<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Frank Leitner on Medium]]></title>
        <description><![CDATA[Stories by Frank Leitner on Medium]]></description>
        <link>https://medium.com/@frank.leitner?source=rss-7eea262d9b94------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/2*MeyZdzttZ1LjXiZ0axa86A.jpeg</url>
            <title>Stories by Frank Leitner on Medium</title>
            <link>https://medium.com/@frank.leitner?source=rss-7eea262d9b94------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 11:15:20 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@frank.leitner/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[STRIDE Is a Technique, Not a Threat Modeling Process]]></title>
            <link>https://meetcyber.net/stride-is-a-technique-not-a-threat-modeling-process-813b9c8683f1?source=rss-7eea262d9b94------2</link>
            <guid isPermaLink="false">https://medium.com/p/813b9c8683f1</guid>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[product-security]]></category>
            <category><![CDATA[secure-by-design]]></category>
            <category><![CDATA[security-engineering]]></category>
            <category><![CDATA[threat-modeling]]></category>
            <dc:creator><![CDATA[Frank Leitner]]></dc:creator>
            <pubDate>Wed, 15 Apr 2026 14:14:47 GMT</pubDate>
            <atom:updated>2026-04-15T14:14:47.600Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nFFJBRWHl_WK-KkeyCttHA.png" /></figure><p>STRIDE produces threats. That is what it is designed to do. The problem is that producing threats is the middle part of threat modeling, and most teams treat it as the whole thing.</p><p>The conflation is understandable. STRIDE is structured, teachable, and generates visible output quickly. You draw a data flow diagram, apply six categories (Spoofing, Tampering, Repudiation, Information Disclosure, Denial of Service, Elevation of Privilege) against each element in the diagram, and end up with a list. That list is real. It represents potential attacker actions against real parts of your architecture. But a list of threats is not a threat model. It is an intermediate artifact that only becomes useful inside a process that STRIDE does not provide.</p><p>What STRIDE does not tell you is what to include in that data flow diagram, or how fine-grained the decomposition should be. It does not tell you which of the threats it generates are worth acting on, in what order, or how to connect them to the security controls that will actually address them. Those decisions sit outside the technique, and they determine whether the output becomes a work plan or a catalog that gets reviewed once and never referenced again.</p><h3>What STRIDE Actually Does</h3><p>STRIDE is a structured elicitation checklist. Applied to each element of a data flow diagram, it prompts an analyst to ask whether that element is vulnerable to each of six threat categories. For a process, that means asking whether it can be spoofed, whether its data can be tampered with, whether actions against it can be repudiated, and so on. For a data flow, it asks whether the transmission can be intercepted, altered, or interrupted. The technique produces a hypothesis for each applicable combination: “An attacker could spoof the authentication service to redirect session tokens.”</p><p>That hypothesis is worth having. The discipline of applying a fixed set of categories to every element in the diagram reduces the chance of obvious threats being missed through oversight or familiarity bias. It is more systematic than open-ended brainstorming, and it produces comparable outputs across sessions when different analysts apply it to similar systems.</p><p>Those properties make STRIDE an effective elicitation technique. They do not make it a threat modeling methodology. Elicitation is one step. The inputs that precede it and the processing that follows it are what make that step useful.</p><blockquote>STRIDE answers one question rigorously. Threat modeling requires answers to three.</blockquote><h3>The Scoping Problem STRIDE Ignores</h3><p>Before applying STRIDE to anything, someone has to decide what the threat model covers. That decision shapes every output the technique produces, and STRIDE offers no guidance on making it.</p><p>Scope drives volume. A threat model covering an entire product spans a different attack surface than one scoped to a specific interface, a particular deployment scenario, or a single feature under review. Applying STRIDE uniformly without first bounding what is in scope is the primary mechanism by which threat modeling exercises generate more threats than any team can process.</p><p>Two things anchor all scope decisions before any other question is resolved: the product’s intended purpose and the constraints and assumptions of its intended operational environment. A remote firmware update service designed for internet-connected consumer devices and the same service embedded within an air-gapped industrial network may share a code base. The assets that matter, the threat actors who are realistic, and the parts of the architecture that require deep examination differ completely. Intended purpose and operational environment together define the frame within which every other scope decision is made.</p><p>Within that frame, scope has four further dimensions to resolve. The product boundary establishes where your system ends and adjacent infrastructure begins. The threat actor definition establishes who you are modeling against and what capabilities you are attributing to them. The asset set defines what you are trying to protect and against what classes of harm. The deployment context establishes the specific configuration and connectivity assumptions that apply to this threat model rather than to the product in the abstract.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/611/1*CKFsoMoWISg0OZDrGuJEfw.png" /></figure><p>None of these dimensions emerge from applying STRIDE. They are inputs that determine what the technique operates against. A team that starts with the technique before resolving scope has made implicit decisions about all four dimensions, and implicit decisions are rarely the right ones.</p><blockquote>Define the wrong scope, and STRIDE will generate the right threats for the wrong system….</blockquote><h3>The Decomposition Level STRIDE Cannot Choose For You</h3><p>Once scope is defined, the architecture needs to be decomposed into the elements that STRIDE operates against: processes, data stores, data flows, and external entities in a data flow diagram (DFD). This is the second structural gap: STRIDE assumes that decomposition is already complete. It applies identically at any level of granularity and provides no signal about whether that level is appropriate. A design review for a new internet-facing authentication mechanism needs a different level of detail than an internal read-only connection between two parts within an closed system. STRIDE generates output at either level without distinguishing between them.</p><p>Decomposition granularity is not a neutral choice. Too coarse, and threats that only become visible at a more specific level of the architecture will not surface. Too granular, and the number of diagram elements multiplies combinatorially with the six STRIDE categories, producing a threat count that far exceeds the team’s capacity to evaluate it during a real development cycle.</p><p>The criterion for deciding how deep to decompose at any given level is worst-case impact, not risk or likelihood. At each decomposition step, examine what each segment of the architecture contains: which data assets, which functions, which interfaces to adjacent systems. Ask what happens if an attacker achieves full control of or complete access to that segment in the worst case. If the impact ceiling is low (the segment processes non-sensitive data, controls no safety-critical function, and offers no path to higher-value assets), a coarse view is sufficient. If the impact ceiling is significant, decompose one level further and apply the same question to each of the resulting sub-segments. Repeat until either the impact assessment says you can stop or you have reached the lowest addressable level of the system.</p><h4>The Impact Ceiling Decision at Each Decomposition Step</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WxzmxtikhpE19f_vXBPGSA.png" /></figure><p>This is iterative by design. The first pass often reveals that only one or two subsystems warrant deeper examination. Applying the same six STRIDE categories at the same level of granularity across all segments would have generated threats in the low-impact areas that no engineering team could realistically act on. Impact-driven decomposition bounds the analysis to the parts of the system where the threats that surface are worth addressing.</p><p>In practice, teams tend to decompose more finely than necessary because detail signals thoroughness. The effect is a threat model that is technically exhaustive and operationally useless: it contains more entries than any engineering sprint can address, which means triaging it becomes a second, unplanned project.</p><blockquote>The threat model that covers everything is the one that changes nothing.</blockquote><h3>What Happens After STRIDE</h3><p>Assume scope is correctly defined, decomposition is at the right level, and STRIDE has been applied. The output is a structured list of threats against every element in the diagram. The next question is what to do with it.</p><p>Raw STRIDE output is not actionable. It is a catalog. Every entry represents a category-based hypothesis derived from applying a fixed checklist to an architecture element. The technique generates these entries uniformly, regardless of whether the underlying attack is realistic in the specific deployment context, whether exploitation would yield meaningful impact, or whether the system already has controls that address it. STRIDE does not know the difference between a theoretical threat and an exploitable one. It does not distinguish between a threat that has been mitigated and one that has been ignored.</p><p>Two capabilities are needed to turn a catalog into a work plan. Prioritization assigns each threat a risk rating based on exploitability, potential impact, and the presence or absence of existing mitigations, then produces a ranked list that engineering can act on within realistic resource constraints. Traceability connects each prioritized threat to a specific security control or design decision, creating a record that links threat analysis to implementation requirements and, eventually, to verification activities.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4uyLufAsTmE-4z-8oa7v7Q.png" /></figure><p>STRIDE provides neither prioritization nor traceability. It generates the input to those steps and stops. Teams that do not run those steps produce a document that demonstrates they applied a structured technique, without demonstrating that any security decision was made as a result.