GitHub - sastrophy/siteiq: AI-powered security analysis, SEO optimization, and global performance testing in one platform.

21 min read Original article ↗

SiteIQ - Website Intelligence Platform

A comprehensive website analysis and security testing platform featuring security testing, SEO analysis, GEO testing, and LLM security testing.

⚠️ Security Notice

This application is designed to run LOCALLY (localhost) only.

Do NOT expose this application to the internet without proper security configuration. If you must deploy it remotely:

  1. Enable Authentication: Set SITEIQ_REQUIRE_AUTH=true and SITEIQ_API_KEY=your-secret-key
  2. Use HTTPS: Deploy behind a reverse proxy with TLS
  3. Restrict Access: Use firewall rules to limit access

Security Features

Feature Environment Variable Default
API Key Auth SITEIQ_REQUIRE_AUTH=true + SITEIQ_API_KEY=xxx Disabled
SSRF Protection SITEIQ_SSRF_PROTECTION=true Enabled
Rate Limiting SITEIQ_RATE_LIMIT=true Enabled
Rate Limit (requests) SITEIQ_RATE_LIMIT_REQUESTS=10 10/min
CSRF Protection Built-in Enabled
Input Sanitization Built-in Enabled

Screenshots

Dashboard

SEO Tests

LLM Tests

Scan Results

Help Page

Features

Current: Security Testing (OWASP Top 10)

  • A01:2021 - Broken Access Control

    • Directory traversal (LFI/RFI)
    • IDOR (Insecure Direct Object References)
    • CSRF protection validation
    • Open redirects
  • A02:2021 - Cryptographic Failures

    • SSL/TLS configuration
    • Certificate validation
    • HTTPS enforcement
    • Mixed content detection
    • TLS 1.3 preferred protocol detection
    • Certificate Transparency (SCT) validation
    • OCSP stapling check
  • A03:2021 - Injection

    • SQL injection (Classic, Union, Blind, Time-based)
    • NoSQL injection
    • Command injection
    • XSS (Reflected, Stored vectors, DOM-based)
    • SSTI (Server-Side Template Injection) - Jinja2, Twig, FreeMarker, Smarty, ERB
    • XXE (XML External Entity) - File read, blind XXE, SVG-based, SOAP
  • A05:2021 - Security Misconfiguration

    • Security headers (CSP, HSTS, X-Frame-Options, etc.)
    • Server information disclosure
    • Debug mode detection
    • Default pages
  • 2026 Modern Security Headers

    • Cross-Origin-Opener-Policy (COOP)
    • Cross-Origin-Embedder-Policy (COEP)
    • Trusted Types CSP directive
    • Sec-Fetch-* metadata headers (Site, Mode, Dest, User)
    • Reporting-Endpoints header
    • NEL (Network Error Logging)
    • Private Network Access (CORS-RFC1918)
  • A06:2021 - Vulnerable and Outdated Components

    • WordPress version detection
    • Plugin enumeration
  • A07:2021 - Identification and Authentication Failures

    • Brute force protection
    • Username enumeration
    • Session management
    • Cookie security flags
  • A10:2021 - Server-Side Request Forgery (SSRF)

    • URL parameter injection
    • Webhook endpoint testing
  • API Security

    • GraphQL introspection disclosure
    • Swagger/OpenAPI documentation exposure
    • Mass assignment vulnerabilities
    • CORS misconfiguration
    • API information disclosure
  • Secrets Detection

    • API keys in responses (AWS, Stripe, OpenAI, GitHub, etc.)
    • Configuration file exposure (.env, .config, etc.)
    • JavaScript secrets scanning
    • Source map exposure
    • Git repository exposure
  • Subdomain Takeover

    • Dangling DNS detection (NXDOMAIN)
    • S3 bucket takeover
    • Azure subdomain takeover
    • GitHub Pages takeover
    • Service fingerprinting (25+ cloud services)

SEO Analysis

  • On-Page SEO

    • Meta tags (title, description, viewport)
    • Heading structure (H1-H6 hierarchy)
    • Image optimization (alt text, dimensions)
    • URL structure analysis
  • Technical SEO

    • Robots.txt validation
    • Sitemap.xml validation
    • Canonical tags
    • Mobile friendliness
  • Structured Data

    • Schema markup (JSON-LD) validation
    • Open Graph tags
    • Twitter Cards
  • Performance SEO

    • Page load time
    • Compression detection
    • Caching headers
    • Core Web Vitals (via PageSpeed API)
  • International SEO

