Almost every SaaS company preparing for its first SOC 2 report asks the same question: do we actually need a penetration test? Compliance platforms, auditors and enterprise customers all seem to treat it as mandatory. The AICPA framework itself is a bit less direct. This guide covers what the Trust Services Criteria actually say, why auditors still expect a pentest, and what a test needs to look like for its report to hold up as audit evidence.
The short answer is no, not explicitly. The longer answer is that skipping it is one of the fastest ways to get an exception, a qualified opinion or a stalled enterprise deal.
SOC 2 reports are built on the AICPA Trust Services Criteria (TSC). The criteria are principles-based, not prescriptive. They describe what a control environment should achieve, not the specific tools or activities you have to use to get there.
Penetration testing appears in the "points of focus" that sit under some criteria, most notably:
Points of focus are guidance, not hard requirements. In principle you could satisfy CC4.1 and CC7.1 with other controls. In practice, few organizations can show equivalent assurance without a pentest.
Auditors have to form an opinion on whether your controls are designed well (Type I) and operated effectively over time (Type II). For a cloud-hosted product that takes customer data, the most persuasive evidence that your security controls work is an independent third party trying to break them and documenting what happened.
There are three practical pressures:
So "required" may not be the right word, but "expected" certainly is.
A well-scoped pentest produces evidence that touches several criteria at once. Mapping findings to these controls ahead of time makes the auditor's job easier and cuts down on follow-up requests.
If you also report against the Availability or Confidentiality categories, findings about resilience or data exposure support those criteria too.
Not all pentest reports are equal in an auditor's eyes. A cheap automated scan dressed up as a "penetration test" will not hold up. These are the elements auditors consistently look for.
Your SOC 2 report contains a system description covering the infrastructure, software, people, procedures and data in scope. The pentest scope must line up with it. If your system description includes a production AWS environment, a customer-facing web app, a public API and a Kubernetes cluster, the pentest should cover all four. A gap between the two is a red flag for the auditor.
Auditors want to know the test followed a repeatable standard. References to OWASP WSTG, the OWASP API Security Top 10, PTES or NIST SP 800-115 show the testing was structured and not ad hoc. The report should state which methodology was used and how it was applied.
For a Type II report, the pentest has to fall within the observation window, usually 3 to 12 months. A test completed two weeks before the window opens does not count as evidence of operating effectiveness during that window.
Each finding should include a severity rating (CVSS v3.1 or v4.0 is common), affected assets, evidence, and specific remediation guidance. Auditors use severity ratings to judge whether your remediation timelines were reasonable.
This is where many organizations fall short. Auditors don't expect a clean report. They expect proof that you handled what was found. A retest letter confirming critical and high findings were fixed, together with tickets showing when each was resolved, closes the loop on CC4.2 and CC7.1.
The test should be done by a party independent of the people who built and run the system. Auditors may also ask about tester qualifications, such as OSCP, OSWE, CRTO or equivalent experience.
Redfox Cybersecurity's penetration testing services are scoped against your SOC 2 system description from the start, so the report maps directly onto the controls your auditor will test.
The industry standard is at least once a year, with extra testing after significant changes. What counts as significant:
For a Type I report, which is a point-in-time assessment, a pentest done shortly before the report date shows the controls are designed appropriately. For a Type II report, schedule the pentest early enough in the observation window to leave time for remediation and retest before the window closes. A good rule is to test in the first third of the period.
The technical depth of the test decides whether the report actually shows control effectiveness. Below is a realistic view of the phases a professional engagement covers and the tooling experienced testers use to produce auditable evidence.
Good engagements start by turning the system description into a machine-readable scope file. This prevents out-of-scope testing and gives the auditor a clear record of what was covered.