</p><blockquote>Raw STRIDE output is just inventory. Prioritization is what turns it into a work plan.</blockquote><h3>The Process STRIDE Belongs In</h3><p>STRIDE has a defined function: structured threat elicitation against a decomposed architecture. It performs that function well. The confusion arises because the function is visible and bounded, while the process it belongs in is neither.</p><p>A properly structured threat modeling process has three phases. The first is scoping and decomposition: establishing the system boundary, defining assets and threat actors, selecting a deployment context, and producing a data flow diagram at a granularity matched to the decision the threat model is meant to support. The second is threat elicitation, which is where STRIDE belongs. The third is risk-based prioritization and traceability: scoring threats against a defined risk model, filtering for those above the product’s risk threshold (taking into account the product’s intended purpose), linking each to a control and connecting the output to security requirements and test cases.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/829/1*BHC98iWjMdRx-KpEAY6Nog.png" /></figure><p>STRIDE is phase 2 of that process. Running phase 2 without phase 1 means applying the technique to an implicitly bounded system whose decomposition level was not chosen deliberately. Running phase 2 without phase 3 means producing output that cannot drive engineering decisions. Both omissions are common, and their consequences are different in kind.</p><p>Omitting phase 3 produces a catalog that cannot be acted on. The output exists; it simply has no decision value attached.</p><p>Omitting phase 1 produces something even worse: a list with no boundary and no triage criterion. There is no definition of what is in scope, no basis for distinguishing a relevant threat from an irrelevant one, no asset criticality signal to tell the team where to look harder and where to stop, and no natural termination point for the analysis. The list grows until the team runs out of time or patience. What makes this particularly costly is that the problem is invisible in the output itself. A STRIDE pass conducted without Phase 1 produces a document that looks exactly like a STRIDE pass conducted with it. The incompleteness is in what was never decided, not in what is missing from the page.</p><p>Organizations that have adopted STRIDE as their threat modeling practice typically have solid threat generation. The gaps are in what precedes and follows it.</p><blockquote>A technique without a process is a step without a staircase.</blockquote><h3>Closing</h3><p>The dominance of STRIDE in threat modeling discussions reflects something real: it is a well-designed elicitation framework that produces structured output from a defined input. Those properties are valuable. They are also the reason it gets treated as a complete solution when it is not.</p><p>Teams that run STRIDE without first resolving scope and decomposition generate threats faster than they can evaluate them, against a system boundary that was never deliberately chosen. Teams that stop at STRIDE output without prioritization and traceability produce a catalog that satisfies the appearance of having done threat modeling while leaving the actual security decisions unmade.</p><blockquote>STRIDE did not cause either problem. The absence of a process around it did.</blockquote><p>Threat models that look complete but produce no actionable decisions are not a neutral outcome. They provide cover for skipping the work while the architecture decisions that are hardest to reverse get made without security input. By the time the threat model is revisited, the design is fixed.</p><p>I consult on CRA and IEC 62443 implementation for manufacturers of software-driven products. You can reach me on <a href="https://www.linkedin.com/in/frankleitner/">LinkedIn </a>or my <a href="https://www.frank-leitner.com/">website</a>.</p><p><em>Originally published at </em><a href="https://www.linkedin.com/pulse/stride-technique-threat-modeling-process-frank-leitner-fbhmf/?trackingId=FY29cDdKTZC7eZB5%2Frir4Q%3D%3D"><em>https://www.linkedin.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=813b9c8683f1" width="1" height="1" alt=""><hr><p><a href="https://meetcyber.net/stride-is-a-technique-not-a-threat-modeling-process-813b9c8683f1">STRIDE Is a Technique, Not a Threat Modeling Process</a> was originally published in <a href="https://meetcyber.net/">MeetCyber</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Audit Finding That Nobody Owns]]></title>
            <link>https://meetcyber.net/the-audit-finding-that-nobody-owns-292b393e14f3?source=rss-7eea262d9b94------2</link>
            <guid isPermaLink="false">https://medium.com/p/292b393e14f3</guid>
            <category><![CDATA[software-security]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[risk-management]]></category>
            <category><![CDATA[product-security]]></category>
            <category><![CDATA[security-governance]]></category>
            <dc:creator><![CDATA[Frank Leitner]]></dc:creator>
            <pubDate>Tue, 14 Apr 2026 06:39:49 GMT</pubDate>
            <atom:updated>2026-04-14T06:39:49.371Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*vtSgk9KuVNZ2EcKz" /></figure><p>Security audits produce findings. Findings get logged. And then, in many organizations, they age.</p><p>Not because someone decided the risk was acceptable. Not because the finding was disputed. But because nobody was specifically responsible for fixing it, and no deadline was attached.</p><p>That distinction matters more than it sounds. A finding that is tracked lives in a spreadsheet. A finding that is managed has an owner, a remediation path, and a fixed-by date. Most audit processes are excellent at the first part and structurally silent on the second.</p><p>The result is predictable. The list grows. The findings from last year’s audit resurface in this year’s. The security team reports on volume and closure rates. Leadership sees a dashboard. Nobody is asking the harder question: which of these findings has an owner who is accountable for remediation, not just aware of the problem?</p><p>Awareness is not accountability. A finding that five people know about but nobody owns is, functionally, unmanaged. And unmanaged findings do not stay static. The environment changes, threat actors evolve, and yesterday’s medium-severity finding can become tomorrow’s incident.</p><p>Fixing this is not primarily a tooling problem. It is a governance problem. Someone needs to assign ownership at the point of finding creation, not weeks later during a review meeting. That owner needs a deadline and the organizational authority to act.</p><blockquote>A security audit without a remediation accountability structure is not a security process. It is a documentation exercise.</blockquote><p>I consult on CRA and IEC 62443 implementation for manufacturers of software-driven products. You can reach me on <a href="https://www.linkedin.com/in/frankleitner/">LinkedIn </a>or my <a href="https://www.frank-leitner.com/">website</a>.</p><p><em>Originally published at </em><a href="https://www.linkedin.com/posts/frankleitner_cybersecurity-securitygovernance-riskmanagement-activity-7442835237044264960-NL6w?utm_source=share&amp;utm_medium=member_desktop&amp;rcm=ACoAAANvnncB9GERsfzPwI0-xmeXP6B2a4uXiOg"><em>https://www.linkedin.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=292b393e14f3" width="1" height="1" alt=""><hr><p><a href="https://meetcyber.net/the-audit-finding-that-nobody-owns-292b393e14f3">The Audit Finding That Nobody Owns</a> was originally published in <a href="https://meetcyber.net/">MeetCyber</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Secure by Design vs Secure by Default]]></title>
            <link>https://meetcyber.net/secure-by-design-vs-secure-by-default-63f14543268e?source=rss-7eea262d9b94------2</link>
            <guid isPermaLink="false">https://medium.com/p/63f14543268e</guid>
            <category><![CDATA[software-security]]></category>
            <category><![CDATA[secure-by-design]]></category>
            <category><![CDATA[secure-by-default]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[product-security]]></category>
            <dc:creator><![CDATA[Frank Leitner]]></dc:creator>
            <pubDate>Wed, 08 Apr 2026 23:33:02 GMT</pubDate>
            <atom:updated>2026-04-08T23:33:02.918Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*klIlkppE_sMRjyCc_UJkNg.png" /></figure><p>“Secure by Design” and “Secure by Default”.</p><p>Two terms often used in the same breath and interchangeably. <br>Meaning very different things.</p><blockquote><strong>Secure by Design is about how you build.</strong></blockquote><p>It means security is considered from the start, not bolted on after the architecture is set. Threat modeling happens before design decisions are finalized. Security requirements are derived from risk analysis, not added as a checklist at the end of the sprint. The development process itself is structured to prevent classes of vulnerabilities from being introduced in the first place.</p><blockquote><strong>Secure by Default is about what ships.</strong></blockquote><p>It means the product, out of the box, is configured in the most secure way that is practical for the intended use case. No unnecessary services enabled. Authentication required. Logging on. Passwords not “admin/admin”.</p><p>A product can be built with Secure by Design principles and still ship with insecure defaults.<br>And a product with locked-down defaults can still be full of architectural vulnerabilities.</p><p>You need both.<br>They address different problems at different points in the lifecycle.<br>Confusing them creates blind spots.</p><blockquote><strong>Secure by Design prevents vulnerabilities.<br>Secure by Default prevents unnecessary exposure.</strong></blockquote><p>The EU Cyber Resilience Act requires both. Not as a philosophical exercise, but as verifiable obligations. That distinction matters when you’re trying to demonstrate compliance, not just claim it.</p><p>In practice, the design gap is usually the harder problem.<br>Insecure defaults are embarrassing.<br>Security architecture mistakes are expensive.</p><p>And much harder to fix once the product ships.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=63f14543268e" width="1" height="1" alt=""><hr><p><a href="https://meetcyber.net/secure-by-design-vs-secure-by-default-63f14543268e">Secure by Design vs Secure by Default</a> was originally published in <a href="https://meetcyber.net/">MeetCyber</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Security requirement traceability vs. security requirement coverage]]></title>