    • Hreflang validation
    • Language targeting
  • 2026 SEO Standards

    • llms.txt / llms-full.txt (AI crawler instructions)
    • Favicon presence and size validation
    • Speculation Rules API (prerendering/prefetching)

GEO Testing

  • Multi-Location Accessibility

    • Site accessibility from multiple regions
    • Geo-blocking detection
    • Response code consistency
  • Latency Analysis

    • Response times by region
    • Latency variance detection
    • CDN performance
  • Geo-Targeted Content

    • Content variation detection
    • Language switching
    • Currency detection
  • Regional Compliance

    • GDPR indicators (EU)
    • CCPA indicators (California)
    • Cookie consent presence
  • International SEO

    • Hreflang validation
    • Content-Language headers

WordPress-Specific Tests

  • Version detection
  • User enumeration (REST API, author parameter, login errors)
  • XML-RPC vulnerabilities (including pingback)
  • Plugin detection and version exposure
  • Configuration file exposure
  • Debug log exposure
  • wp-admin accessibility

LLM Security Testing (131 Tests)

Test your LLM-powered API endpoints for security vulnerabilities:

  • Prompt Injection

    • Direct prompt injection attacks
    • Indirect injection (RAG/context attacks)
    • Instruction override attempts
  • Jailbreaking

    • DAN-style jailbreaks
    • Role-play bypasses
    • Context manipulation
  • Persona/Character Continuation

    • Story continuation attacks
    • "Grandma exploit" style attacks
    • Fictional world framing
    • Split personality manipulation
  • Educational Framing

    • Academic research claims
    • Security researcher impersonation
    • Teaching example manipulation
    • Ethical hacking framing
  • Developer Mode

    • Fake debug/admin mode activation
    • Sudo command injection
    • Legacy version claims
    • Configuration override attempts
  • Encoding Bypass

    • Base64, ROT13, Hex, Binary encoding
    • Leetspeak and Pig Latin
    • Unicode escapes and Morse code
  • Nested Encoding

    • Double/triple Base64
    • Base64 + ROT13 combinations
    • Hex + Base64 chains
    • Multi-layer encoding bypass
  • Language Switching

    • Multilingual filter bypass (Spanish, French, German, Chinese, etc.)
    • Mixed language attacks
  • Multi-turn Manipulation

    • Fake conversation history injection
    • Trust building attacks
    • Context window overflow
  • Few-Shot Jailbreaking

    • Malicious example injection
    • Behavior cloning via examples
    • Reinforcement framing
    • Pattern demonstration attacks
  • Completion Baiting

    • Sentence/code completion traps
    • Pattern completion manipulation
    • JSON/list completion attacks
    • Fill-in-the-blank exploitation
  • Context Boundary Attacks

    • Token limit exploitation
    • Attention dilution attacks
    • Middle injection techniques
    • Newline/unicode flooding
  • Negation Logic

    • Opposite day attacks
    • Double negative confusion
    • Logical contradiction exploitation
    • Exception logic bypass
  • Token Manipulation

    • Token splitting/concatenation
    • Acronym expansion attacks
    • Phonetic bypass
    • Variable substitution tricks
  • System Prompt Leakage

    • Prompt extraction attempts
    • Configuration disclosure
    • Instruction revelation
  • Denial of Wallet (DoW)

    • Token multiplication attacks
    • Context window stuffing
    • Cost exploitation detection
    • Rate limiting verification
  • Tool/Function Abuse

    • Function call injection
    • Tool enumeration
    • Privilege escalation via tools
  • Indirect URL Injection

    • URL parameter injection
    • Markdown link attacks
    • Data URI injection
  • PII Handling

    • SSN, credit card, password exposure
    • API key leakage
    • Cross-session data leakage
  • Markdown/HTML Injection

    • XSS via LLM output
    • Phishing link injection
    • Tracking pixel injection
  • Unicode/Homoglyph Attacks

    • Zero-width character injection
    • Cyrillic/Greek homoglyphs
    • RTL override attacks
    • Full-width character bypass
  • Emotional Manipulation

    • Urgency/Emergency appeals
    • Authority impersonation
    • Guilt/Sympathy exploitation
    • Reverse psychology
  • RAG Poisoning

    • Document context injection
    • Metadata manipulation
    • Fake source attribution
    • Context overflow attacks
  • Model Fingerprinting

    • Model identity disclosure
    • Version detection
    • Capability enumeration
  • Training Data Extraction

