especially relevant for embedded C and C++. Defects involving memory access, integer operations, pointers, and resource management can create exploitable weaknesses. Secure coding practices and standards such as CERT and weakness classifications such as CWE can help identify these problems before they become vulnerabilities. Static analysis can continuously detect potential weaknesses without executing the software. Unit and regression testing can verify behavior, structural coverage can show how thoroughly relevant code has been exercised, and requirements traceability can demonstrate that security requirements have corresponding verification activities. Generative AI makes these controls even more important. Developers can now generate code, suggest fixes, refactor implementations, and create tests much faster. But faster code generation does not eliminate the need for verification. It increases the importance of automated controls over what enters the software baseline. AI-generated code can be subjected to the same coding standards, static analysis, automated tests, coverage requirements, and CI/CD quality gates as engineer-written code. AI can also assist verification by helping developers understand findings, remediate coding issues, generate unit tests, and identify coverage gaps. The objective is not to trust or distrust software because AI created it. Confidence should come from verification. Cyber Resilience Requires a Closed Loop The broader lesson from the CRA is that cybersecurity needs to become a closed-loop engineering process. Finding vulnerabilities remains essential, but resilience requires organizations to understand exposure, assess risk, remediate problems, verify changes, maintain traceability, and retain evidence throughout the product life cycle. That means connecting activities that have historically existed in separate domains: software development, cybersecurity, testing, vulnerability management, DevOps, and compliance. Organizations that establish those connections will be better prepared not only for the CRA, but for a software environment increasingly defined by connected products, continuous updates, complex supply chains, and AI-assisted development. The cybersecurity question is therefore evolving. It is no longer enough to ask whether an organization found and addressed a vulnerability. Can it demonstrate that the vulnerability was correctly remediated, and provide the engineering evidence that the software can still be trusted?