            <link>https://meetcyber.net/security-requirement-traceability-vs-security-requirement-coverage-abdf5dad94a6?source=rss-7eea262d9b94------2</link>
            <guid isPermaLink="false">https://medium.com/p/abdf5dad94a6</guid>
            <category><![CDATA[cyber-resilience-act]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[product-security]]></category>
            <category><![CDATA[software-security]]></category>
            <category><![CDATA[security-engineering]]></category>
            <dc:creator><![CDATA[Frank Leitner]]></dc:creator>
            <pubDate>Tue, 07 Apr 2026 20:14:25 GMT</pubDate>
            <atom:updated>2026-04-08T23:41:23.286Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*TSSD3VWcsnTEWLha" /></figure><p>A completed risk assessment mapped to security requirements looks like security engineering. Often, it is just the first half of it.</p><p>Coverage answers whether a requirement exists for each identified risk. Traceability answers whether that requirement can be followed through design, implementation, testing, and release. The first is a completeness check. The second is an evidence chain.</p><p>A product can have full coverage and zero traceability. Every risk has a requirement, but none of them can be confirmed as actually met. From an engineering standpoint, the work was done. From a conformity standpoint, it is invisible.</p><p>This distinction has direct consequences under the CRA. Module A self-assessment requires compiling a technical file that demonstrates conformity with Annex I essential requirements. A requirements list mapped to a risk assessment is a coverage artifact. What survives scrutiny is a traceable chain: risk, requirement, design decision, test result, release state.</p><p>The gap surfaces during conformity prep, not during development. Teams find they wrote requirements against the risk assessment but never linked them to architecture decisions, never referenced them in test plans, never confirmed them in acceptance criteria. The requirements existed. The thread is gone.</p><blockquote>Coverage without traceability is documentation. Traceability is what makes it defensible.</blockquote><p>I consult on CRA and IEC 62443 implementation for manufacturers of software-driven products. You can reach me on <a href="https://www.linkedin.com/in/frankleitner/">LinkedIn </a>or my <a href="https://www.frank-leitner.com/">website</a>.</p><p><em>Originally published at </em><a href="https://www.linkedin.com/posts/frankleitner_a-completed-risk-assessment-mapped-to-security-activity-7447176350882316288--ENx"><em>https://www.linkedin.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=abdf5dad94a6" width="1" height="1" alt=""><hr><p><a href="https://meetcyber.net/security-requirement-traceability-vs-security-requirement-coverage-abdf5dad94a6">Security requirement traceability vs. security requirement coverage</a> was originally published in <a href="https://meetcyber.net/">MeetCyber</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Gap Analysis and Readiness Assessment: Not the Same Thing]]></title>
            <link>https://meetcyber.net/gap-analysis-and-readiness-assessment-not-the-same-thing-3e9d4c7cd67e?source=rss-7eea262d9b94------2</link>
            <guid isPermaLink="false">https://medium.com/p/3e9d4c7cd67e</guid>
            <category><![CDATA[compliance]]></category>
            <category><![CDATA[product-management]]></category>
            <category><![CDATA[software-security]]></category>
            <category><![CDATA[product-security]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <dc:creator><![CDATA[Frank Leitner]]></dc:creator>
            <pubDate>Mon, 06 Apr 2026 20:34:45 GMT</pubDate>
            <atom:updated>2026-04-06T20:34:45.428Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*kCa9ZDlo3bsQJuel" /></figure><p>A completed gap analysis feels like progress. The requirements are mapped. The gaps are documented. The remediation plan has a timeline. Leadership has signed off. What it does not tell you is whether any of that is executable. An organization can run a rigorous, accurate gap analysis against CRA Annex I, produce a credible remediation roadmap, and still arrive at December 2027 non-compliant, not because the gaps were wrong, but because the capacity to close them was never verified. That is the failure the readiness assessment is designed to prevent. Skipping it does not shorten the preparation process. It just delays the point at which the plan breaks.</p><h3>What a Gap Analysis Actually Tells You</h3><p>A gap analysis measures distance. You take a defined standard or regulatory requirement, map your current state against it, and document what is missing or incomplete. For CRA preparation, this means working through the essential cybersecurity requirements in Annex I, comparing them to your current secure development lifecycle, vulnerability handling, SBOM practices, and conformity documentation.</p><p>Done well, a gap analysis is useful and necessary. It gives you a structured picture of what the regulation requires, where you are today, and how large the gap is for each requirement. For CRA, this involves input from multiple teams: development leads, product managers, DevOps, legal, and compliance. The output is a list of gaps prioritized by compliance impact and implementation effort.</p><p>That output is also the limit of what a gap analysis can tell you. It describes a distance. It does not tell you whether you can cover it.</p><blockquote><em>A completed gap analysis is the most dangerous moment to stop asking questions.</em></blockquote><h3>What a Readiness Assessment Adds</h3><p>A readiness assessment asks a different question. Given the gaps identified, does the organization have the capacity, process maturity, and tooling to close them within the available time?</p><p>Those three dimensions are separable and each can fail independently.</p><p>Capacity is headcount, budget, and availability. A gap list might show that implementing SBOM generation and automation requires four months of dedicated engineering effort. The readiness question is whether that effort is actually available, or whether the same engineers are already committed to product roadmap deliverables through the end of the year. A gap analysis does not answer this. A readiness assessment does.</p><p>Process maturity is more specific than it sounds. An organization can have a documented vulnerability management procedure and still lack the organizational muscle to run it under real conditions. IEC 62443–4–1 separates Maturity Level 2, where a procedure is written and personnel are trained, from Level 3, where that procedure has been practiced across the organization for at least one product. The gap matters because the first real test of an unpracticed procedure is not a dry run: it is an actual incident, an actual notified body review, or an actual regulator request. Handoff points that look clean in documentation break down when a report arrives outside business hours, when ownership between intake and triage was never operationally confirmed, or when the evidence required for a 72-hour notification turns out to live in three separate systems with no single owner. A readiness assessment surfaces those gaps before execution begins. A gap analysis has no mechanism to find them at all.</p><p>Tooling creates lead times that do not appear on gap analysis timelines because they precede the work the gap analysis measured. Before SBOM generation can be implemented, build pipelines need to be instrumented, tooling needs to be evaluated, procured, and integrated, and the output needs to be validated against real component data. That sequence takes months and must be complete before the implementation work starts. The same applies to vulnerability intake infrastructure: a ticketing system with the right workflow and ownership model is a prerequisite for the process, not a byproduct of it. If a readiness assessment reveals that tooling selection has not started, the realistic implementation timeline shifts by the length of that procurement and integration cycle, whether the gap analysis acknowledged it or not.</p><blockquote><em>The work the gap analysis measured cannot start until the infrastructure work it ignored is finished.</em></blockquote><h3>Why Skipping the Readiness Assessment Makes the Gap Analysis Misleading</h3><p>A gap analysis that does not account for organizational readiness produces an implementation plan with structurally optimistic timelines. The gaps are real. The plan to close them may not be.</p><p>This matters practically for CRA because there are two deadlines, not one. The mandatory reporting obligations for actively exploited vulnerabilities and severe incidents apply from September 2026: a 24-hour early warning followed by a 72-hour detailed notification to ENISA. That process needs to be operational, not planned. For organizations that have not yet built the intake, triage, and notification infrastructure, September 2026 is already close. The December 2027 enforcement deadline for the full set of Annex I requirements adds a second horizon, but it does not reduce the pressure on the first. Reaching December 2027 with a tested and verifiable process requires roughly eighteen months of focused, multi-team effort with dedicated resources.</p><p>Each phase has dependencies: you cannot document processes that do not exist, cannot automate pipelines whose process logic is still undefined, and cannot conduct a meaningful conformity readiness review before the underlying implementations are complete.</p><p>If the readiness assessment reveals that engineering teams are available at 50% capacity, the realistic timeline extends accordingly. If it reveals that the vulnerability management process exists on paper but has never been run against an actual report, the maturity gap requires time that a gap analysis did not account for. If it reveals that SBOM tooling selection has not started, that procurement and integration cycle sits before the implementation timeline, not inside it.</p><p>Notified bodies, for products that require third-party conformity assessment, compound the problem further. They have finite capacity, and that capacity is being reserved as enforcement approaches. An organization that completes its gap analysis and then discovers in a readiness assessment that it cannot realistically finish implementation before Q3 2027 has little margin to secure assessment capacity. The timeline consequences of a late readiness assessment are not recoverable by working harder.</p><blockquote><em>A accurate gap analysis and a failing project are not contradictions. They are a sequence.