    • Memorized content extraction
    • PII leakage from training
    • Code memorization probing
  • Cross-Tenant Leakage

    • Session confusion attacks
    • Memory probing
    • Tenant isolation testing
  • Instruction Hierarchy

    • System prompt override
    • Priority escalation
    • Boundary delimiter injection
  • Authentication

    • Unauthenticated access testing
    • API key validation
  • Hallucination Induction

    • Fake library/package requests
    • Non-existent CVE fabrication
    • Fake API endpoint generation
    • Package hallucination attacks
  • ASCII Art Jailbreak

    • Visual text representation bypass
    • Figlet/block letter attacks
    • Dot matrix encoding
    • Box drawing character injection
  • Refusal Suppression

    • Refusal word blocking
    • Output format constraints
    • Character limit abuse
    • JSON format forcing
  • Cipher/Encryption Games

    • Custom cipher bypass (Atbash, vowel shift)
    • Number-letter encoding
    • First-letter acrostics
    • Invented cipher attacks
  • Recursive Prompt DoS

    • Quine-style self-replication
    • Infinite explanation loops
    • Fibonacci expansion attacks
    • Binary tree generation
  • Semantic Dissociation

    • Educational misdirection attacks
    • Fictional framing bypass
    • Positive spin on harmful content
    • Metaphor and abstract game theory attacks
  • Fine-tuning Data Inference

    • Internal documentation probing
    • Company-specific knowledge extraction
    • HR/personnel data inference
    • Customer data probing
  • Adversarial Suffixes

    • Control character injection
    • Token boundary exploitation
    • Unicode BOM injection
    • Special token attempts
  • Implicit Instruction Following

    • Hidden document commands
    • Conversational pattern inference
    • False memory/authorization claims
    • Behavioral classification bypass
  • Sensitive File Output

    • RAG context file disclosure
    • Environment variable extraction
    • Source code leakage via context
    • Cloud/infrastructure config exposure

2025 Advanced LLM Security Tests (OWASP LLM Top 10 2025)

  • MCP/Tool Attacks

    • Line jumping (tool description injection)
    • Tool impersonation and hijacking
    • Context manipulation via tools
    • Cross-tool pollution attacks
    • MCP server authentication bypass
  • Memory/Context Poisoning

    • Echo Chamber attacks (90%+ success on GPT-4o/Gemini)
    • MemoryGraft (persistent compromise)
    • MINJA (memory injection via queries)
    • Cross-session pollution
    • Fabricated conversation history
  • Chain-of-Thought Manipulation

    • H-CoT (Hijacking Chain-of-Thought)
    • <think> token injection
    • Reasoning step disruption
    • Safety reasoning hijacking
    • CoT forging attacks
  • Structured Output Attacks

    • Chain Enum Attack (96.2% ASR on GPT-4o)
    • JSON key injection (duplicate key override)
    • Schema constraint bypass
    • Output format manipulation
    • Constrained decoding exploitation
  • Vector/Embedding Attacks

    • Embedding extraction attempts
    • Vector similarity exploitation
    • Cross-tenant vector leakage
    • RAG retrieval manipulation
    • Semantic collision attacks
  • CVE Attack Patterns

    • CVE-2025-32711 (EchoLeak) patterns
    • CVE-2025-54135 (CurXecute) patterns
    • GitHub Copilot RCE patterns
    • Document-embedded prompt injection
    • Log-to-prompt injection
  • Unbounded Consumption

    • Model extraction query patterns
    • Compute exhaustion attacks
    • Economic denial of service
    • Token multiplication attacks
    • Rate limit evasion techniques
  • Multimodal Attack Simulation

    • Image-based injection patterns
    • Audio transcription attacks
    • PDF hidden prompt detection
    • Cross-modal exploitation
    • OCR bypass techniques
  • Supply Chain Attacks

    • Backdoor trigger detection
    • Plugin impersonation
    • Poisoned model claims
    • Dependency confusion patterns
    • Training data poisoning indicators
  • Cognitive Overload / Paradox Attacks

    • Instruction paradox exploitation
    • Liar paradox / self-reference attacks
    • Trolley problem coercion
    • Policy contradiction exploitation
    • Meta-cognitive overload
    • Parallel task flooding
    • Utilitarian ethics bypass
  • Multi-Agent Compromise

    • Agent-to-agent message poisoning
    • Agent chain injection
    • Privilege escalation via agents
    • Orchestrator impersonation
    • Low-to-high privilege attacks
    • Pipeline stage injection
    • Agent delegation attacks
  • Misinformation / Disinformation

