In short
A penetration test, or pentest, is a simulated attack conducted with the system owner’s permission to uncover vulnerabilities that a real attacker could exploit. It differs from a configuration audit or a security scan in that it involves actual exploitation: the tester does not merely assume that a vulnerability exists—they prove it. The process concludes with a report that classifies each vulnerability by severity and indicates how to fix it.
Black box, gray box, or white box
| Approach | What the tester knows | What it simulates |
|---|---|---|
| Black box | Only the URL or the name of the target | An external attacker with no access whatsoever |
| Gray box | One or more user accounts | A compromised client, partner, or account |
| White box | Accounts, documentation, and sometimes the source code | A comprehensive audit, or an in-depth one |
The gray box approach is often the best compromise: most serious vulnerabilities (access to another customer’s data, privilege escalation) can only be detected once a connection has been established.
The main types of penetration tests
- Web application penetration testing: websites, extranets, SaaS applications. The most commonly requested service.
- API penetration testing: the interfaces that power mobile apps and partner systems, which are often less secure than the website itself.
- External penetration test: anything exposed to the Internet (VPNs, email systems, forgotten services).
- Internal penetration test: what an attacker who is already inside the network can do (compromised workstation, Active Directory).
- Mobile, cloud, and IoT penetration tests, as well as social engineering tests (simulated phishing).
The Steps of a Penetration Test
- Scope: scope, approach, dates, emergency contacts, rules of engagement.
- Written authorization: required. Without it, accessing a system is a criminal offense (Article 323-1 of the French Criminal Code).
- Reconnaissance and mapping of the attack surface.
- Identifying and exploiting vulnerabilities, without causing any damage.
- Report and findings.
- Verification of corrections (follow-up audit).
The details of each step are explained in our methodology guide.
What does a good penetration test report include?
- A summary for management: overall risk level, major vulnerabilities, and priorities.
- For each vulnerability: description, severity (often a CVSS score), proof of exploitation, actual impact, recommended fix, and references (CWE, OWASP).
- The scope of the test and what was not tested—an honest report also states its limitations.
The Legal Framework for a Penetration Test
In France, fraudulently accessing or remaining in a computer system is a crime under Article 323-1 of the French Criminal Code, punishable by three years’ imprisonment and a fine of 100,000 euros, and up to five years and 150,000 euros if data is deleted or modified or if the system’s operation is disrupted. Case law holds that the offense is committed even if the system was inadequately protected.
What distinguishes a penetration test from an attack is therefore authorization: a written mandate, signed by someone with the authority to authorize the test on these systems, specifying the scope and dates. When a component is hosted by a third party (cloud, hosting provider, SaaS), the third party’s terms of service must also permit it.
- Mandate or letter of authorization: who authorizes it, within what scope, and when.
- Rules of engagement: prohibited actions (denial-of-service attacks, data tampering), operating hours, and emergency contacts.
- A confidentiality agreement and, if personal data may be accessed, a data processing agreement (GDPR).
When should you conduct a penetration test?
- Before an application goes live or a major update is implemented.
- Regularly: at least once a year, and more often if the application is evolving rapidly.
- For compliance requirements (ISO 27001, NIS2, PCI DSS), client requirements, insurer requirements, or due diligence purposes.
- After an incident, to verify that no other vulnerabilities remain.
Frequently asked questions
What is the difference between a penetration test and a security audit?
“Security audit” is a broad term (covering organization, configuration, and compliance). A penetration test is a technical form of this, which involves actually attacking the system to prove the vulnerabilities.
Can a penetration test disrupt my production environment?
A well-conducted penetration test is non-destructive: no denial-of-service, no data tampering, and the ability to stop at any time. These rules are established in writing before the test begins.
What is a PASSI service provider?
An audit service provider certified by ANSSI under the PASSI standard. This certification is required for certain public agencies and regulated operators; for others, it is not mandatory.
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.
