// guides

Penetration Testing for SOC 2, ISO 27001 and DORA Audits

Penetration testing SOC 2, ISO 27001 and DORA audit evidence mapped to a single engagement scope

Penetration testing for SOC 2 is not mandatory by name, ISO 27001 sets no test frequency, and DORA mandates two different kinds of test. Auditors in all three regimes still sample the same four things: when the test ran, what it covered, who ran it, and proof that findings were fixed or formally accepted. That overlap lets one well-scoped engagement serve all three audits. TLPT is the exception.

The short answer

  • SOC 2 does not mandate a penetration test, but Type II auditors expect one dated inside the review period (CC4.1, CC7.1).
  • ISO 27001:2022 sets no frequency; Annex A 8.8, 8.29 and 5.35 accept pentest evidence, and your own policy sets the bar.
  • DORA requires yearly testing of critical-function systems and TLPT every three years for designated entities.
  • One scoped engagement with a contracted retest can satisfy SOC 2, ISO 27001 and DORA annual testing; TLPT stays separate.

Key takeaways

  • SOC 2 never mandates a pentest by name, but CC4.1 lists it as a separate evaluation and Type II auditors expect one dated inside the review period.
  • ISO 27001:2022 sets no test frequency; Annex A 8.8, 8.29 and 5.35 accept pentest evidence, and auditors hold you to the frequency your own policy states.
  • DORA requires yearly testing of systems supporting critical or important functions, plus TLPT at least every three years for entities their authority designates.
  • Auditors accept retest letters, linked change records or signed risk acceptance with a compensating control, not a status column.
  • One well-scoped engagement covers SOC 2, ISO 27001 and DORA yearly testing; TLPT is always a separate exercise.

Is penetration testing required for SOC 2?

No. The AICPA Trust Services Criteria never require a penetration test by name, yet almost every SOC 2 Type II auditor expects one. The expectation comes from criteria that are hard to satisfy any other way.

  • CC4.1 requires ongoing and separate evaluations of controls. Its points of focus name penetration testing as an example of a separate evaluation.
  • CC7.1 requires procedures that detect newly discovered vulnerabilities and configuration changes that introduce weakness. Scanning covers part of this. Manual exploitation testing covers the rest.
  • CC3.2 requires risk identification and analysis. A pentest report is one of the few artifacts that shows risk that has been demonstrated, not merely assumed.

The test also produces evidence for CC7.2, which covers monitoring for anomalies. Whether the test set off your alerts (or never did) tells the auditor whether detection works.

Points of focus are guidance, not requirements, so in theory you can present equivalent separate evaluations instead. In practice, a SaaS vendor that offers only automated scans for a Type II gets questions from the auditor, and harder ones from enterprise customers reading the report.

How often do you need a penetration test for SOC 2 Type II?

At least once inside each review period, which for most companies means annually. A Type II report covers a window, usually 6 to 12 months, and the auditor tests whether controls operated throughout it. A test dated before the window opens is not evidence for that window. Type I assesses design at a single point in time, so a test from the prior 12 months usually suffices.

Three timing mistakes come up again and again:

  • Testing in the last two weeks of the period. The report lands with open findings and no time to fix or retest. The auditor sees a control that found problems and no evidence of a response.
  • Ignoring a mid-period release. If you shipped a new API, region or auth flow after the test, the report no longer describes the system in scope. Significant change triggers retesting, not just the annual date.
  • Letting the date drift. If the period runs January to December and last year’s test was in November, keep this year’s near the same point so coverage stays continuous.

Schedule the test in the first half of the period, so remediation and retest evidence fit inside it.

What evidence does a SOC 2 auditor need from a penetration test?

A dated report showing scope, method, tester identity, findings with severity, and proof that each finding was handled. Expect the auditor to sample the full report, not the executive summary. A report that survives sampling contains:

  • Test dates inside the review period.
  • A scope statement that maps to the system description in the SOC 2 report: the same applications, environments, cloud accounts and APIs.
  • Methodology (OWASP, PTES, NIST SP 800-115 or equivalent) and approach: black, grey or white box, and which authenticated roles were used.
  • The tester’s organization and a statement of independence from the team that builds and runs the system.
  • Findings with severity, affected assets and reproduction detail.
  • Remediation tracking: ticket references, fix dates, retest results or risk acceptance.

The most common gap is scope mismatch. The system description says “customer-facing platform, APIs and AWS production”, and the pentest covered the marketing site and the login page. An auditor who reads both documents will flag it.

Write the pentest scope from the system description, not from whatever the vendor quoted last year. Our SOC 2 penetration testing engagements start from that document for this reason.

How long is a penetration test report valid?

No standard sets an expiry. Auditors and customers treat anything older than 12 months as stale, and any report goes stale once the tested system changes significantly. For SOC 2 Type II, the only report that counts is one dated inside the review period.