</em></blockquote><h3>What a Readiness Assessment Looks Like in Practice</h3><p>A readiness assessment is not a second gap analysis. It evaluates each significant workstream against three variables: who owns it and whether they are available at the level the plan assumes; what tooling or process dependencies must be resolved before the workstream can start; and whether the current process maturity is sufficient to sustain the change once implemented, or whether organizational capability needs to be built first.</p><p>When the answers are bad, the output is not a failed assessment. It is a revised timeline, a reprioritized sequence, or a resourcing decision that needs to be made before the remediation plan is approved. That friction is useful. It is cheaper to encounter before execution starts than after two quarters of slippage have made the December 2027 buffer unrecoverable.</p><p>A plan built on an honest readiness assessment will absorb the friction of execution. A plan built on a gap analysis alone will encounter it instead.</p><blockquote><em>Bad answers in a readiness assessment are cheap. The same answers discovered mid-execution are not.</em></blockquote><h3>The Conversation to Have Now</h3><p>The structural failure of skipping a readiness assessment is not visible at plan approval. The gaps are documented, the timeline looks reasonable, and the workstreams are assigned. What is not visible is that the timeline was built without a load-bearing capacity estimate. The engineers assigned to SBOM implementation are at 60% availability because no one checked. The vulnerability management process has never been practiced end-to-end because the team assumed the documentation was sufficient. The tooling procurement is in month three of a process the plan allocated six weeks for. Each of those failures is independent. Together, they produce a plan that held together on paper until mid-execution, when the dependencies between workstreams made the slippage unrecoverable.</p><p>For product managers and executives responsible for CRA preparation: if your current roadmap was built from a gap analysis without a readiness assessment, you have a list of intentions organized chronologically. The remediation plan describes what needs to happen. It does not establish that your organization can make it happen at the pace the timeline assumes.</p><blockquote><em>A gap without a route is just a distance measurement. A route without a readiness check is just optimism with a Gantt chart attached.</em></blockquote><p>I consult on CRA and IEC 62443 implementation for manufacturers of software-driven products. You can reach me on <a href="https://www.linkedin.com/in/frankleitner/">LinkedIn </a>or my <a href="https://www.frank-leitner.com/">website</a>.</p><p><em>Originally published at </em><a href="https://www.linkedin.com/pulse/gap-analysis-readiness-assessment-same-thing-frank-leitner-epbbf"><em>https://www.linkedin.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3e9d4c7cd67e" width="1" height="1" alt=""><hr><p><a href="https://meetcyber.net/gap-analysis-and-readiness-assessment-not-the-same-thing-3e9d4c7cd67e">Gap Analysis and Readiness Assessment: Not the Same Thing</a> was originally published in <a href="https://meetcyber.net/">MeetCyber</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[ISMS Policy vs. Product Security Policy]]></title>
            <link>https://meetcyber.net/isms-policy-vs-product-security-policy-d1f982919506?source=rss-7eea262d9b94------2</link>
            <guid isPermaLink="false">https://medium.com/p/d1f982919506</guid>
            <category><![CDATA[iec-62443]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[cyber-resilience-act]]></category>
            <category><![CDATA[product-security]]></category>
            <category><![CDATA[secure-by-design]]></category>
            <dc:creator><![CDATA[Frank Leitner]]></dc:creator>
            <pubDate>Thu, 02 Apr 2026 08:28:07 GMT</pubDate>
            <atom:updated>2026-04-02T08:28:07.736Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*r50dLrHwDf9706F-" /></figure><p>Most manufacturers have a security policy. The question is what it actually covers.</p><p>An information security policy governs how the organization protects its own assets: networks, data, access controls, incident response. This is ISO 27001 territory. It answers: how do we secure what we run?</p><p>A product security policy is different. It defines how security obligations attach to the product development process specifically: which products are in scope for structured threat modeling, what security requirements need to be documented and verified before release, how vulnerability handling responsibilities are assigned, what the update support commitment looks like over the product’s lifecycle.</p><p>The CRA assumes this second document exists. Its obligations under Article 13 require manufacturers to have systematic processes for exactly these things, tied to the product, not the organization’s internal IT environment. When a manufacturer’s gap analysis surfaces for the first time, the ISMS policy frequently comes up as evidence. It is not the right document.</p><p>The gap rarely shows up in audits because most organizations have not yet been audited against the CRA. It shows up when someone tries to map existing documentation to Annex I requirements and finds that the organizational policy says nothing about threat modeling gates, security verification criteria, or end-of-support declarations.</p><p>An ISMS policy protects the organization from threats. A product security policy commits the organization to building products that protect its customers from threats. These are different scopes, targetting different audiences and having different enforcement points.</p><p>Having one does not substitute for the other.</p><p>I consult on CRA and IEC 62443 implementation for manufacturers of software-driven products. You can reach me on <a href="https://www.linkedin.com/in/frankleitner/">LinkedIn </a>or my <a href="https://www.frank-leitner.com/">website</a>.</p><p><em>Originally published at </em><a href="https://www.linkedin.com/feed/update/urn:li:activity:7444994497723289601/"><em>https://www.linkedin.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d1f982919506" width="1" height="1" alt=""><hr><p><a href="https://meetcyber.net/isms-policy-vs-product-security-policy-d1f982919506">ISMS Policy vs. Product Security Policy</a> was originally published in <a href="https://meetcyber.net/">MeetCyber</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why Least Privilege Fails in Practice]]></title>
            <link>https://meetcyber.net/why-least-privilege-fails-in-practice-4eab4712c048?source=rss-7eea262d9b94------2</link>
            <guid isPermaLink="false">https://medium.com/p/4eab4712c048</guid>
            <category><![CDATA[least-privilege]]></category>
            <category><![CDATA[technical-debt]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[product-security]]></category>
            <category><![CDATA[security-engineering]]></category>
            <dc:creator><![CDATA[Frank Leitner]]></dc:creator>
            <pubDate>Tue, 31 Mar 2026 18:08:52 GMT</pubDate>
            <atom:updated>2026-03-31T23:01:40.125Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*dYBgpRi4I1lIjqOD" /></figure><p>Least privilege is one of the few security principles everyone agrees on.</p><p>The implementation rarely matches that consensus. The usual version looks like this: user accounts get scoped roles, someone documents it in the access control policy, the box gets checked. What the policy does not cover are the non-human principals that actually run the system. Service accounts, API keys, inter-process calls, build pipelines: the identities that authenticate and act without a human behind them. Service accounts carry administrative rights because narrowing them down was inconvenient at the time. API keys accumulate scope across integrations and rarely get revisited once they work. Inter-process calls assume trust because they stay inside the perimeter, and build pipelines hold production credentials long after the person who provisioned them has moved on.</p><p>These are not just edge cases. They are the default outcome when least privilege gets treated as a user account configuration task rather than a system-wide constraint.</p><p>The deeper problem is the gap between what a policy states and what an environment actually enforces. A policy can say “minimal necessary permissions for all principals.” An environment where service accounts were provisioned years ago, by people who are no longer there, against requirements that have since changed, is a different thing entirely. Auditing the policy does not close that gap. Only auditing the environment does.</p><p>Least privilege applied to human users is a start. Privilege escalation through a compromised service account is a documented failure mode in post-incident reports, not a theoretical risk. So is an attacker moving from a single API key to adjacent systems because the key’s scope was never restricted. So are supply chain attacks through build pipelines with credentials that should have been rotated years ago.</p><p>The principle is sound. The scope of its application is where implementations fall short.</p><p>I consult on CRA and IEC 62443 implementation for manufacturers of software-driven products. You can reach me on <a href="https://www.linkedin.com/in/frankleitner/">LinkedIn </a>or my <a href="https://www.frank-leitner.com/">website</a>.</p><p><em>Originally published at </em><a href="https://www.linkedin.com/feed/update/urn:li:activity:7444632195291455489"><em>https://www.linkedin.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4eab4712c048" width="1" height="1" alt=""><hr><p><a href="https://meetcyber.net/why-least-privilege-fails-in-practice-4eab4712c048">Why Least Privilege Fails in Practice</a> was originally published in <a href="https://meetcyber.net/">MeetCyber</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Security Control vs Security Requirement]]></title>
            <link>https://meetcyber.net/security-control-vs-security-requirement-dbe3fee8388f?source=rss-7eea262d9b94------2</link>
            <guid isPermaLink="false">https://medium.com/p/dbe3fee8388f</guid>
            <category><![CDATA[product-security]]></category>
            <category><![CDATA[cyberse]]></category>
            <category><![CDATA[cyber-resilience-act]]></category>
            <category><![CDATA[secure-by-design]]></category>
            <category><![CDATA[software-security]]></category>
            <dc:creator><![CDATA[Frank Leitner]]></dc:creator>
            <pubDate>Mon, 30 Mar 2026 11:01:01 GMT</pubDate>
