Sovra Recon·Guide · Web Penetration Testing

Web Application Penetration Testing: What is Tested and How

By Solan Barlet, co-founder of Sovra Recon
Updated October 29, 2026

In short

A web application penetration test involves attacking a website, extranet, or SaaS application to find vulnerabilities that could be exploited to steal data, hijack an account, or take control of the server. It covers the categories of the OWASP Top 10—primarily access control, the leading cause of serious vulnerabilities—as well as the application’s specific business logic, which no scanner can test.

What Is Tested

  • Access control: Can a user view or modify another user’s data (IDOR), or access administrative functions?
  • Authentication: passwords, MFA, password reset, session tokens, and JWTs.
  • Injections: SQL, system commands, templates, and cross-site scripting (XSS).
  • Server-Side Request Forgery (SSRF): Can the server be tricked into querying internal resources?
  • Configuration: security headers, exposed files, overly verbose error messages, and obsolete components.
  • Business logic: bypassing a payment, reusing a discount coupon, or skipping a validation step.

Why conduct a connected test (gray box)

Most serious vulnerabilities in a web application are only visible once a user is authenticated: that is where the data and permissions reside. Providing two accounts of the same level and one account of a higher level allows for testing client isolation and privilege escalation—the two most costly risks.

Preparing for a Web Penetration Test

  • Choose the environment: production (the most realistic) or an identical pre-production environment.
  • Create dedicated test accounts with fictitious data.
  • Notify the hosting provider and the operations team, and whitelist the test IP addresses in case a WAF blocks all traffic.
  • Designate a contact who can be reached during the test in case a critical vulnerability is discovered.

Frequently asked questions

Is a scanner enough to secure a web application?

No. A scanner detects known vulnerabilities and configuration errors, but not access control or business logic vulnerabilities, which are the most serious and most common.

Should testing be done in production?

This is the most realistic approach, provided the test is non-destructive and well-defined. A strictly identical pre-production environment is a valid alternative.

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.