# soc2_pentest_scope.yaml
engagement: "SOC 2 Type II Annual Pentest"
audit_period: "2026-01-01 to 2026-12-31"
test_window: "2026-03-02 to 2026-03-20"
methodology: ["OWASP WSTG v4.2", "OWASP API Top 10 2023", "NIST SP 800-115"]
in_scope:
web_apps:
- https://app.example.com
apis:
- https://api.example.com/v2
cloud:
- provider: aws
accounts: ["111122223333", "444455556666"]
kubernetes:
- cluster: prod-eks-use1
out_of_scope:
- https://status.example.com # third-party hosted
- denial_of_service_testing
contacts:
emergency: secops@example.com
[cta]
Testing begins by mapping everything an attacker can see. This confirms CC6.6 boundary protections and often surfaces shadow IT that is missing from your asset inventory.
# Enumerate subdomains from passive sources
subfinder -d example.com -all -recursive -o subdomains.txt
# Probe live hosts with tech fingerprinting
httpx -l subdomains.txt -title -tech-detect -status-code \
-follow-redirects -o live_hosts.txt
# Service discovery on in-scope IP ranges
nmap -sS -sV -p- --min-rate 2000 -T4 -iL scoped_ips.txt -oA external_full
# Templated detection of known misconfigurations and exposures
nuclei -l live_hosts.txt -severity critical,high,medium \
-etags dos,fuzz -rl 50 -o nuclei_external.txt
[cta]
Each discovered asset is checked against the system description. A forgotten staging environment holding production data is a common finding, and it is far better for a tester to find it than an attacker.
Encryption in transit is a frequent auditor focus under CC6.7. Testers check protocol versions, cipher suites and certificate configuration, and keep the output as evidence.
# Full TLS assessment with exportable evidence
testssl.sh --severity MEDIUM --jsonfile-pretty tls_results.json \
--htmlfile tls_report.html https://app.example.com
# Quick cipher enumeration
nmap --script ssl-enum-ciphers -p 443 app.example.com
[cta]
For most SaaS companies the application layer is where the real risk sits. This phase is mainly manual and is done in Burp Suite Professional against the OWASP WSTG and API Security Top 10. The areas auditors care most about are:
Testers usually work with two accounts per role across at least two tenants, so cross-tenant access can be checked across every object type the API exposes. If your product is multi-tenant, Redfox Cybersecurity's web application and API penetration testing puts tenant isolation first, because it is the finding that matters most to your auditor and your customers.
Most SOC 2 environments live in AWS, Azure or GCP. A configuration review against CIS benchmarks provides evidence for CC6.1, CC6.6 and CC7.1 in one pass.
# AWS: CIS benchmark and SOC 2 mapped checks with Prowler
prowler aws --compliance soc2_aws cis_2.0_aws \
--output-formats csv html json-ocsf \
--output-directory ./evidence/prowler
# Multi-cloud posture review with ScoutSuite
scout aws --profile audit-readonly --report-dir ./evidence/scoutsuite
# Azure equivalent using Prowler
prowler azure --az-cli-auth --compliance cis_2.0_azure \
--output-directory ./evidence/prowler-azure
[cta]
Prowler's SOC 2 compliance mapping is especially useful because it tags each check with the related Trust Services Criterion, which saves hours of manual mapping. Typical findings include overly permissive IAM policies, public storage buckets, missing CloudTrail coverage in some regions, and security groups exposing management ports.
If your system description includes containers, auditors expect them in scope. Testers assess cluster configuration, image vulnerabilities and RBAC.
# CIS Kubernetes benchmark checks on EKS worker nodes
kubectl apply -f https://raw.githubusercontent.com/aquasecurity/kube-bench/main/job-eks.yaml
kubectl logs job/kube-bench > ./evidence/kube-bench-eks.txt
# Scan production images for known vulnerabilities and misconfigurations
trivy image --severity HIGH,CRITICAL --format json \
-o ./evidence/trivy_api.json registry.example.com/api:2026.03.1
# Cluster-wide misconfiguration and RBAC review
trivy k8s --report summary --compliance k8s-cis-1.23 cluster
# Identify overly broad RBAC bindings
kubectl get clusterrolebindings -o json | \
jq -r '.items[] | select(.roleRef.name=="cluster-admin") |
"\(.metadata.name): \(.subjects[]?.name)"'
[cta]
When corporate identity systems such as Entra ID, Okta or Active Directory govern access to production, auditors increasingly expect them in scope. This phase tests whether a compromised employee laptop or phished account could reach customer data. The report should document which paths were tried, which controls stopped them (conditional access, privileged access management, network segmentation) and where gaps exist. That gives evidence for CC6.1, CC6.2 and CC6.3 on user access provisioning and least privilege.