    • Political propaganda generation
    • Health misinformation requests
    • Executive/government impersonation
    • Conspiracy theory amplification
    • Astroturfing / fake grassroots
    • Demographic bias content
    • Election interference content
    • Financial panic / bank run content
  • Package Hallucination

    • Non-existent package suggestions
    • Supply chain attack via fake packages
    • Typosquatting package names
    • Bulk requirements hallucination
  • Glitch Token Attacks

    • Known glitch tokens (SolidGoldMagikarp, petertodd)
    • Tokenization boundary exploitation
    • Null/invisible character tokens
    • Mixed script token confusion
  • Crescendo Attacks

    • Gradual benign-to-harmful escalation
    • Academic/fiction framing escalation
    • Multi-stage context building
    • Trust-then-exploit patterns
  • CBRN Content

    • Chemical weapon information requests
    • Biological agent/toxin information
    • Radiological/nuclear content
    • Dual-use research concerns
  • Code Chameleon

    • JSON-embedded injection
    • XML/YAML structure attacks
    • SQL-style query injection
    • GraphQL/CSV hidden prompts
  • Math Prompt Framing

    • Harmful requests as word problems
    • Optimization problems for attacks
    • Game theory for attack strategy
    • Bayesian reasoning bypass
  • Persuasion Techniques

    • Fake expert endorsement
    • Fabricated organizational authority
    • Social proof pressure
    • Fake statistics/evidence
  • Snowball Hallucination

    • False premise acceptance
    • Future event fabrication
    • Pseudoscience confirmation
    • Conspiracy detail generation

Emerging LLM Security Tests (2024-2025 Research)

  • Skeleton Key Jailbreak (Microsoft Research)

    • Behavior augmentation attacks
    • Safety training bypass via role modification
    • Multi-turn skeleton key patterns
  • Many-Shot Jailbreaking (Anthropic Research)

    • In-context learning exploitation
    • Faux dialogue injection (100+ examples)
    • Long context window abuse
  • Visual Prompt Injection

    • Hidden text in images/PDFs
    • OCR-based injection
    • Steganographic payloads
  • Audio Prompt Injection

    • Ultrasonic hidden commands
    • Speech-to-text manipulation
    • Audio steganography
  • Agent Workflow Hijacking

    • LangChain/AutoGPT exploitation
    • Tool chain manipulation
    • Agent loop injection
  • Guardrail Bypass

    • NeMo Guardrails evasion
    • Llama Guard bypass
    • Safety classifier manipulation
  • System Prompt Reconstruction

    • Iterative extraction techniques
    • Binary search reconstruction
    • Token-by-token extraction
  • Prefilling Attacks (Claude-specific)

    • Response prefill manipulation
    • Assistant turn injection
    • Partial completion attacks
  • Function Parameter Injection

    • OpenAI function calling abuse
    • Tool parameter manipulation
    • Schema constraint bypass
  • Reasoning Trace Manipulation

    • o1/thinking model CoT hijacking
    • <think> token injection
    • Reasoning step override
  • Tool Result Injection

    • Poisoning tool outputs
    • Fake API responses
    • Tool result manipulation
  • Context Window Smuggling

    • Unicode padding attacks
    • Invisible character injection
    • Token boundary exploitation
  • Instruction Boundary Confusion

    • XML/JSON policy format injection
    • Delimiter confusion
    • System/user boundary attacks
  • Jailbreak Transfer

    • Cross-model attack portability
    • Universal jailbreak patterns
    • Model-agnostic exploits
  • Fine-Tuning Poisoning Detection

    • Backdoor trigger detection
    • Poisoned model indicators
    • Training data manipulation signs
  • Prompt Caching Attacks

    • Cache reconstruction
    • Shared cache injection
    • Cross-tenant cache leakage
  • Constitutional AI Bypass

    • RLHF guardrail evasion
    • Constitutional principle manipulation
    • Self-critique bypass
  • Model Merging Vulnerabilities

    • Merged model security gaps
    • Weight averaging exploits
    • Safety degradation in merges
  • Quantization Vulnerabilities

    • INT4/INT8 safety degradation
    • Precision loss exploitation
    • Quantized model jailbreaks
  • Speculative Decoding Attacks

    • Draft model exploitation
    • Speculative output manipulation
    • Parallel decoding attacks
  • Output Handling (OWASP LLM05) (NEW)

