README.md

🔐 AI-Powered Security Scanner

A security engineering tool for detecting high-impact authorization, logic, and configuration vulnerabilities in modern web applications and APIs.

📌 Overview

The AI-Powered Security Scanner is a purpose-built security engineering system designed to assist penetration testers and security teams in identifying real-world, high-impact vulnerabilities that are commonly missed by traditional automated scanners.

Unlike signature-based tools that prioritize breadth over accuracy, this project focuses on:

Behavioral analysis

Access-control validation

Context-aware vulnerability detection

The scanner combines deterministic security testing logic with AI-assisted analysis to produce findings that are:

Explainable

Reproducible

Suitable for professional security reports

🧠 Design Principles

🔎 Signal Over Noise

Focus on vulnerabilities that matter in real production systems.

⚙️ Deterministic Core, Assisted Intelligence

All scanning logic is transparent and reproducible

AI is used only for analysis and explanation

🛡️ Security-Engineer First

Findings are written the way a human security engineer would report them.

📊 Auditability

Every result is traceable to:

Specific request

Specific response

Clear decision path

🧩 Security Coverage

The scanner prioritizes high-risk vulnerability classes:

🔓 Authorization bypasses (IDOR, missing access checks)

🔁 Business logic flaws

🔑 Authentication & token handling issues

⚠️ Security misconfigurations

🌐 Insecure API behavior & exposure

These issues commonly lead to:

Account compromise

Data leakage

Privilege escalation

🤖 Role of AI

AI is not used for blind vulnerability discovery.

Instead, it is used to:

Analyze behavioral differences in HTTP responses

Explain the root cause of issues

Assess technical and business impact

Generate developer-friendly remediation guidance

The scanning engine remains fully deterministic, ensuring all findings are verifiable.

🏗 Architecture Overview

. ├── app.py # API entrypoint ├── scanner/ │ ├── core.py # Request orchestration & scan flow │ ├── ai_analyzer.py # AI-assisted analysis layer │ └── detectors/ │ ├── auth.py # Authentication & authorization tests │ ├── idor.py # IDOR detection logic │ ├── jwt.py # Token & JWT analysis │ └── headers.py # Security header checks ├── reports/ │ └── generator.py # Structured report generation └── requirements.txt

⚙️ Installation

Prerequisites

Python 3.9+

Playwright (browser-level behavior analysis)

Install Dependencies pip install -r requirements.txt playwright install chromium

🔧 Configuration

Optional: Enable AI-Assisted Analysis export OPENAI_API_KEY="your-api-key"

If not set, the scanner runs using deterministic logic only.

▶️ Running the Scanner

uvicorn app:app --reload --host 0.0.0.0 --port 8000

Service URL:

http://localhost:8000

📡 Example Scan Request

curl -X POST "http://localhost:8000/api/scan/url" \ -H "Content-Type: application/json" \ -d '{ "target": "https://httpbin.org", "scan_type": "full", "max_depth": 2 }'

📄 Output

Each scan includes:

Affected endpoint

Vulnerability classification

Reproduction logic

Impact assessment

Severity estimation

Remediation guidance

Designed for:

JSON output

PDF reports

Professional pentest delivery

⚖️ Ethical Use

🚨 This tool is intended only for educational and authorized security testing.

Do NOT scan systems without explicit permission from the owner.

Unauthorized use may be illegal.