            <atom:updated>2026-04-03T23:23:53.331Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*l3qPh3A7Mv_vPxeQqtmqKQ.png" /></figure><p>A security control and a security requirement are not the same thing. Treating them as interchangeable is one of the more consistent mistakes in 62443–4–2 implementation work.</p><p>A control is a solution: role-based access control, encrypted storage, session timeout. These are answers. <br>A requirement defines the problem that needs solving: which assets need protection, from which threats, at what level of rigor. <br>One is a design decision. The other is the basis for evaluating that decision.</p><p>The gap matters because, without the requirement, you cannot determine whether the control is sufficient. You cannot determine whether it is proportionate to the risk it is supposed to address. You cannot even confirm it is relevant to the threat scenario that prompted it. You are left with an implementation that may be technically sound but cannot be defended as adequate.</p><p>In 62443–4–2 work, this shows up regularly. Teams implement the component requirements as a checklist: CR 1.1 covered, CR 2.1 covered, CR 3.1 covered. The controls exist. The traceability back to identified threats and product-specific risk decisions does not. What looks like compliance is often just a list of implemented features with no engineering rationale connecting them to the problem they are supposed to solve.</p><p>This is exactly the gap between 62443–4–1 and 4–2 in practice: controls get implemented, requirements never get operationalized. The requirements are not an input to the control list, they are the justification for it.</p><blockquote>A control without a traceable requirement is an assumption. During an audit, assumptions do not hold.</blockquote><p>I consult on CRA and IEC 62443 implementation for manufacturers of software-driven products. You can reach me on <a href="https://www.linkedin.com/in/frankleitner/">LinkedIn </a>or my <a href="https://www.frank-leitner.com/">website</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=dbe3fee8388f" width="1" height="1" alt=""><hr><p><a href="https://meetcyber.net/security-control-vs-security-requirement-dbe3fee8388f">Security Control vs Security Requirement</a> was originally published in <a href="https://meetcyber.net/">MeetCyber</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[EU Cyber Resilience Act: What Product Teams Should Do Now]]></title>
            <link>https://systemweakness.com/eu-cyber-resilience-act-what-product-teams-should-do-now-a6734ab15136?source=rss-7eea262d9b94------2</link>
            <guid isPermaLink="false">https://medium.com/p/a6734ab15136</guid>
            <category><![CDATA[cyber-resilience-act]]></category>
            <category><![CDATA[product-security]]></category>
            <category><![CDATA[secure-by-design]]></category>
            <category><![CDATA[software-security]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <dc:creator><![CDATA[Frank Leitner]]></dc:creator>
            <pubDate>Fri, 27 Mar 2026 18:56:00 GMT</pubDate>
            <atom:updated>2026-03-31T23:03:42.887Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NqWHJI7_WHNeyW6ix7irXg.png" /></figure><p>The EU Cyber Resilience Act enforcement begins December 2027, less than two years away. For manufacturers of software products, the preparation window is closing rapidly. Yet many product organizations are still in the early stages of understanding what compliance actually requires.</p><p>The CRA fundamentally changes how software-driven products are brought to market in the EU. It mandates secure-by-design practices, vulnerability handling processes, and demonstrable security throughout the product lifecycle. For product teams, this is not a documentation exercise. It is a multi-stream engineering program: building management systems, and embedded control products, these aren’t minor adjustments, they require substantial technical implementation.</p><p>This article outlines a realistic 18-month preparation roadmap based on my experience helping product teams prepare for regulatory requirements. Even this timeline represents minimum viable preparation. Organizations with complex product portfolios, legacy systems, or immature security processes should expect to need significantly more time.</p><blockquote>CRA doesn’t ask you to improve your security. It asks you to prove it.</blockquote><h3>Timeline Overview</h3><p>Here’s what realistic CRA preparation looks like, starting today. The sections below explain what each phase actually involves. Starting in March 2026, this roadmap targets completion by September 2027, providing a three-month buffer before December 2027 enforcement. The timeline looks like this:</p><ul><li>Months 1–2 focus on scope assessment with the critical product inventory.</li><li>Months 2–4 handle gap analysis requiring cross-functional assessment.</li><li>Months 3–8 establish the vulnerability process including legal review and infrastructure.</li><li>Months 4–10 implement SBOM generation with automation tooling.</li><li>Months 5–12 enhance the SDLC through tool integration and training.</li><li>Months 10–14 create the documentation framework and templates.</li><li>Months 14–16 prepare for conformity assessment through internal audit.</li><li>Months 16–18 provide final preparation and buffer for unexpected issues.</li></ul><p>Many activities run in parallel. SBOM implementation overlaps with vulnerability process establishment. SDLC enhancement occurs alongside most other activities. However, dependencies exist: you can’t document processes that don’t exist yet, and automation requires process clarity.</p><blockquote>Starting today puts you on the critical path. Starting later removes your margin for error.</blockquote><h3>Understanding Your Product’s CRA Scope (Month 1–2)</h3><p>Before investing effort in compliance activities, you need clarity on which products actually fall under CRA scope. This seems straightforward but often isn’t.</p><p>The CRA applies to products with digital elements that connect to networks or other devices. For industrial product portfolios, this typically includes industrial automation controllers, building automation systems, embedded software in industrial equipment, IoT gateways and edge devices, and products with remote access or update capabilities.</p><p>However, gray areas exist. Pure cloud services have different treatment than products with digital elements. Component manufacturers face different obligations than system integrators. Products solely for internal use within a company may be out of scope. Critical products listed in CRA Annex III face additional third-party conformity assessment requirements.</p><p>For organizations with ten or more product lines, this assessment alone can take six to eight weeks. You need to examine each product’s architecture, connectivity, update mechanisms, and intended use. Don’t rush this step: misclassifying products wastes resources through unnecessary compliance work or dangerous non-compliance (and rushed last minute compensation attempts).</p><p>Create an inventory of your product portfolio with preliminary scope assessment and risk prioritization. Document your reasoning for scope decisions, particularly for borderline cases. This documentation becomes important if you later face questions from market surveillance authorities.</p><blockquote>Misclassifying products wastes resources or creates dangerous non-compliance.</blockquote><h3>Conduct a CRA-Focused Gap Analysis (Month 2–4)</h3><p>Once you know which products require CRA compliance, assess how far your current practices fall from the requirements. The CRA’s essential cybersecurity requirements in Annex I cover secure development processes, vulnerability handling, secure default configurations, protection against unauthorized access, secure update mechanisms, and logging and monitoring capabilities.</p><p>A thorough gap analysis examines your secure development lifecycle maturity, vulnerability management and disclosure processes, supply chain security and software bill of materials capabilities, and security documentation and conformity assessment readiness.</p><p>If you’ve already implemented IEC 62443 for industrial security, you have a head start. The standards overlap significantly in areas like security requirements specification, threat modeling, and secure development practices. However, CRA adds specific obligations around vulnerability disclosure timelines, SBOM provision, and end-of-support declarations that may not be covered by existing IEC 62443 implementations.</p><p>This assessment takes time because it requires input from multiple teams. You’ll need to interview development leads about current security practices, product managers about vulnerability handling, IT and DevOps about build systems and component tracking, and legal and compliance teams about documentation and disclosure processes.</p><p>Map your current processes against CRA Annex I requirements systematically. For each requirement, document current state, gap severity, implementation effort, and dependencies. Prioritize gaps by their impact on compliance and the effort required to close them. Some gaps are quick wins that build momentum. Others require months of infrastructure work and should start immediately.</p><blockquote>Some gaps are quick wins. Others require months of infrastructure work and should start immediately.</blockquote><h3>Establish Your Vulnerability Handling Process (Month 3–8)</h3><p>For many industrial product organizations, vulnerability handling represents the largest gap. Historically, industrial products operated in isolated networks with infrequent updates. The CRA requires a completely different approach, it is a regulated obligation with defined expectations around responsiveness and communication and no longer coluntary best practice.</p><p>You need a complete vulnerability handling process that includes intake mechanisms for vulnerability reports from external researchers, internal security teams, and automated scanning tools. Triage procedures must assess severity and exploitability, typically using CVSS scoring. You need coordinated disclosure timelines that meet CRA’s requirements, including acknowledging reports within a reasonable timeframe and providing remediation within appropriate periods based on severity.</p><p>The process must include patch management and update mechanisms that can deliver security fixes to deployed products, customer communication channels for security advisories and notifications, and a CSIRT or designated security contact point visible to researchers and customers.</p><p>Many organizations consider adopting CSAF (Common Security Advisory Framework) or VEX (Vulnerability Exploitability eXchange) formats early. These standardized formats facilitate automated vulnerability information exchange and are increasingly expected by enterprise customers.</p><p>This implementation requires significant infrastructure. You need ticketing systems to track vulnerability reports, secure communication channels for coordinated disclosure, advisory publication platforms (often integrated with product documentation sites), and potentially bug bounty or vulnerability rewards program infrastructure.