Significant change means a new authentication model, a new public API, a cloud migration, a new tenant isolation design, or an AI agent given tool access to production data. Each warrants targeted testing, even if the annual test is months away.

Penetration testing SOC 2 evidence: auditor sampling a pentest report against system scope
Auditors sample the full report, not the executive summary.

Does ISO 27001 require annual penetration testing?

No. ISO/IEC 27001:2022 does not require penetration testing at any frequency. It requires you to manage technical vulnerabilities, test security during development and review security independently, then justify your choices in the risk assessment and Statement of Applicability (SoA).

Frequency is yours to set, but once set you must follow it. Certification bodies run surveillance audits every year and recertification every three years. If your procedure says “annual external penetration test” and the auditor finds no test in 14 months, that is a nonconformity against your own documented process, not against the standard.

That is the practical trap: overcommitting in policy. Write a frequency you will actually meet. Annual testing plus testing on significant change is the defensible default, with scope tied to the assets your risk assessment rates as high risk.

Which ISO 27001 Annex A controls does penetration testing satisfy?

  • Annex A 8.8, Management of technical vulnerabilities. The main one. Scanning finds known issues; a pentest shows which are exploitable and finds the logic flaws scanners miss. ISO/IEC 27002 guidance mentions penetration testing as a way to identify vulnerabilities.
  • Annex A 8.29, Security testing in development and acceptance. A test of a new system or major change before go-live is direct evidence.
  • Annex A 5.35, Independent review of information security. Testing by people outside the operating team contributes to independent review, though it does not replace internal audit under clause 9.2.

One engagement covers all three if the report states which systems were tested, which were tested pre-release, and who did the work. Our ISO 27001 penetration testing is scoped from your SoA for that reason.

What does DORA require for penetration testing?

DORA requires financial entities to run a testing program and to test all ICT systems and applications supporting critical or important functions at least yearly. Regulation (EU) 2022/2554 has applied since 17 January 2025. Unlike SOC 2 and ISO 27001, it writes testing obligations into law.

  • Article 24 requires a risk-based digital operational resilience testing program, proportionate to size and risk profile, with yearly testing of systems supporting critical or important functions.
  • Article 25 lists the tests the program can include: vulnerability assessments and scans, open source analysis, network security assessments, gap analyses, physical security reviews, source code reviews where feasible, scenario-based tests, compatibility, performance and end-to-end testing, and penetration testing.
  • Article 26 adds threat-led penetration testing (TLPT) for entities their competent authority identifies.
  • Article 27 sets requirements for the testers who perform TLPT.

Under Article 25, penetration testing is one item on a menu, not a standalone mandate. Supervisors will still expect the program for critical functions to include real exploitation testing of internet-facing and high-risk systems, with findings prioritized and fixed. Scanning everything annually but going no deeper will not read as a credible program.

What is TLPT and who needs it under DORA?

TLPT is an intelligence-led red team exercise against live production systems supporting critical or important functions. Designated entities must run it at least every three years. Most smaller entities will never be designated. Larger banks, insurers, payment firms and market infrastructure operators should assume they will be.

The criteria sit in the Commission’s delegated regulation on TLPT, adopted in 2025. They weigh systemic importance, ICT risk profile and impact on the financial sector, and the authority can adjust frequency for an individual entity. The method follows TIBER-EU, revised in 2025 to align with DORA:

  1. Preparation. Scope agreed with the authority, a control team set up inside the entity, providers procured.
  2. Testing. A threat intelligence provider builds scenarios from relevant actors, and red team testers execute them against production. Active red teaming must last at least 12 weeks.
  3. Closure. Red and blue teams replay the attack, a summary goes to the authority, and the entity submits a remediation plan. The authority then attests that the test took place.

ICT third-party providers supporting critical functions are in scope. Where testing a shared provider entity by entity would harm its other clients, Article 26 allows pooled testing.

TLPT has an authority in the loop from scoping to attestation. It cannot be repackaged from an ordinary pentest report. Our DORA TLPT service covers the full process.

How independent must a penetration tester be?

Independent enough that the people who built or run the system are not grading their own work. The exact bar differs by framework.

  • SOC 2: no requirement for an external firm. An internal team works if it is organizationally separate from the system owners and the auditor can evaluate its objectivity and competence. Most companies hire a third party because customers reading the report expect one.
  • ISO 27001: Annex A 5.35 requires independent review, which internal staff outside the tested area’s management line can provide. Certification auditors look at reporting lines.
  • DORA TLPT (Article 27): the strictest bar. Testers need proven capability, certification by an accreditation body in a Member State or adherence to formal codes of conduct, and professional indemnity insurance. Internal testers need authority approval and conflict-of-interest safeguards, and every third TLPT must use external testers. Significant credit institutions under the ECB’s Single Supervisory Mechanism must use external testers only.

