Cyber resilience testing checks whether an organization can respond to and recover from a severe but plausible disruption, then requires evidence of the result. Penetration testing asks whether an attacker can get in. Resilience testing assumes the attacker has already succeeded and validates response and recovery rather than prevention.
What Is Cyber Resilience Testing?
Cyber resilience testing is a structured exercise that assumes a serious disruption has already happened. It checks, against objectives set in advance, whether people, systems, and business decisions restore critical services. The output records whether each objective was met and attaches the supporting evidence. Reflex Security publishes this article and builds crisis simulations for the people layer of resilience testing; it does not provide technical recovery testing.
The clearest public example of cyber resilience testing is the European Central Bank's 2024 cyber resilience stress test. Its scenario assumed that every preventive measure had failed and that a cyberattack had severely affected the databases of each bank's core systems. The ECB therefore focused on how banks would respond and recover, "rather than on how they would prevent it."
The test covered 109 banks under direct ECB supervision. All of them answered a questionnaire and submitted documentation. A sample of 28 went further: each performed an actual IT recovery test, provided evidence that it had succeeded, and received supervisors on site. The results fed the ECB's 2024 Supervisory Review and Evaluation Process.
The ECB stress test applied to participating banks under its direct supervision; it did not create a general obligation for other organizations. This article borrows the design without extending the mandate and applies the pattern to a cryptocurrency exchange. An EU crypto-asset service provider authorized under MiCA already carries testing duties under DORA. For that firm, the ECB design provides a useful structure rather than an additional rule.
Why CISOs Need to Test Resilience, Not Just Measure It
Measurement describes expected capability, while testing observes performance against a defined scenario.
Although a maturity score, control-coverage percentage, or recovery time objective remains useful, none demonstrates that a service can be restored under test conditions. For example, a four-hour recovery objective becomes observed evidence only when the organization restores the service within that period and records the result.
The distinction matters when a CISO signs an insurer’s questionnaire, a client’s due-diligence response, or a board risk report. Assessments document the expected capability; tests add observed results and timestamps. Cyber Resilience Assessment: How to Measure Your Readiness explains where an assessment ends and testing begins.
What Should a Cyber Resilience Test Actually Validate?
A cyber resilience test should validate outcomes that no document can prove on its own. Seven targets cover most of the ground. Several echo the expectations the ECB restated after its test: banks should be able to meet their own recovery objectives, assess dependencies on critical third-party ICT providers, and estimate the direct and indirect losses of an attack.
Recovery against your own objectives. Start the clock when the incident is declared. Stop it when a business owner confirms the service works. A restore job finishing is only an intermediate event.
Restore order of critical services. The team should bring services back in the order the business ranked them. Engineers under pressure may restore what is easiest first. Disaster Recovery Tabletop Exercise: A Practical Guide for 2026 covers recovery sequencing; the resilience test checks whether the agreed order held.
Third-party and critical-provider dependencies. Can the team reach the provider through the contracted channel? Does the provider's actual response match what the runbook assumes?
Decision rights and escalation. Record who declared the incident, who authorized each irreversible action, and how long each step waited for a name.
Customer and regulator communication. Measure the time to the first external statement, note who approved it, and check whether support, social, and regulatory channels said the same thing.
Loss estimation. Finance should produce a first estimate of direct and indirect losses while the event is still running. That number drives most of the business decisions.
Corrective-action follow-through. Findings from the last test should be closed. The ECB likewise asked banks to show they could implement lessons learned.
Each target needs a pass condition that an observer can check. Define it before the test begins so the result is evaluated against an agreed threshold rather than criteria created afterward.
Cyber Resilience Testing vs. Penetration Testing vs. Tabletop Exercises
Penetration tests, tabletop exercises, and resilience tests answer three different questions. Most security programs already run the first two, so a yearly penetration test plus an annual tabletop may appear to provide resilience testing. Neither method, however, requires the organization to restore a service against a predefined objective.
Tabletops usually fill the middle of that program as discussion-based cyber resilience exercises involving a scenario, a facilitator, and a group of decision-makers. What Is a Tabletop Exercise? Cybersecurity Guide for 2026 covers the format in full. The table compares the three methods by what each one proves.
Method | Question answered | What is attacked or exercised | Who participates | Evidence produced | Main blind spot |
Penetration test | Can an attacker get in, and how far? | Infrastructure, applications, and controls | External or internal testers and security engineers | Findings report of exploited weaknesses | Stops at compromise. Says nothing about recovery or business decisions |
Tabletop exercise | Does the team understand the plan? | The plan, through discussion | Cross-functional leads and a facilitator | Notes and an after-action summary | Nothing is restored, so stated recovery capabilities go unchecked |
Cyber resilience test | Can the organization respond and recover within its own objectives? | Recovery systems, people, and business decisions together | Technical, operations, executive, and provider staff | Objectives marked met or not met, with logs and timestamps | Requires coordinated technical and business participation. A technical-only version misses the people layer |
The three methods are complementary. Penetration testing shows whether technical controls withstand a skilled attacker, while resilience testing begins from the assumption that prevention failed. Tabletop exercises can expose plan gaps before a full recovery test. For a tired exercise program, improving its scenarios and facilitation model is usually more useful than dropping the format.
