ModernPentestModernPentest

Concepts

Key terminology and concepts used throughout ModernPentest

How It Works

Every pentest is built on your asset map — a live graph of your application. AI agents build that graph, reason over it, and attack it:

Reconnaissance

The recon agent crawls your application and builds an asset map — every page, endpoint, service, and the relationships between them.

Auditing

The auditor reads the whole graph at once and decides what's worth attacking, emitting prioritized investigations.

Proving

A prover confirms or denies each investigation by actively exploiting it — capturing a reproducible proof of concept and chaining deeper from what it confirms.

Validation & Reporting

Findings are deduplicated, critiqued for false positives, prioritized by real risk, and compiled into a report with remediation guidance and SOC 2 mapping.

Application

An application represents a single product or service with a unified backend. In ModernPentest, you register one application per product—not separate entries for different interfaces or platforms.

Attack Surfaces

Applications expose functionality to users through attack surfaces. ModernPentest supports two surface types:

SurfaceDescriptionWhat We Test
WebUser-facing websites and web appsOWASP Top 10, session management, authentication flows, client-side vulnerabilities
APIProgrammatic REST/GraphQL endpointsOWASP API Top 10, BOLA, injection, authorization bypass

Most modern applications have both surfaces—a web dashboard and an API that share users, data, and business logic. Configure both to get complete coverage. You can also register web-only or API-only applications.

Examples

Product TypeSurfacesConfiguration
SaaS with dashboard + public APIWeb + APIWeb URL + API URL with OpenAPI spec
Internal admin toolWeb onlyWeb URL with auth credentials
Developer platform (API-first)API onlyAPI URL with OpenAPI spec
Mobile app backendAPI onlyAPI URL (the mobile app hits this API)

Automatic Platform Detection

You don't need to configure platform-specific settings. Our AI agents automatically detect technologies like:

  • Supabase - RLS policies, storage buckets, edge functions
  • Firebase - Firestore rules, RTDB permissions, storage rules
  • Vercel/Next.js - Environment variables, API routes, middleware
  • Convex - Database access, function permissions

If you add Supabase to your project next month, our agents will detect it and run platform-specific tests—no configuration changes needed.

See Adding Applications for setup details.

Application Settings

Each application has:

  • Surfaces - Web and/or API configurations (editable anytime)
  • Authentication - Test credentials for each surface
  • Pentest configuration - Rate limits, exclusions, schedules

Vulnerability

A vulnerability is a security weakness that could be exploited by an attacker. We categorize vulnerabilities by:

Severity Levels

LevelDescriptionExampleAction
CriticalImmediate exploitation risk with severe impactSQL injection in login formFix within hours
HighSerious security impact, readily exploitableStored XSS in user commentsFix within 24-48 hours
MediumModerate risk, may require specific conditionsMissing rate limiting on APIFix within 1-2 weeks
LowMinor issues with limited impactVerbose error messagesFix when convenient
InfoObservations and recommendationsMissing security headersReview and acknowledge

Common Vulnerability Types

Injection Flaws

  • SQL Injection - Malicious SQL queries
  • Command Injection - OS command execution
  • XSS - Cross-site scripting attacks

Access Control Issues

  • IDOR - Accessing other users' data
  • Privilege Escalation - Gaining unauthorized access levels
  • Missing Authorization - Unprotected endpoints

Authentication Problems

  • Weak Sessions - Predictable or insecure session tokens
  • Credential Issues - Password policy weaknesses
  • JWT Vulnerabilities - Token security problems

Vulnerability Status

Vulnerabilities move through these statuses as you address them:

StatusMeaning
OpenNew vulnerability, not yet addressed
In RemediationBeing worked on
RemediatedFix has been applied, pending verification
FixedFix confirmed via rescan
Risk AcceptedRisk acknowledged, won't fix
False PositiveNot a real vulnerability

Pentest (Penetration Test)

A pentest is a security assessment where we actively test your application for vulnerabilities. Unlike passive security scans, pentests attempt to exploit weaknesses just like a real attacker would.

Our Approach

ModernPentest performs automated penetration testing using AI agents:

  1. Reconnaissance - Mapping your application into an asset graph
  2. Auditing - Reasoning over the graph to decide what to attack
  3. Proving - Confirming findings by actively exploiting them
  4. Reporting - Documenting validated results with remediation guidance

Finding

A finding is a single discovered vulnerability or security issue. Each finding includes:

  • Title - Brief description (e.g., "SQL Injection in /api/users")
  • Severity - Risk level (Critical, High, Medium, Low, Info)
  • Location - Where it exists (URL, parameter, code location)
  • Evidence - Proof the vulnerability exists
  • Remediation - How to fix it
  • References - CWE, OWASP, and external documentation

Report

A report is a document summarizing pentest results. ModernPentest generates several report types:

Pentest Report

Comprehensive technical report including:

  • Executive summary
  • All findings with details
  • Remediation guidance
  • Evidence and proof-of-concept
  • Risk scoring

See Understanding Results for more details.

SOC 2 Report

Auditor-ready compliance documentation including:

  • Testing methodology
  • Trust Services Criteria mapping (CC4.1, CC7.1)
  • Vulnerability findings
  • Remediation status
  • Testing timeline

See SOC 2 Reports for more details.

Agent

An agent is an AI-powered specialist that performs specific security tests. Rather than running generic scans, agents understand the context of your application and test intelligently.

Agent Types

AgentRoleWhat It Does
ReconnaissanceDiscoveryBuilds the asset map — pages, endpoints, services, technologies, auth flows
AuditorReasoningReads the whole graph and emits prioritized investigations
ProverExploitationConfirms or denies each investigation by actively exploiting it
ValidationCriticFilters false positives, finalizes severity, verifies remediation

Platform-specific weaknesses (Supabase RLS, Firebase rules, and so on) aren't a separate agent — recon detects the platform and tags the asset, and the auditor targets it. Learn more in AI Agents.

Pentest Execution

A pentest is a single execution of the testing pipeline. When you click "Start Pentest", a pentest begins. Each pentest:

  • Has a unique ID for tracking
  • Records all agent activity
  • Produces findings
  • Can be stopped or rerun

Pentest States

StateDescription
QueuedWaiting to start
RunningActively testing
CompletedFinished successfully
FailedEncountered an error
CancelledManually stopped

Organization

An organization is your team workspace in ModernPentest. Organizations:

  • Group related applications
  • Share team members and permissions
  • Have unified billing
  • Provide organization-wide reporting

Team members can have different roles:

  • Admin - Manage applications and team
  • Member - Run pentests, view results

Let us know at support@modernpentest.com if you would need more fine grained roles in your org.

Next Steps

Last updated: June 14, 2026

On this page