In short
A penetration test follows six steps: defining the scope, obtaining written authorization, conducting reconnaissance of the attack surface, identifying and controlled exploitation of vulnerabilities, reporting, and verifying fixes. This method is based on public standards, primarily the OWASP Web Security Testing Guide for web applications and the PTES (Penetration Testing Execution Standard).
1. Scoping
Together, we define what will be tested (domains, applications, APIs, environments), the approach (black box, gray box, or white box), the testing windows, and the contacts to notify in the event of a critical discovery. A clearly defined scope is the first guarantee of a useful test without surprises.
2. Authorization
No attack may begin without a mandate signed by a person authorized to approve the test on these systems. If any part of the system is hosted by a third party (cloud provider, hosting provider), the third party’s terms of service must permit it. This document protects both parties: without it, accessing a system constitutes a criminal offense under Article 323-1 of the French Criminal Code.
3. Reconnaissance
We map everything an attacker would see: subdomains, pages, forms, API endpoints, technologies and versions, and authentication mechanisms. This is often where the first surprises emerge: a forgotten test environment, an exposed administration interface.
4. Discovery and Exploitation
We test each vulnerability category—access control, injection, authentication, session, configuration, and business logic—following the OWASP classifications. Each hypothesis is confirmed through a real-world but controlled exploitation: we demonstrate that we can read data that we should not be able to see, without modifying or destroying it.
Vulnerabilities are then chained together whenever possible: combining two medium-severity vulnerabilities can sometimes result in critical access.
5. Report
The report classifies each vulnerability by severity (CVSS score), along with evidence, impact, and the recommended remediation. A non-technical summary enables management to set priorities. It also specifies what could not be tested.
6. Verification of fixes
Once the fixes have been deployed, the identified vulnerabilities are retested to confirm that they have indeed been resolved. Without this step, a report remains nothing more than a list of intentions.
What Changes with AI Agents
At Sovra Recon, we follow these six steps. Steps 3 and 4 are carried out by offensive AI agents working in parallel in a fully automated manner, and a vulnerability is only included in the report if it has been verified. Details of our process are presented on our Methodology page.
Frequently asked questions
Which standards does a penetration test follow?
Primarily the OWASP Web Security Testing Guide (WSTG) for the web, the OWASP API Security Top 10 for APIs, the PTES for the general process, and the ANSSI PASSI standard for qualified service providers.
How many steps are involved in a penetration test?
There are generally six steps: scoping, authorization, reconnaissance, exploitation, reporting, and verification of fixes. Some standards provide more detail, but the logic remains the same.
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.