</p><p>Legal review is essential. Your vulnerability disclosure policy must balance transparency with liability concerns. Your legal team needs to review disclosure timelines, communication templates, and coordination procedures with affected third parties.</p><p>Draft your vulnerability disclosure policy and publish it prominently. Assign clear roles for vulnerability intake, triage, remediation, and disclosure. Establish intake mechanisms such as security email addresses and web forms. Conduct dry-run scenarios with intentionally introduced test vulnerabilities to validate your process end-to-end before you need it for real incidents.</p><p>The five-month timeline for this work is realistic only if you have dedicated resources. Part-time effort easily extends this to six or eight months.</p><blockquote>Reality: Without intake, triage, remediation, and disclosure infrastructure, you don’t have compliance</blockquote><h3>Get Your SBOM House in Order (Month 4–10)</h3><p>Software bill of materials generation is consistently underestimated by organizations new to the practice. The CRA requires providing SBOMs for products with digital elements, enabling customers to understand supply chain risks and component vulnerabilities.</p><p>Legacy products often lack complete component visibility. Build systems may not systematically track all dependencies. Third-party commercial components may come without their own SBOMs, requiring negotiation with suppliers. Manual SBOM generation might work for a handful of products but doesn’t scale to continuous delivery of dozens of product variants.</p><p>Take a phased approach. Start with a pilot using one representative product. Manually identify all software components including open source libraries, commercial third-party components, internally developed modules, and system dependencies. Choose an SBOM format, SPDX and CycloneDX are the leading standards, each with different tooling ecosystems and strengths. Generate your first SBOM manually to understand the challenges, gaps in your component knowledge, and tooling requirements.</p><p>This manual pilot typically takes four to six weeks but provides invaluable learning. You’ll discover undocumented dependencies, components without clear version tracking, and third-party elements that require supplier engagement.</p><p>Following the pilot, establish tooling and automation. Research and select SBOM generation tools appropriate for your technology stack. Procure and configure these tools, which may require budget approval cycles. Integrate SBOM generation into your CI/CD pipelines so every build automatically produces an SBOM. Implement validation and quality checks to catch missing or incorrect component information.</p><p>This automation phase takes three to four months for most organizations. Build system integration requires dedicated development time. Different products may need different tooling approaches. Testing and validation across your product portfolio reveals edge cases and exceptions.</p><p>Generate your first SBOM manually, then build an automation roadmap based on lessons learned. Don’t assume vendor promises about automated SBOM generation will work without significant configuration and validation effort.</p><blockquote>Reality: SBOM automation is build system work, not documentation work.</blockquote><h3>Enhance Your Secure Development Lifecycle (Month 5–12)</h3><p>The CRA requires demonstrable secure-by-design practices throughout the product lifecycle. This goes beyond security as a final testing gate to security integrated into every development phase.</p><p>Key SDLC enhancements include security requirements in product specifications from the earliest planning stages, threat modeling integrated into design phase for new features and products, automated security testing in CI/CD pipelines through static analysis, dynamic analysis, and dependency scanning, security-focused code review practices and training, and security testing and validation before release with documented results.</p><p>Implementing these enhancements requires bridging security and development teams. Embed security requirements directly into existing sprint and development workflows rather than creating parallel security processes that teams will bypass. Conduct regular threat modeling sessions with architects and senior developers to build security thinking into design culture. Avoid positioning security as gatekeeping, frame it as risk-informed decision making that keeps products viable in regulated markets.</p><p>This cultural and process shift takes six months minimum. You need to select and implement security tooling, which requires evaluation, procurement, and integration. Teams need training on threat modeling, secure coding practices, and security testing interpretation. Initial implementations will be clumsy and slow: processes need refinement based on real experience. Tool configurations require tuning to reduce false positives while catching real issues.</p><p>Start by assessing current SDLC maturity against secure development frameworks like IEC 62443–4–1 or NIST SSDF. Select two or three priority enhancements with the highest compliance impact and feasibility. Pilot these enhancements with one team before attempting broader rollout. Document what works and what doesn’t, then refine your approach before scaling.</p><p>Trying to implement everything simultaneously across all teams will fail. Phased, iterative improvement succeeds where big-bang transformations collapse.</p><blockquote>Reality: Security integrated into every sprint beats a security gate at the end, every time.</blockquote><h3>Prepare Your Documentation Framework (Month 10–14)</h3><p>The CRA requires demonstrating compliance through technical documentation. You can’t retrofit this documentation at the end, it must reflect actual security activities throughout development.</p><p>Start documenting systematically now. Security risk assessments and threat models for each product should be documented with identified threats, risk ratings, and mitigation decisions. Secure SDLC evidence includes tools used, process descriptions, and training records showing teams are actually following secure practices. Vulnerability handling records should log every report, assessment, and resolution from the moment you establish the process. Third-party component analysis needs documented risk decisions about component selection, version choices, and known vulnerability acceptance. Security testing results from each release should be retained as evidence of verification activities.</p><p>Don’t wait until conformity assessment to discover you lack evidence of activities you actually performed but didn’t document.</p><p>Create reusable templates that scale across products. A threat model template ensures consistent analysis across different products. A security test report template standardizes evidence collection. A risk assessment template facilitates comparable decision documentation.</p><p>Establish clear documentation ownership and review cycles. Each product team should know who maintains security documentation, where it’s stored, when it’s reviewed and updated, and how it integrates with conformity assessment needs.</p><p>This documentation infrastructure takes three to four months to establish properly. Templates need creation and refinement, teams need training on their use, storage and retrieval systems need implementation, and initial documentation for existing products needs creation.</p><blockquote>Reality: Six months of undocumented development creates a six-month compliance gap.</blockquote><h3>Conformity Assessment Preparation (Month 14–16)</h3><p>With technical implementation largely complete, prepare for formal conformity assessment. Most products will follow self-assessment under Module A of the CRA. Critical products listed in Annex III require third-party assessment by notified bodies.</p><p>Conduct an internal readiness review. Systematically verify gap remediation for each item identified in your initial gap analysis. Check documentation completeness against CRA requirements. Conduct an internal audit or dry-run assessment simulating the rigor of external review.</p><p>This review invariably uncovers remaining issues. A process documented on paper may not be consistently followed in practice. Evidence may exist but not be easily retrievable. Technical implementations may not fully meet requirements upon close inspection.</p><p>If your product requires third-party assessment, begin notified body engagement early. Research and select a notified body with expertise in your product domain. Understand their specific requirements, timelines, and costs. Submit preliminary documentation to identify gaps before formal assessment.</p><p>Notified bodies are experiencing high demand as CRA enforcement approaches. Capacity constraints alone may become a bottleneck independent of your technical readiness. Waiting until month sixteen to initiate contact may mean you can’t secure assessment capacity before the deadline.</p><p>Complete your internal compliance review, address any remaining gaps, and initiate notified body discussions if applicable. Budget both time and money for this phase, external assessments are neither quick nor inexpensive.</p><blockquote>Notified bodies won’t wait for you. Neither will the deadline.</blockquote><h3>Final Preparation and Buffer (Month 16–18)</h3><p>The final months before your target completion date focus on conformity documentation finalization and preparing for market requirements.</p><p>Prepare your EU Declaration of Conformity, the formal statement that your product meets CRA requirements. Ensure CE marking readiness with proper labeling and documentation. Plan customer communications about your CRA compliance, security update commitments, and vulnerability disclosure process. Complete final documentation review, ensuring everything is current, accurate, and accessible.</p><p>Critically, maintain buffer time. Unexpected issues always emerge in final stages. A supplier might not provide needed component documentation. A notified body might request additional evidence. An internal audit might reveal a process gap. Market surveillance authorities might publish new guidance requiring adjustments.</p><p>Organizations that plan to finish exactly at the deadline will miss it. Those who plan to finish months early might actually finish on time.</p><blockquote>Plan to finish months early. You might actually finish on time.</blockquote><h3>Why This Timeline Is Still Aggressive</h3><p>This eighteen-month timeline assumes dedicated resources. If your security team handles CRA preparation alongside other responsibilities, or if development teams contribute only part-time, expect significant timeline extension. Full-time equivalent resource allocation matters enormously.