    • XSS injection in LLM outputs
    • SQL injection in generated queries
    • Command injection in output
    • SSRF via generated URLs
  • Excessive Agency (OWASP LLM06) (NEW)

    • Unauthorized tool access
    • Tool privilege escalation
    • Cross-tenant tool abuse
    • Dangerous tool chaining
    • HITL (Human-in-the-Loop) bypass
  • 2025-2026 MCP & Agent Attacks (NEW)

    • MCP Tool Poisoning (CVE-2025-6514)
    • Shadow Escape (zero-click MCP hijacking)
    • Confused Deputy (cross-privilege agent abuse)
    • ReAct2Shell (CVE-2025-55182)
    • LangChain Injection (CVE-2025-68664)
  • 2025-2026 Advanced Jailbreak Vectors (NEW)

    • Diffusion Attacker (optimized adversarial prompts)
    • Content Concretization, Sequential Break
    • Immersive World, Cascade Multi-Turn
    • Emoji Attacks, Script-Shaped Payloads
    • Evaluation Framing, Multilingual Safety
  • 2025-2026 RAG & Alignment Attacks (NEW)

    • Poisoned RAG (single & multimodal)
    • RAG Jamming (retrieval disruption)
    • Alignment Faking, Agentic Misalignment
    • Sycophancy Exploitation, Reward Hacking
  • Extraction & Side-Channel (NEW)

    • Write Primitive Extraction
    • PLeak (token-by-token prompt reconstruction)
    • Whisper Leak (side-channel probing)
  • Toxicity Scoring (NEW)

    • ML-based toxicity detection (Perspective API, Toxic-BERT)
    • Hate speech, threat, and bias detection
    • Signature-based fallback scoring
  • Adversarial Optimization (NEW)

    • Static suffix jailbreaks
    • Random search optimization
    • Genetic algorithm optimization
  • Red-Team Orchestration (NEW)

    • LLM-as-Attacker automated multi-turn attacks
    • System prompt extraction orchestration
    • Gradual escalation orchestration
  • Stateful Multi-Turn (NEW)

    • True multi-turn crescendo (separate HTTP requests)
    • Context poisoning across turns
    • Persona drift attacks
    • Trust building exploitation

Quick Start

1. Setup

cd siteiq
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

2. Run Web Application

Open http://localhost:5000 in your browser.

3. Or Run via CLI

# Run all tests against a target
python3 -m pytest --target-url=https://example.com

# Run with HTML report
python3 -m pytest --target-url=https://example.com --html=report.html

# Run specific test categories
python3 -m pytest --target-url=https://example.com -m sql_injection
python3 -m pytest --target-url=https://example.com -m xss
python3 -m pytest --target-url=https://example.com -m wordpress

Web Interface

SiteIQ includes a Jenkins-like web interface for running scans:

  • Dashboard - Enter URL and select test categories
  • Live Console - Real-time test output streaming
  • Results Page - Findings organized by severity
  • Scan History - Track previous scans
  • Help Guide - Comprehensive usage documentation

Access the help guide at http://localhost:5000/help

Command Line Options

Option Description Default
--target-url Target URL to test (required for non-LLM tests) -
--llm-endpoint LLM API endpoint URL (for LLM tests) -
--wordpress-path Path to WordPress installation /blog
--intensity Test intensity: light, medium, aggressive medium
--auth-username Username for authenticated testing -
--auth-password Password for authenticated testing -
--skip-ssl Skip SSL/TLS tests false
--skip-wordpress Skip WordPress-specific tests false

Test Markers

Run specific test categories using pytest markers:

# Security Tests
python3 -m pytest -m sql_injection    # SQL injection tests
python3 -m pytest -m xss              # XSS tests
python3 -m pytest -m csrf             # CSRF tests
python3 -m pytest -m headers          # Security headers tests
python3 -m pytest -m ssl              # SSL/TLS tests
python3 -m pytest -m wordpress        # WordPress tests
python3 -m pytest -m auth             # Authentication tests
python3 -m pytest -m traversal        # Directory traversal tests

# 2026 Security Headers Tests
python3 -m pytest -m coop             # Cross-Origin-Opener-Policy
python3 -m pytest -m coep             # Cross-Origin-Embedder-Policy
python3 -m pytest -m trusted_types    # Trusted Types CSP
python3 -m pytest -m sec_fetch        # Sec-Fetch-* headers
python3 -m pytest -m reporting_endpoints # Reporting-Endpoints header
python3 -m pytest -m nel              # Network Error Logging
python3 -m pytest -m private_network_access # Private Network Access (CORS-RFC1918)