None of the three requires rotating your external provider. Still, a fresh team every two or three years tends to find what the incumbent provider stopped looking for.

Vulnerability assessment vs penetration test vs TLPT

Vulnerability assessment Penetration test TLPT (DORA)
Question answered What known issues exist? What can an attacker actually exploit? Can a real threat actor reach critical functions undetected?
Method Mostly automated scanning Manual exploitation, business logic, chaining Threat-intel scenarios, covert red team, purple team closure
Environment Production or staging Usually production or production-like Live production, mandatory
Typical duration Hours to days 1 to 4 weeks Months, with at least 12 weeks of active testing
Frequency driver Monthly or quarterly (CC7.1, A 8.8) Annual and on major change Every 3 years for designated entities
Tester independence Internal team acceptable Separate from system owners; usually third party Article 27 criteria; external testers every third test
Regulator involvement None None Authority approves scope and issues attestation

The blue team knows about a pentest. It usually does not know about a TLPT. That one property is why TLPT measures detection and response, and why a pentest report cannot be relabeled as one.

What remediation evidence do auditors accept?

Three things: a retest showing the issue is fixed, a dated change record linked to the finding, or a signed management risk acceptance that names an owner, a reason and a compensating control. A spreadsheet with “fixed” in a status column is not evidence.

  • Retest letter. The strongest option. The original tester confirms each finding is closed, with dates and method. Put it in the original contract so it is not a second procurement.
  • Ticket and change trail. Finding ID mapped to a ticket, a pull request or config change, and a deployment date. Good for low and medium findings where a retest is not cost-effective.
  • Management risk acceptance. For findings you will not fix: who accepted the risk, why, which compensating control applies, and when it will be reviewed. Accepting a critical finding with no compensating control will draw an exception.

Set remediation deadlines by severity in your vulnerability management policy, then meet them. The numbers are your choice (one common pattern is 15 days for critical and 30 for high); no framework prescribes them. The auditor tests against the policy you wrote. Under DORA, the TLPT remediation plan also goes to the competent authority, which can follow up.

Can one penetration test satisfy SOC 2, ISO 27001 and DORA?

Yes for SOC 2, ISO 27001 and DORA’s yearly testing under Articles 24 and 25. No for TLPT, which is a separate, regulator-supervised exercise. The overlap is large enough that companies paying for two or three tests could usually do with one scope.

  1. Build scope as the union. Start with the SOC 2 system description, add high-risk assets from the ISO 27001 risk register, then add every system supporting a DORA critical or important function. Resolve conflicts toward broader coverage.
  2. Fix the timing. Test early in the SOC 2 review period and before the ISO surveillance audit, with retest time before both.
  3. Specify the report contents. Dates, scope mapping, methodology, tester identity, an independence statement, finding IDs, severity and a retest annex. One report, three readers.
  4. Include a pre-release component. Testing one new feature before launch gives you Annex A 8.29 evidence at no extra cost.
  5. Contract the retest up front. It is the remediation evidence every framework wants.

One caution keeps this honest: auditors check that a test happened, not that it was good. A scope built to find real risk (authenticated testing across every role, API authorization, cloud IAM paths, AI agents and their tool chains) produces the same audit artifact plus a report your engineers can act on.

If you are a financial entity still mapping DORA obligations, start with NIS2 and DORA readiness before buying either test.

One penetration test scope covering SOC 2, ISO 27001 and DORA requirements at their overlap
Scope as the union of three frameworks, tested once.
DORA TLPT threat-led penetration testing of live financial production systems
TLPT runs on live production, with the authority in the loop.

FAQ

Is a vulnerability scan enough for SOC 2?

Usually not for a Type II. Scans support CC7.1, but CC4.1 expects a separate evaluation, and most auditors and enterprise customers read that as a manual penetration test.

Can our internal security team do the SOC 2 penetration test?

Yes, if the team is organizationally independent of the system owners and the auditor can assess its competence. Many companies still hire a third party because customers reviewing the report expect it.

How often should we run a penetration test for ISO 27001?

The standard sets no frequency. Annual testing plus targeted tests after significant change is the defensible default, and whatever you write in policy is what the certification auditor will check.

Does every DORA entity need TLPT?

No. Only financial entities identified by their competent authority must run TLPT, at least every three years. All entities must still test systems supporting critical or important functions at least yearly.

Can a regular penetration test count as DORA TLPT?

No. TLPT requires threat intelligence scenarios, testing on live production, at least 12 weeks of active red teaming, and a competent authority that agrees the scope and issues the attestation. A standard pentest can feed DORA's yearly testing program, but not TLPT.

agent::offense scopes one engagement against your SOC 2 system description, ISO 27001 risk register and DORA critical functions, with the retest included. See how our compliance penetration testing works.



// get started

Work with AgentOffense

Tell us about your target and goals. We’ll reply with scope and a fixed-price quote — usually within one business day.

./request_engagement