</p><p>The timeline assumes moderate complexity. Organizations with dozens of legacy products, distributed development teams across multiple sites, or very immature security processes need substantially more time. Products with complex supply chains or novel technologies face additional challenges.</p><p>Parallel activities have dependencies that constrain true parallelization. SBOM automation requires SDLC process clarity about build systems and component management. Documentation requires completed implementations to document. Conformity assessment requires everything else to be finished.</p><p>Many organizations will need twenty-four months or more for comprehensive preparation. If you’re reading this in 2026 and haven’t begun the process, you’re already on the critical path with no room for delays.</p><blockquote>Reality: Part-time CRA preparation delivers part-time results, with a fixed deadline.</blockquote><h3>Common Pitfalls to Avoid</h3><p>Don’t treat CRA purely as a legal or compliance exercise, and leadership cannot outsource this to legal departments or certification bodies. It requires significant technical implementation that legal teams can’t accomplish alone. Engineering, product management, and security teams must drive the technical work.</p><p>Don’t underestimate SBOM automation effort. Manual approaches might work for a few products but don’t scale to continuous delivery of multiple product variants. Plan for substantial build system integration work.</p><p>Identify what can proceed in parallel versus what has hard dependencies. Sequential execution of every step extends your timeline beyond the available window. Some tasks must finish before others begin, but many can overlap with proper planning.</p><p>If you’ve already implemented IEC 62443, leverage that work. The standards overlap significantly in secure development, threat modeling, and security architecture. Don’t duplicate effort, but also recognize CRA adds specific requirements around vulnerability disclosure and SBOM that may not be covered.</p><p>Don’t wait for final clarity on every regulatory detail. Core obligations are clear enough to start implementation. Using uncertainty as an excuse for delay means you’ll still lack clarity but now also lack time.</p><p>Finally, ensure sufficient resource allocation. This work requires dedicated time from development, security, and product management teams. Adding CRA preparation on top of existing full workloads guarantees both missed deadlines and poor quality outcomes.</p><blockquote>Reality: CRA is a technical program. Not a compliance checkbox.</blockquote><h3>Moving Forward</h3><p>CRA readiness is achievable with structured, realistic preparation — but it requires starting immediately. This eighteen-month timeline represents minimum viable preparation for organizations beginning today. Don’t underestimate the effort required.</p><p>Organizations starting now can meet the December 2027 deadline, but delays compound rapidly. A two-month slip in gap analysis delays everything that follows. Infrastructure procurement delays ripple through dependent activities. Resource constraints stretch every phase beyond initial estimates.</p><p>If you’re unsure whether your current roadmap realistically meets December 2027, a focused gap assessment now is significantly cheaper than late remediation in 2027. Reach out for a realistic assessment of your specific timeline needs.</p><p>The teams that will succeed are those that treat CRA preparation as a significant technical program requiring dedicated resources, executive support, and realistic planning. Those that treat it as a compliance checkbox to address later will find themselves scrambling in late 2027 with insufficient time to implement required technical controls.</p><blockquote>The December 2027 deadline is fixed, your preparation timeline should reflect that reality.</blockquote><p>I consult on CRA and IEC 62443 implementation for manufacturers of software-driven products. You can reach me on <a href="https://www.linkedin.com/in/frankleitner/">LinkedIn </a>or my <a href="https://www.frank-leitner.com/">website</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a6734ab15136" width="1" height="1" alt=""><hr><p><a href="https://systemweakness.com/eu-cyber-resilience-act-what-product-teams-should-do-now-a6734ab15136">EU Cyber Resilience Act: What Product Teams Should Do Now</a> was originally published in <a href="https://systemweakness.com/">System Weakness</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[CRA Presumption of Conformity: What It Means and Why IEC 62443 Is the Right Implementation Path]]></title>
            <link>https://meetcyber.net/cra-presumption-of-conformity-what-it-means-and-why-iec-62443-is-the-right-implementation-path-b6795d49b29e?source=rss-7eea262d9b94------2</link>
            <guid isPermaLink="false">https://medium.com/p/b6795d49b29e</guid>
            <category><![CDATA[software-security]]></category>
            <category><![CDATA[iec-62443]]></category>
            <category><![CDATA[cyber-resilience-act]]></category>
            <category><![CDATA[product-security]]></category>
            <category><![CDATA[cybersecurity]]></category>
            <dc:creator><![CDATA[Frank Leitner]]></dc:creator>
            <pubDate>Thu, 26 Mar 2026 11:31:01 GMT</pubDate>
            <atom:updated>2026-03-31T23:04:33.210Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*154PGNgB4vSVF6Zf" /></figure><p>The EU Cyber Resilience Act tells manufacturers what they must achieve. It does not tell them how.</p><p>Annex I sets out the essential cybersecurity requirements. Products must be designed and developed securely. Vulnerabilities must be handled systematically. Updates must be delivered. Personal data must be protected. The obligations are clear in principle. What remains open is the question every product security team eventually arrives at: what gives us the most defensible path to demonstrating we actually meet them?</p><p>The CRA’s answer is presumption of conformity. It is worth understanding precisely, because it is frequently misread in both directions, either overstated as a broadly available compliance shortcut, or dismissed as irrelevant to implementation strategy. Neither reading is accurate.</p><h3>What Presumption of Conformity Actually Means</h3><p>The CRA, like most EU product regulation, operates on a presumption mechanism tied to harmonized standards. When a manufacturer implements a harmonized standard that has been formally published in the Official Journal of the EU, and that standard covers the relevant essential requirements, the manufacturer is presumed to be in conformity with those requirements.</p><p>The practical effect is a shift in the burden of proof. Without a harmonized standard, a manufacturer must construct an independent case for how each Annex I requirement is met, from first principles, in a form that holds up under conformity assessment. With a harmonized standard, that work has already been done at the standard’s level. Implementing the standard is itself the evidence.</p><p>One important clarification: presumption of conformity is not a guarantee of compliance. A manufacturer can implement a harmonized standard and still fall short of Annex I if the actual practice doesn’t match documented processes, or if the standard doesn’t cover a particular requirement. The presumption shifts the burden. It does not eliminate the need for genuine implementation.</p><blockquote>The standard does the evidentiary work. You do the implementation.</blockquote><h3>How Conformity Assessment Works in Practice</h3><p>Understanding what presumption of conformity delivers requires understanding what a conformity assessment actually involves.</p><p>Most products will follow self-assessment under Module A of the CRA. The manufacturer conducts its own assessment, compiles a technical file demonstrating how the product meets Annex I requirements, issues an EU Declaration of Conformity, and applies CE marking. No external auditor is involved. But “self-assessment” does not mean unchecked. Market surveillance authorities can request the technical file at any point, and that file needs to hold up under scrutiny.</p><p>Products listed in Annex III require mandatory third-party assessment by a notified body. These are the higher-risk critical product categories, e.g. firewalls, VPNs, routers, SIEMs, and others. For these, a notified body reviews the technical documentation, assesses the conformity evidence, and issues a certificate. Notified bodies apply their own rigor, and capacity is already constrained as the December 2027 deadline approaches.</p><p>In both cases, the core challenge is the same: producing credible, traceable evidence that security was engineered into the product and maintained across its lifecycle. A recognized standard provides the structure that makes that evidence coherent rather than ad hoc.</p><blockquote>Self-assessment is not a free pass. It is an unsupervised exam with open-book results.</blockquote><h3>The Current Gap: No Formally Designated Harmonized Standards Yet</h3><p>Here is where the practical picture becomes important to understand clearly.</p><p>On 3 February 2025, the European Commission officially issued Standardization Request M/606 to CEN, CENELEC, and ETSI. The request was accepted in April 2025. The work is structured across four standard types: horizontal framework standards, vulnerability handling standards, technical measures standards, and product-category-specific vertical standards for Annex III product categories.</p><p>Current planning targets publication between August and October 2026 for the first deliverables, with the full set expected by October 2027, shortly before the CRA enters full application. The work is active and progressing on a defined standardization track. But no harmonized standards have yet been formally published in the Official Journal under the CRA.</p><p>This means the presumption of conformity mechanism is currently available in principle and unavailable in practice. Manufacturers implementing IEC 62443 today cannot formally invoke it. That matters for how the current investment should be understood, but it does not change the strategic calculation, as the next sections explain.</p><blockquote>The mechanism exists. For most manufacturers, the path to invoking it is narrower than it appeared.</blockquote><h3>Why IEC 62443 Is the Right Path Anyway</h3><p>The Commission has confirmed it will not cite the horizontal 62443 standards broadly for presumption of conformity. For manufacturers outside the product categories covered by vertical standards, that closes the formal presumption path. It does not change the implementation case.