# 2026 SSL/TLS Tests
python3 -m pytest -m tls13_preferred  # TLS 1.3 preferred protocol
python3 -m pytest -m certificate_transparency # Certificate Transparency (SCT)

# API Security Tests
python3 -m pytest -m api_security     # All API security tests
python3 -m pytest -m graphql          # GraphQL introspection
python3 -m pytest -m swagger          # Swagger/OpenAPI exposure
python3 -m pytest -m mass_assignment  # Mass assignment attacks
python3 -m pytest -m cors             # CORS misconfiguration

# Secrets Detection Tests
python3 -m pytest -m secrets          # All secrets detection
python3 -m pytest -m config_exposure  # Config file exposure
python3 -m pytest -m js_secrets       # JavaScript secrets
python3 -m pytest -m sourcemaps       # Source map exposure
python3 -m pytest -m git_exposure     # Git repository exposure

# SSTI (Template Injection) Tests
python3 -m pytest -m ssti             # All SSTI tests
python3 -m pytest -m jinja2           # Jinja2 injection
python3 -m pytest -m twig             # Twig (PHP) injection
python3 -m pytest -m freemarker       # FreeMarker (Java) injection
python3 -m pytest -m smarty           # Smarty (PHP) injection
python3 -m pytest -m erb              # ERB (Ruby) injection

# Subdomain Takeover Tests
python3 -m pytest -m subdomain_takeover  # All takeover tests
python3 -m pytest -m s3_takeover         # S3 bucket takeover
python3 -m pytest -m azure_takeover      # Azure takeover
python3 -m pytest -m github_takeover     # GitHub Pages takeover

# XXE (XML External Entity) Tests
python3 -m pytest -m xxe              # All XXE tests
python3 -m pytest -m xxe_file_read    # File read via XXE
python3 -m pytest -m xxe_blind        # Blind XXE detection
python3 -m pytest -m xxe_svg          # SVG-based XXE
python3 -m pytest -m xxe_soap         # SOAP endpoint XXE

# SEO Tests
python3 -m pytest -m seo              # All SEO tests
python3 -m pytest -m meta_tags        # Meta tags analysis
python3 -m pytest -m headings         # Heading structure
python3 -m pytest -m images           # Image optimization
python3 -m pytest -m robots           # Robots.txt tests
python3 -m pytest -m sitemap          # Sitemap tests
python3 -m pytest -m schema           # Schema markup tests
python3 -m pytest -m opengraph        # Open Graph tests
python3 -m pytest -m twitter          # Twitter Card tests
python3 -m pytest -m performance      # Performance SEO
python3 -m pytest -m pagespeed        # PageSpeed API tests
python3 -m pytest -m hreflang         # Hreflang tests

# 2026 SEO Tests
python3 -m pytest -m llms_txt         # llms.txt AI crawler tests
python3 -m pytest -m favicon          # Favicon validation
python3 -m pytest -m speculation_rules # Speculation Rules API

# GEO Tests
python3 -m pytest -m geo              # All GEO tests
python3 -m pytest -m accessibility    # Geo accessibility tests
python3 -m pytest -m latency          # Response time tests
python3 -m pytest -m content          # Geo content tests
python3 -m pytest -m compliance       # Regional compliance
python3 -m pytest -m cdn              # CDN tests