Redfox Cybersecurity's internal and cloud identity assessments are built to show auditors how your access controls hold up against realistic attack paths, not just that they exist on paper.
A report sitting in a shared drive is not evidence on its own. Auditors want to see the full lifecycle: finding, ticket, fix, verification.
Define remediation timelines in your vulnerability management policy and stick to them. A common baseline:
Your auditor will sample findings and compare remediation dates against these SLAs. Missed SLAs without documented risk acceptance are a common source of exceptions.
Linking pentest findings to your ticketing system creates a clean audit trail. Here is a simple script that imports findings from a JSON report into Jira with severity-based due dates:
import json
from datetime import date, timedelta
from jira import JIRA
SLA_DAYS = {"Critical": 15, "High": 30, "Medium": 90, "Low": 180}
jira = JIRA(server="https://example.atlassian.net",
basic_auth=("secops@example.com", "API_TOKEN"))
with open("pentest_findings.json") as f:
findings = json.load(f)
for item in findings:
due = date.today() + timedelta(days=SLA_DAYS[item["severity"]])
issue = jira.create_issue(
project="SEC",
summary=f"[Pentest 2026] {item['id']}: {item['title']}",
description=(
f"Severity: {item['severity']} (CVSS {item['cvss']})\n"
f"Affected: {', '.join(item['assets'])}\n"
f"TSC Mapping: {', '.join(item['tsc'])}\n\n"
f"Remediation:\n{item['remediation']}"
),
issuetype={"name": "Vulnerability"},
duedate=due.isoformat(),
labels=["soc2", "pentest-2026", item["severity"].lower()],
)
print(f"Created {issue.key} due {due}")
[cta]
Once critical and high findings are fixed, the testing firm should verify each fix and issue a retest letter or updated report. This independent confirmation is far stronger evidence than an internal "fixed" status on a ticket.
Organize everything the auditor will ask for in one place:
soc2-evidence-2026/
├── 01_scope/
│ ├── soc2_pentest_scope.yaml
│ └── rules_of_engagement_signed.pdf
├── 02_reports/
│ ├── pentest_report_2026-03.pdf
│ └── retest_letter_2026-04.pdf
├── 03_tool_output/
│ ├── prowler/
│ ├── kube-bench-eks.txt
│ └── tls_report.html
├── 04_remediation/
│ ├── jira_export_pentest_2026.csv
│ └── risk_acceptances_signed.pdf
└── 05_detection/
└── siem_alerts_during_test_window.csv
[cta]
The detection folder is often overlooked. If your SIEM or EDR flagged the testers' activity, export those alerts. They are direct evidence for CC7.2, showing your monitoring controls work.
These are the issues that most often lead to follow-up requests or exceptions:
The right partner makes your audit smoother. When evaluating providers, ask:
Redfox Cybersecurity offers SOC 2-aligned penetration testing with TSC-mapped findings, a retest, and a customer-ready attestation letter, built for teams on an audit deadline. Current packages and pricing for web application, API, cloud and network engagements are listed on the Redfox Cybersecurity website.
SOC 2 doesn't explicitly require penetration testing, but in practice it is one of the most important pieces of evidence you can give an auditor. Pentesting appears as a point of focus under CC4.1, supports CC6 and CC7 controls, and is something enterprise customers will ask for no matter what the framework says.
To make your pentest count:
Treated as a compliance checkbox, a pentest produces a PDF. Treated as a real control evaluation, it gives you evidence your auditor can rely on and a security posture your customers can trust. If your audit window is coming up, talk to Redfox Cybersecurity about scoping a test that fits your timeline.