Between the tabletop and the full recovery test sits a fourth format: simulation of the response itself. Reflex Security runs this kind of exercise, with an adaptive adversary and pressured stakeholders reacting to team decisions. It shares the tabletop’s participants and produces a timestamped exercise record, but it restores no systems. It therefore runs alongside a technical recovery test. The difference between structured templates and an adaptive platform is covered separately.
How to Stress-Test People, Technology & Business Decisions Together
Cyber resilience stress testing should place technology, people, and business decisions under one scenario because real incidents affect all three layers. The three hypothetical designs below apply to a cryptocurrency exchange. Every number is a parameter for the organization to set, not a benchmark. Each design follows the same chain: objective, method, evidence, and pass criterion.
The usual objection is that stress-testing a live trading venue is too risky. Most of the test need not touch production. Where risk permits, selected technical recovery steps can use production during an approved maintenance window. The remaining technical work can run against a production-equivalent replica, while the people and business layers use simulated injects in real time. A controlled production restore can then reduce the remaining fidelity gap. The evidence pack should identify which environment produced each result. Simulating a live crisis covers the injected side in more depth.
Test Design One: Hot-Wallet Compromise During Peak Volume
Monitoring flags unauthorized signing activity on a hot wallet while trading volume runs at three times its 30-day average. Engineers work the technical layer on a replica of the wallet and signing infrastructure. The designated exercise participant is paged through the real paging tool under an agreed test protocol. Executives get the halt-or-continue question with the partial information they would actually hold at that minute. The people and business rows can run as a Reflex simulation while engineers work the replica.
Test Design Two: Custody Provider Outage
The exchange's third-party custody provider stops answering withdrawal requests for one asset, with no explanation. Nothing inside the exchange is compromised. That is deliberate, because the whole test is about a dependency. The team escalates through the contacts named in the contract. Treasury reconciles balances it can no longer confirm. Leadership decides whether to suspend the asset and what to tell the customers who hold it.
Test Design Three: Social-Media Panic Plus Withdrawal Surge
A false claim that the exchange has been hacked spreads on social media, and withdrawal requests climb to five times normal within the hour. No system is down. Communications, support, and treasury carry the scenario between them. The technical layer only has to show that withdrawals keep processing within published limits. The real risk is a public statement that contradicts what support agents tell customers in parallel.
Test design | Layer | Objective | Evidence | Pass criterion |
Hot-wallet compromise | Technology | Freeze hot-wallet withdrawals and move signing to a clean path | Freeze and signing logs with timestamps | Freeze applied within 10 minutes of detection; no withdrawal signed after it |
Hot-wallet compromise | People | Escalate to a named incident commander | Paging acknowledgements, bridge transcript | Commander on the bridge within 15 minutes |
Hot-wallet compromise | Business | Decide whether to halt trading | Decision log: name, time, information available | Decision made by someone holding that authority, within 30 minutes, rationale recorded |
Custody provider outage | Technology | Reconcile affected balances | Reconciliation report | Reconciliation complete before any resumption |
Custody provider outage | People | Reach the provider through contracted channels | Call records against contract terms | Provider reached within its contracted response time, or the miss escalated and logged |
Custody provider outage | Business | Suspend or keep the asset live, then notify holders | Suspension decision, customer notice | Notice published within 60 minutes of the decision |
Social-media panic | Technology | Keep withdrawals within published limits | Queue and processing logs | No withdrawal held past published limits without an announced change |
Social-media panic | People | Align support with the public statement | Statement drafts, support scripts, timestamps | Support scripts match the statement within 30 minutes of its release |
Social-media panic | Business | Decide on liquidity measures | Treasury decision log | Any limit change approved by a named owner and announced before enforcement |
Every pass criterion in the table describes an outcome an observer can verify. This allows the exercise to produce a defensible result rather than relying on participants’ descriptions of what they would do.
Turning Cyber Resilience Exercises Into Measurable Evidence
The logs from a combined test become usable evidence only when they are organized into a record an outside reviewer can check. Three disciplines create that record.
A verdict on every objective. The evidence pack marks each objective met or not met. It records a near miss as not met and notes the margin, keeping the assessment tied to the pass condition established before the test. The relevant timestamp, transcript excerpt, or restore log line sits beside the verdict so a reviewer can confirm it independently.