# LLM Security Tests (use --llm-endpoint instead of --target-url)
python3 -m pytest -m llm --llm-endpoint=https://api.example.com/chat           # All LLM tests (131 tests)
python3 -m pytest -m llm_injection --llm-endpoint=https://api.example.com/chat # Prompt injection
python3 -m pytest -m llm_jailbreak --llm-endpoint=https://api.example.com/chat # Jailbreaking
python3 -m pytest -m llm_leakage --llm-endpoint=https://api.example.com/chat   # System prompt leak
python3 -m pytest -m llm_dos --llm-endpoint=https://api.example.com/chat       # Denial of Wallet
python3 -m pytest -m llm_data --llm-endpoint=https://api.example.com/chat      # Data exfiltration
python3 -m pytest -m llm_encoding --llm-endpoint=https://api.example.com/chat  # Encoding bypass
python3 -m pytest -m llm_language --llm-endpoint=https://api.example.com/chat  # Language switching
python3 -m pytest -m llm_multiturn --llm-endpoint=https://api.example.com/chat # Multi-turn attacks
python3 -m pytest -m llm_tools --llm-endpoint=https://api.example.com/chat     # Tool/function abuse
python3 -m pytest -m llm_url --llm-endpoint=https://api.example.com/chat       # Indirect URL injection
python3 -m pytest -m llm_pii --llm-endpoint=https://api.example.com/chat       # PII handling
python3 -m pytest -m llm_markdown --llm-endpoint=https://api.example.com/chat  # Markdown/HTML injection
python3 -m pytest -m llm_unicode --llm-endpoint=https://api.example.com/chat   # Unicode/homoglyph bypass
python3 -m pytest -m llm_emotional --llm-endpoint=https://api.example.com/chat # Emotional manipulation
python3 -m pytest -m llm_rag --llm-endpoint=https://api.example.com/chat       # RAG poisoning
python3 -m pytest -m llm_fingerprint --llm-endpoint=https://api.example.com/chat # Model fingerprinting
python3 -m pytest -m llm_training --llm-endpoint=https://api.example.com/chat  # Training data extraction
python3 -m pytest -m llm_tenant --llm-endpoint=https://api.example.com/chat    # Cross-tenant leakage
python3 -m pytest -m llm_hierarchy --llm-endpoint=https://api.example.com/chat # Instruction hierarchy
python3 -m pytest -m llm_rate --llm-endpoint=https://api.example.com/chat      # Rate limiting
python3 -m pytest -m llm_auth --llm-endpoint=https://api.example.com/chat      # Auth bypass

# NEW LLM Jailbreak Tests
python3 -m pytest -m llm_persona --llm-endpoint=https://api.example.com/chat   # Persona/character jailbreak
python3 -m pytest -m llm_educational --llm-endpoint=https://api.example.com/chat # Educational framing
python3 -m pytest -m llm_devmode --llm-endpoint=https://api.example.com/chat   # Developer mode bypass
python3 -m pytest -m llm_completion --llm-endpoint=https://api.example.com/chat # Completion baiting
python3 -m pytest -m llm_nested --llm-endpoint=https://api.example.com/chat    # Nested encoding bypass
python3 -m pytest -m llm_boundary --llm-endpoint=https://api.example.com/chat  # Context boundary attacks
python3 -m pytest -m llm_fewshot --llm-endpoint=https://api.example.com/chat   # Few-shot jailbreaking
python3 -m pytest -m llm_negation --llm-endpoint=https://api.example.com/chat  # Negation logic bypass
python3 -m pytest -m llm_token --llm-endpoint=https://api.example.com/chat     # Token manipulation

# ADVANCED LLM Tests
python3 -m pytest -m llm_hallucination --llm-endpoint=https://api.example.com/chat # Hallucination induction
python3 -m pytest -m llm_ascii --llm-endpoint=https://api.example.com/chat     # ASCII art jailbreak
python3 -m pytest -m llm_refusal --llm-endpoint=https://api.example.com/chat   # Refusal suppression
python3 -m pytest -m llm_cipher --llm-endpoint=https://api.example.com/chat    # Cipher game bypass
python3 -m pytest -m llm_recursive --llm-endpoint=https://api.example.com/chat # Recursive prompt DoS
python3 -m pytest -m llm_semantic --llm-endpoint=https://api.example.com/chat  # Semantic dissociation
python3 -m pytest -m llm_finetune --llm-endpoint=https://api.example.com/chat  # Fine-tuning data inference
python3 -m pytest -m llm_adversarial --llm-endpoint=https://api.example.com/chat # Adversarial suffix bypass
python3 -m pytest -m llm_implicit --llm-endpoint=https://api.example.com/chat  # Implicit instruction following
python3 -m pytest -m llm_fileoutput --llm-endpoint=https://api.example.com/chat # Sensitive file output