</p><p>The substantive overlap between 62443 and CRA Annex I is structural, not contingent on citation status. IEC 62443–4–1 maps directly to how the CRA expects products to be developed. IEC 62443–4–2 maps directly to what the CRA expects products to do. That overlap does not disappear because the horizontal presumption path is unavailable. What it means in practice is that manufacturers implementing 62443 must construct their compliance case explicitly, mapping their implementation to Annex I requirements with traceable evidence, rather than relying on a standard citation to do that work for them. The evidentiary burden is higher. The technical foundation is the same.</p><p>IEC 62443–4–1 defines a secure development lifecycle, including structured practices for security requirements management, threat and risk analysis, secure design, implementation guidance, verification activities, and vulnerability management. This maps to the CRA’s core expectations for how products are developed: security is not added at the end, it is engineered throughout.</p><p>The CRA’s vulnerability handling obligations, covering coordinated disclosure processes, patch delivery timelines, and the need for systematic intake and triage, map to the vulnerability management practice group in 62443–4–1. The standard requires documented processes for tracking and remediating vulnerabilities across the product lifecycle, including defined processes for working with external researchers, which directly aligns with what the CRA expects.</p><p>IEC 62443–4–2 defines technical security requirements at the component level: authentication controls, communication integrity and confidentiality, protection against unauthorized access, and requirements for update integrity verification. These map to the CRA’s product-level security requirements in Annex I, part I, which specify that products must resist unauthorized access, protect data in transit and at rest, and deliver authenticated updates.</p><p>This overlap is substantial. For organizations that have already implemented 62443, the CRA does not require starting over. It requires extending and adapting.</p><blockquote>IEC 62443 doesn’t describe security. It engineers it.</blockquote><h3>Where IEC 62443 Currently Falls Short</h3><p>Stating the overlap accurately requires stating the gaps equally accurately. There are four areas where the current 62443 standard does not fully cover CRA obligations, and where manufacturers need to build separate compliance tracks.</p><h4>SBOM provision</h4><p>The CRA requires manufacturers to identify and manage software components and dependencies across the product lifecycle. In practice, this makes SBOM generation the most viable implementation approach. For many products, that means maintaining a machine-readable inventory of components, including third-party software. IEC 62443–4–1 addresses supply chain security in principle and recommends practices for managing third-party components, but it does not mandate structured SBOM generation and disclosure as a required product deliverable. For many industrial product organizations, this represents the largest practical implementation gap.</p><h4>Mandatory reporting timelines</h4><p>The CRA requires manufacturers to report actively exploited vulnerabilities or severe incidents to ENISA within 24 hours of becoming aware of them, with a follow-up report within 72 hours. This specific regulatory notification obligation has no direct counterpart in 62443–4–1. The standard addresses internal vulnerability management process maturity, but it does not define legally bound external reporting timelines to regulatory authorities. These are fundamentally different obligations, and the reporting infrastructure required to meet them needs to be built independently.</p><h4>End-of-support declarations</h4><p>The CRA requires manufacturers to define and communicate a minimum security support period for each product, to deliver security updates throughout that period, and to notify users explicitly when the support period ends. IEC 62443 addresses lifecycle security in a general sense, but it does not require this kind of formal, market-facing commitment as a compliance deliverable.</p><h4>Privacy and data protection</h4><p>CRA Annex I requires that products minimize the processing of personal data, protect personal data by default, and prevent unauthorized access to personal data stored or transmitted by the product.</p><p>IEC 62443 was not designed for that problem space. Its confidentiality requirements are rooted in industrial operations, not GDPR-style privacy engineering. For products that handle personal data, this creates a separate compliance track that 62443 alone does not cover.</p><blockquote>Knowing where a standard ends is as important as knowing where it begins.</blockquote><h3>What the EN Adaptation Is Doing to Close These Gaps</h3><p>The standardization work currently underway is adapting the 62443 series into harmonized European standards, with two distinct outcomes. The horizontal amendments to 4–1 and 4–2 will strengthen the technical and lifecycle alignment with CRA requirements but will not provide presumption of conformity. The vertical standards for specific Annex III product categories will. Both tracks are in active development.</p><p>EN IEC 62443–4–2:2019/A11:2026 is the primary vehicle on the technical side. The amendment work includes adapting existing component requirements to the CRA framework and introducing targeted enhancements to close identified Annex I gaps. It also defines applicability criteria, aligns the evaluation approach with expected conformity assessment artefacts, and specifies Security Level-based acceptance criteria for each requirement. This last point connects directly to the Applicability Assessment Requirements approach introduced in prAA, which shifts the framework from declaring a security level to individually justifying each requirement decision. The planned publication target for this amendment is October 2026.</p><p>EN IEC 62443–4–1:2018/A11:2026 runs in parallel on the lifecycle side. This amendment adds requirements around documenting intended use and security context, and further specifies the expected development artefacts that demonstrate a compliant lifecycle process has been followed. This directly addresses the CRA’s demand for traceable evidence that the product was designed securely from the start, rather than asserted after the fact.</p><p>In addition, six vertical standards are planned for specific Annex III product categories, including VPNs, routers, SIEMs, and firewalls. These will provide product-specific technical guidance where the horizontal amendments leave room for interpretation, and are expected by October 2026.</p><p>One important caveat: the amendment work is clearly directional, and the scope is explicitly aligned with closing CRA gaps. But the degree to which the published amendments will fully address the SBOM, mandatory reporting, end-of-support, and privacy gaps in their final form remains to be confirmed once the text is published. The privacy gap in particular sits outside the natural scope of what a 62443-derived standard would address, and manufacturers should not assume it will be covered.</p><blockquote>The direction is clear. The completeness is not yet confirmed.</blockquote><h3>What Manufacturers Should Do Now</h3><p>Three actions follow directly from this analysis.</p><h4>1. Implement IEC 62443–4–1 and 4–2 now.</h4><p>Not because horizontal presumption of conformity is coming, but because it structures your evidence base in a way that maps substantively to CRA Annex I, and already provides the most credible and auditable compliance path available today. If vertical standards covering your product category are cited, a well-documented 62443 implementation positions you to claim formal presumption with minimal additional work. If they are not, your 62443 implementation is still the strongest foundation for constructing a traceable self-assessment case. Either way, the implementation investment is not at risk.</p><h4>2. Address the four gaps as dedicated workstreams</h4><p>SBOM infrastructure, ENISA reporting processes, end-of-support policy documentation, and privacy engineering are not optional CRA obligations that harmonization will eventually solve for you. They need dedicated implementation now, regardless of where the EN adaptation lands. Build them alongside your 62443 work, not after it.</p><h4>3. Explicitly map IEC 62443 implementation to CRA Annex I</h4><p>Do not assume auditors or notified bodies will connect the dots. For each Annex I requirement your 62443 implementation addresses, document how, with traceable references to your actual processes and artefacts. For manufacturers covered by vertical standards, a well-structured 62443 implementation means the transition to a formal presumption position should be a documentation update, not a technical re-assessment. For everyone else, that same documentation discipline is what makes a self-assessment case hold up under scrutiny.</p><blockquote>Waiting for certainty is a decision. It just isn’t a good one.</blockquote><h3>Closing</h3><p>Presumption of conformity is not a compliance shortcut. It is a legal mechanism that rewards manufacturers who invested in structured, standards-based implementation before the regulatory deadline arrived.</p><p>The harmonization work continues, and vertical standards for specific Annex III product categories remain on track for citation. The horizontal presumption path for 62443–4–1 and 4–2 has been ruled out, and manufacturers should plan accordingly. Organizations that read the current absence of formally designated standards as a reason to wait are misreading the situation.</p><p>For manufacturers covered by vertical standards, that window is closing. For everyone else, the question is not when formal presumption will arrive. It is how well your evidence base holds up without it.</p><blockquote>Presumption of conformity is not a shortcut for those who waited. It rewards early structure.</blockquote><p>I consult on CRA and IEC 62443 implementation for manufacturers of software-driven products. You can reach me on <a href="https://www.linkedin.com/in/frankleitner/">LinkedIn </a>or my <a href="https://www.frank-leitner.com/">website</a>.</p><p><em>Originally published at </em><a href="https://www.linkedin.com/pulse/cra-presumption-conformity-what-means-why-iec-62443-right-leitner-cxrbf/"><em>https://www.linkedin.com</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b6795d49b29e" width="1" height="1" alt=""><hr><p><a href="https://meetcyber.net/cra-presumption-of-conformity-what-it-means-and-why-iec-62443-is-the-right-implementation-path-b6795d49b29e">CRA Presumption of Conformity: What It Means and Why IEC 62443 Is the Right Implementation Path</a> was originally published in <a href="https://meetcyber.net/">MeetCyber</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>