Remediation tracked to closure. Every not-met result becomes an item with an owner, a due date, and a retest. The item closes only when a later test shows that the objective was met. Completing the ticket without a successful retest does not demonstrate that the underlying gap was resolved.
A pack built for its readers. An auditor, an insurer, an institutional client running due diligence, and a regulator will each read it. For firms within its scope, DORA resilience testing adds requirements on top of this structure.
A cyber resilience testing framework links each objective to evidence and remediation, then repeats the cycle with the next scenario. Cyber Resilience Metrics: How to Measure What Actually Matters explains how to compare results across controlled runs.
From Technical Stress Tests to the People Pen Test
A network penetration test probes technical controls. The people pen test applies pressure to the response process and observes how the team performs. Both produce findings about weaknesses, but only the penetration test targets technical infrastructure.
Reflex Security’s homepage puts it directly: “You already run technical pen tests for your infrastructure. Reflex is the ‘people pen test’ for your team.” Inside a combined program, that is the layer Reflex fills. Penetration tests continue testing controls, while technical recovery tests prove restores. A Reflex simulation tests the decisions between them, with AI agents representing the adversary and stakeholders and reacting to each call. Its After Action Report ties every finding to the moment in the simulation that produced it. Cyber Crisis Simulation: The CISO’s Guide to the Next Generation of Cyber Exercises explains how that evidence is captured.
At the exchange above, the hot-wallet and social-media designs carry most of their risk in the people and business rows. A technical test cannot evaluate those decisions. Select one scenario, mark each objective met or not met, and compare Reflex with traditional tabletop formats before choosing an approach.
See what your team does under real pressure.
Put your readiness to the test. Sixty minutes, your real environment, no slides.
Book A demo
Frequently Asked Questions
How often should an organization run cyber resilience testing?
No regulator sets one cadence for every sector. The ECB's cyber test was the targeted topic of its 2024 supervisory stress test, and earlier targeted years covered liquidity and climate risk instead. A workable pattern is one full combined test a year, another after any material change such as a new custody provider, and shorter people-layer exercises in between.
Is cyber resilience testing required by regulation?
For some organizations, yes. ECB-supervised banks took part in the 2024 stress test, and UK regulators use CBEST on systemically important firms. In the EU, DORA Article 11(6) requires in-scope financial entities, including MiCA-authorized crypto-asset service providers, to test their response and recovery plans at least yearly. Elsewhere, obligations depend on license and jurisdiction, so confirm them with counsel.
What does "severe but plausible" mean in a resilience test scenario?
The ECB used the phrase to describe its 2024 test, which gauged how banks would respond to and recover from a severe but plausible cybersecurity incident. Severe means the scenario defeats preventive controls: the ECB assumed all of them failed. Plausible means it could happen to this organization, given its real threats and dependencies, rather than as a cinematic worst case.
Can a cyber resilience test be run without disrupting production systems?
Mostly, yes. Technical recovery can run against a production-equivalent replica or inside a maintenance window. The people and business layers can run on simulated injects, through a facilitated format or a platform such as Reflex. Using a replica reduces operational risk, but the evidence becomes less representative as the test environment diverges from production. The evidence pack should therefore state which environment produced each result.
What is CBEST and how does it relate to resilience testing?
CBEST is the Bank of England's intelligence-led penetration testing framework, part of the supervisory toolkit it shares with the PRA and FCA since 2014. Testers mimic real attackers targeting a firm's important business services, and the assessment also examines detection and response. It complements a recovery test: CBEST asks how far an informed attacker gets, while an ECB-style test assumes the attacker succeeded.
Who should sign off on the results of a cyber resilience test?
The executive who owns the affected business service, not the team that ran the test. The CISO typically owns the design and the evidence, while a second-line risk function or internal audit confirms each met or not-met verdict. Keeping the person who ran the test apart from the person who accepts it stops a near miss from being filed as a pass.
References
Bank of England. (2024). CBEST threat intelligence-led assessments: Implementation guide for CBEST participants (2024 ed.). Bank of England. https://www.bankofengland.co.uk/financial-stability/operational-resilience-of-the-financial-sector/cbest-threat-intelligence-led-assessments-implementation-guide (Page last updated August 25, 2026.)
European Central Bank. (2024, July 26). ECB concludes cyber resilience stress test [Press release]. ECB Banking Supervision. https://www.bankingsupervision.europa.eu/press/pr/date/2024/html/ssm.pr240726~06d5776a02.en.html
European Parliament & Council of the European Union. (2022, December 14). Regulation (EU) 2022/2554 of the European Parliament and of the Council of 14 December 2022 on digital operational resilience for the financial sector. Official Journal of the European Union. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32022R2554
