SVG Security Analysis Toolkit - Tool Documentation

This toolkit contains specialized Python tools for analyzing potentially malicious SVG files and detecting security mechanisms.

Overview

The toolkit consists of four main tools:

1. extract.py - Static analysis for obfuscated URL extraction
2. extract_dynamic.py - Dynamic analysis using box-js for JavaScript execution
3. cf_probe.py - Cloudflare protection detection (Turnstile, CAPTCHA, etc.)
4. encoder.py - Test case generator for creating obfuscated SVG samples

Tool #1: extract.py - Static SVG URL Extractor

Purpose

A pattern-based static analysis tool that decodes obfuscated URLs from malicious SVG files without executing any code. This is the safer option for initial analysis.

Key Features

• XOR Decoding: Automatic detection and decoding of XOR-encrypted payloads
• Character Arithmetic: Decodes String.fromCharCode patterns with numeric calculations
• Base64 Extraction: Finds and decodes Base64-encoded URLs
• Email Detection: Identifies email addresses used in URL concatenation
• Multiple Obfuscation Strategies: Handles various deobfuscation techniques

Usage

# Analyze a single SVG file
python3 extract.py -i malicious.svg

# Verbose output showing decoded JavaScript
python3 extract.py -i malicious.svg -v

# Debug mode with detailed analysis steps
python3 extract.py -i malicious.svg -d

# Read from stdin
python3 extract.py < malicious.svg

Key Detection Patterns

1. Secretkey XOR Pattern: Detects String.fromCharCode(115,101,99,...) ("secretkey")
2. XOR Loop Detection: Identifies for loops with parseInt and XOR operations
3. Data URI Scripts: Extracts and decodes Base64 content from data: URIs
4. Nested atob() + Variable: Handles atob("base64") + variableName patterns

Dependencies

• Python 3.6+ (uses only standard library)

Example Output

DYNAMIC SVG URL EXTRACTOR
=========================================================
SVG #1 Results:
----------------------------------------
Obfuscation: XOR decoding (key length: 9)
Extracted URLs:
  → https://malicious.example.com/[email protected]

Tool #2: extract_dynamic.py - Dynamic SVG JavaScript Analyzer

Purpose

A sophisticated dynamic analysis tool that extracts and analyzes JavaScript embedded in SVG files using box-js. This tool executes JavaScript in a sandboxed environment and is designed to extract final, complete URLs even when built through complex obfuscation.

Key Features

• JavaScript Execution: Uses box-js to safely execute SVG scripts
• Advanced Hook System: 300+ line JavaScript payload with comprehensive monitoring
• URL Construction Tracking: Captures URLs built dynamically through string concatenation
• Final URL Prioritization: Distinguishes between complete URLs and partial components
• ActiveX/WScript Monitoring: Specialized Windows script object detection
• Multiple Sink Detection: Monitors location, window.open, fetch, XHR, Image.src, etc.

Dependencies

# Required: box-js must be installed globally
npm install -g box-js

# Python requirements (optional)
pip install -r requirements.txt

Usage

# Basic analysis of a single SVG
python3 extract_dynamic.py -i malicious.svg

# Analyze with custom output directory
python3 extract_dynamic.py -i malicious.svg -o analysis_results

# Analyze directory of SVGs with extended timeout
python3 extract_dynamic.py -i svg_samples/ -t 60

# Keep temporary files for debugging
python3 extract_dynamic.py -i malicious.svg --keep-temp

# Save URL blocklist to file
python3 extract_dynamic.py -i malicious.svg --blocklist-out urls.txt

# Mask email addresses for privacy
python3 extract_dynamic.py -i malicious.svg --mask-emails

Advanced Features

Hook System (PRELUDE_JS)

The tool injects a comprehensive JavaScript payload that hooks:

• Location Operations: location.href, location.assign(), location.replace()
• Window Methods: window.open(), navigation methods
• HTTP Requests: fetch(), XMLHttpRequest.open()
• String Operations: String.prototype.concat, string manipulation
• Decoders: atob(), decodeURI(), unescape()
• Dynamic Execution: eval(), Function(), setTimeout()

Pattern Recognition

Automatically detects and handles:

• Simple hex decoding patterns
• Hex-encoded dynamic execution
• XOR + dynamic eval construction
• Character mapping + find() execution
• Arithmetic character code patterns
• Common obfuscation techniques

Example Output

=== Processing: malicious.svg ===
  Wrote temp JS -> /tmp/boxjs_svg_ABC/test.boxed.js
  Running box-js (timeout 30s) with sink hooks...
  [+] box-js finished. Results in: boxjs_out/malicious.svg.boxjs_results

  *** FINAL COMPLETE URLs FOUND ***:
    → https://evil.example.com/[email protected]
  
  Emails found:
    → [email protected]

=== FINAL ANALYSIS SUMMARY ===
🎯 COMPLETE URLs EXTRACTED (1 total):
======================================================
  → https://evil.example.com/[email protected]
======================================================

Tool #3: cf_probe.py - Cloudflare Protection Detection

Purpose

A web reconnaissance tool for detecting Cloudflare security mechanisms including Turnstile, CAPTCHA systems, browser verification, and other protection measures across redirect chains.

