In short
For a SaaS provider, penetration testing serves two purposes: to identify vulnerabilities that could expose one customer’s data to another, and to demonstrate the platform’s security to buyers, auditors, and investors. The number one risk in a multi-tenant application is the lack of authorization at the object level (BOLA), ranked first in the OWASP API Security Top 10 for 2023: changing an identifier in a request is sometimes enough to access another tenant’s data. Since the code changes every week, an annual test no longer reflects the platform’s actual state: testing must be conducted after every significant change.
The core risk: isolation between clients
A multi-tenant application serves all its clients from the same infrastructure, often the same database. The separation between clients then relies entirely on the code: a forgotten filter in a request, an API route that fails to verify who owns an object—and one company’s data becomes accessible to another.
- Objects belonging to another tenant: modifying the identifier of an invoice, a project, or a user in a request (BOLA, or IDOR).
- Falsifiable tenant identifier: a “tenant_id” transmitted by the client in a header, the request body, or an improperly validated token.
- Administrative functions: a standard user who invites members, changes roles, or views the billing information for another account.
- Exports, searches, and dashboards: features that aggregate data but fail to apply client-specific filtering.
- Files: predictable or permanent download links to other clients’ documents.
- Caches, message queues, and background tasks: data served to the wrong client due to a lack of isolated keys.
- Integrations: SSO, SCIM provisioning, webhooks, and API keys that are not properly restricted to their tenant.
The OWASP Cloud Tenant Isolation project identifies these types of errors in PaaS and SaaS offerings. To test them thoroughly, at least two distinct client organizations are required, each with multiple roles.
APIs: The Primary Attack Surface
In a modern SaaS system, the web interface is just one API client among many: mobile apps, integrations, and clients that automate their usage. The API therefore exposes everything the platform is capable of doing. The OWASP API Security Top 10 (2023 edition) lists object-level authorization flaws (BOLA) as the top vulnerability, followed by weak authentication and property-level authorization flaws.
- Forgotten versions: an old API version that remains online without the controls added since then (API9 risk, inventory management).
- Unlimited usage: no throughput limits, massive data exports, and skyrocketing infrastructure costs.
- Sensitive business workflows: unlimited creation of trial accounts, abuse of invitations, and circumvention of subscription quotas.
- Keys and tokens: API keys with no expiration or scope, and improperly validated JWTs.
Security questionnaires, ISO 27001, SOC 2: what the report proves
As soon as a SaaS platform targets large enterprises, the procurement department sends a security questionnaire, which commonly includes the question: “Has a penetration test been conducted by a third party within the last twelve months?”
| Request or Repository | What Is Expected | Value of the penetration test |
|---|---|---|
| Client Security Questionnaire | Evidence of a recent third-party test and status of fixes | A summary or certifiable report that can be shared without disclosing the details of the vulnerabilities |
| ISO/IEC 27001:2022 | Controls 8.8 (technical vulnerability management) and 8.29 (security testing during development and acceptance) | Dated evidence of the identification and remediation of vulnerabilities |
| SOC 2 | Report by an independent auditor (CPA) on controls related to security, availability, data integrity, confidentiality, and privacy, in accordance with AICPA criteria | The standard does not explicitly require penetration testing by name, but a recent report is among the pieces of evidence commonly presented to the auditor. |
| DORA (financial sector clients) | For services that support critical or important functions, the contract must provide for the service provider’s participation in penetration tests based on the financial entity’s threat profile (Article 30 of Regulation 2022/2554). | A culture of testing and the ability to cooperate with a client-led test |
A good report distinguishes between two levels: an executive summary that can be shared with a client or auditor, and technical details reserved for the team responsible for fixing the issues.
Fundraising and Due Diligence
During a fundraising round or acquisition, investors scrutinize technical risks just as closely as financial statements. A data leak between customers discovered during the acquisition audit—or immediately afterward—can weigh heavily on negotiations. Having a recent test report, and especially proof that critical vulnerabilities have been fixed and then re-verified, allows you to address these issues before they become a sticking point.
Deploy every week, test once a year?
A SaaS platform delivers continuously: new API endpoints, new roles, new integrations. A penetration test report describes the platform as of the day of the test; a few months and dozens of production deployments later, it no longer proves much. ISO 27001, in fact, requires security testing during development and acceptance (control 8.29), not just once a year.
- Initial comprehensive test: the entire platform, multiple tenants, multiple roles.
- Targeted testing for each significant change: authentication, roles and permissions, billing, new APIs, and integrations.
- Recurring tests: Re-run isolation scenarios with each release to detect regressions.
- Verification of fixes: every fixed vulnerability is retested, and the proof is retained.
This is the principle behind continuous penetration testing (PTaaS). Sovra Recon offers this through SENTINELLE, which provides continuous offensive testing conducted by autonomous AI agents; RAID covers one-time audits, such as those conducted prior to a funding round or a request for proposals.
Frequently asked questions
Our cloud hosting provider is certified: do we still need a penetration test?
Yes. Under the cloud’s shared responsibility model, the security of the application, APIs, configurations, and accounts remains the responsibility of the software vendor. The hosting provider’s certification says nothing about isolation between your customers.
Should a multi-tenant SaaS application be tested in production?
An identical pre-production environment is often preferable, with at least two test organizations. In production, testing is limited to dedicated accounts, without touching actual customer data.
Can the penetration test report be shared with a client?
Yes, in the form of a summary or attestation, often subject to a confidentiality agreement. Technical details of vulnerabilities—especially those that remain unpatched—are not disclosed.
What is the difference between SOC 2 and ISO 27001?
ISO 27001 is an international standard that certifies an information security management system. SOC 2 is an attestation report issued by an auditor according to AICPA criteria, primarily requested by North American clients. In both cases, a penetration test report serves as evidence.
Sources
Find out where an attacker would get in.
Our AI agents actually attack your scope, fully automated, and hand you the proof of every vulnerability.