Key Features

• Redirect Chain Following: Automatically follows HTTP redirects and meta-refresh redirects
• Multiple Detection Patterns: Detects Turnstile through various methods
• JavaScript File Scanning: Analyzes linked JS files for additional sitekeys
• Compression Support: Handles gzip, deflate, and brotli compression
• Comprehensive Protection Detection: Identifies multiple Cloudflare protection types

Dependencies

pip install requests
pip install brotli  # Optional for brotli support

Usage

# Check single URL for Cloudflare protections
python3 cf_probe.py https://example.com

# Scan with JavaScript file analysis
python3 cf_probe.py https://example.com --scan-js

# Process multiple URLs from file
python3 cf_probe.py -i urls.txt

# Custom User-Agent and timeout
python3 cf_probe.py https://example.com -A "Custom-Agent" -t 30

# Disable TLS verification (for testing)
python3 cf_probe.py https://example.com --insecure

# Dump HTML/JS for analysis
python3 cf_probe.py https://example.com --dump analysis_dump/

Detection Capabilities

Turnstile Detection

• data-sitekey attributes
• turnstile.render() JavaScript calls
• turnstile.create() JavaScript calls
• Turnstile API script references
• cf-turnstile CSS classes

Other Cloudflare Protection

• Custom CAPTCHA systems
• Browser verification pages
• CF challenge containers
• CF headers (CF-Ray, server headers)
• DDoS protection messages

Example Output

=== Redirect/Navigation Chain (3 steps) ===

[0] URL: https://example.com
    Status: 301
    Cloudflare Protection: YES (CF Headers)
    CF Header: cf-ray: 123abc456def789

[1] URL: https://protected.example.com
    Status: 200
    Cloudflare Protection: YES (Turnstile, Custom CAPTCHA)
    Turnstile Site key(s): 0x1AAA000000000000000
    CAPTCHA Evidence: Click the computer to verify

=== Summary ===
Cloudflare Protection Detected: Turnstile with key(s): 0x1AAA000000000000000 | Custom CAPTCHA

Tool #4: encoder.py - SVG Test Case Generator

Purpose

Generates obfuscated SVG files with various encoding patterns for testing the analysis tools.

Key Features

• Multiple Obfuscation Patterns: 6 different encoding techniques
• Dynamic Variability: Randomized variable names, structures, and encoding parameters
• Realistic Test Data: Generates random domains and email addresses
• Batch Generation: Can create complete test suites

Usage

# Generate single pattern
python3 encoder.py -u "https://evil.com/c2" -p 1 -o test.svg

# Generate all patterns in directory
python3 encoder.py --random-all -o test_patterns/

# Custom email address
python3 encoder.py -u "https://evil.com" -e "[email protected]" -p 3 -o custom.svg

Obfuscation Patterns

1. XOR + Proxy Pattern: Enhanced XOR encoding with ES6 Proxy for execution
2. Hex Function Pattern: Hex-encoded dynamic execution via Function constructor
3. Object Keys Pattern: XOR + dynamic eval construction using object properties
4. Character Mapping Pattern: Character mapping with Array.find() execution
5. Arithmetic Pattern: Arithmetic character code evaluation
6. Data URI Pattern: Base64 encoded scripts in data URIs

Example Output

$ python3 encoder.py --random-all -o test_suite/
Generated xor_proxy_enhanced: test_suite/pattern_01_xor_proxy_enhanced.svg
  URL: https://r4nd0m.example.com/[email protected]

Generated hex_function: test_suite/pattern_02_hex_function.svg  
  URL: https://evil-site.org/[email protected]

[... 4 more patterns ...]

Testing Workflow

Recommended Analysis Workflow:

# Step 1: Generate test cases
python3 encoder.py --random-all -o test_cases/

# Step 2: Static analysis first (safer)
python3 extract.py -i test_cases/pattern_01_*.svg -v > static_results.txt

# Step 3: Dynamic analysis (in isolated environment)
python3 extract_dynamic.py -i test_cases/ -o dynamic_results/ --blocklist-out malicious_urls.txt

# Step 4: Verify Cloudflare protection on extracted URLs (if safe to do so)
python3 cf_probe.py -i malicious_urls.txt

Installation

# Clone or download the toolkit
cd svg_tools/

# Install Python dependencies
pip install -r requirements.txt

# Install box-js globally (for dynamic analysis)
npm install -g box-js

# Verify installation
python3 extract.py --help
python3 extract_dynamic.py --help
python3 cf_probe.py --help
python3 encoder.py --help