# 2025 ADVANCED LLM Tests (OWASP LLM Top 10 2025)
python3 -m pytest -m llm_mcp --llm-endpoint=https://api.example.com/chat        # MCP/Tool attacks
python3 -m pytest -m llm_memory --llm-endpoint=https://api.example.com/chat     # Memory/context poisoning
python3 -m pytest -m llm_cot --llm-endpoint=https://api.example.com/chat        # CoT manipulation
python3 -m pytest -m llm_structured --llm-endpoint=https://api.example.com/chat # Structured output attacks
python3 -m pytest -m llm_vector --llm-endpoint=https://api.example.com/chat     # Vector/embedding attacks
python3 -m pytest -m llm_cve --llm-endpoint=https://api.example.com/chat        # CVE attack patterns
python3 -m pytest -m llm_consumption --llm-endpoint=https://api.example.com/chat # Unbounded consumption
python3 -m pytest -m llm_multimodal --llm-endpoint=https://api.example.com/chat # Multimodal attacks
python3 -m pytest -m llm_supplychain --llm-endpoint=https://api.example.com/chat # Supply chain attacks
python3 -m pytest -m llm_cognitive --llm-endpoint=https://api.example.com/chat   # Cognitive overload/paradox
python3 -m pytest -m llm_multiagent --llm-endpoint=https://api.example.com/chat  # Multi-agent compromise
python3 -m pytest -m llm_misinfo --llm-endpoint=https://api.example.com/chat     # Misinformation generation

# GARAK/PYRIT GAP Coverage Tests (NEW)
python3 -m pytest -m llm_package_hallucination --llm-endpoint=https://api.example.com/chat  # Package hallucination
python3 -m pytest -m llm_glitch_token --llm-endpoint=https://api.example.com/chat  # Glitch token attacks
python3 -m pytest -m llm_crescendo --llm-endpoint=https://api.example.com/chat     # Crescendo attacks
python3 -m pytest -m llm_cbrn --llm-endpoint=https://api.example.com/chat          # CBRN content
python3 -m pytest -m llm_code_chameleon --llm-endpoint=https://api.example.com/chat # Code chameleon
python3 -m pytest -m llm_math_framing --llm-endpoint=https://api.example.com/chat  # Math prompt framing
python3 -m pytest -m llm_persuasion --llm-endpoint=https://api.example.com/chat    # Persuasion techniques
python3 -m pytest -m llm_snowball --llm-endpoint=https://api.example.com/chat      # Snowball hallucination

# EMERGING LLM Tests (2024-2025 Research)
python3 -m pytest -m llm_skeleton_key --llm-endpoint=https://api.example.com/chat    # Skeleton Key jailbreak
python3 -m pytest -m llm_manyshot --llm-endpoint=https://api.example.com/chat        # Many-shot jailbreaking
python3 -m pytest -m llm_visual_injection --llm-endpoint=https://api.example.com/chat # Visual prompt injection
python3 -m pytest -m llm_audio_injection --llm-endpoint=https://api.example.com/chat  # Audio prompt injection
python3 -m pytest -m llm_agent_hijack --llm-endpoint=https://api.example.com/chat    # Agent workflow hijacking
python3 -m pytest -m llm_guardrail_bypass --llm-endpoint=https://api.example.com/chat # Guardrail bypass
python3 -m pytest -m llm_prompt_reconstruction --llm-endpoint=https://api.example.com/chat # System prompt reconstruction
python3 -m pytest -m llm_prefilling --llm-endpoint=https://api.example.com/chat      # Prefilling attacks
python3 -m pytest -m llm_function_injection --llm-endpoint=https://api.example.com/chat # Function parameter injection
python3 -m pytest -m llm_reasoning_manipulation --llm-endpoint=https://api.example.com/chat # Reasoning trace manipulation
python3 -m pytest -m llm_tool_injection --llm-endpoint=https://api.example.com/chat  # Tool result injection
python3 -m pytest -m llm_context_smuggling --llm-endpoint=https://api.example.com/chat # Context window smuggling
python3 -m pytest -m llm_boundary_confusion --llm-endpoint=https://api.example.com/chat # Instruction boundary confusion
python3 -m pytest -m llm_jailbreak_transfer --llm-endpoint=https://api.example.com/chat # Jailbreak transfer
python3 -m pytest -m llm_finetuning_poison --llm-endpoint=https://api.example.com/chat # Fine-tuning poisoning
python3 -m pytest -m llm_cache_attack --llm-endpoint=https://api.example.com/chat    # Prompt caching attacks
python3 -m pytest -m llm_constitutional_bypass --llm-endpoint=https://api.example.com/chat # Constitutional AI bypass
python3 -m pytest -m llm_model_merging --llm-endpoint=https://api.example.com/chat   # Model merging vulnerabilities
python3 -m pytest -m llm_quantization --llm-endpoint=https://api.example.com/chat    # Quantization vulnerabilities
python3 -m pytest -m llm_speculative_decoding --llm-endpoint=https://api.example.com/chat # Speculative decoding attacks

# 2025-2026 ADVANCED LLM Tests (NEW - 48 tests across 7 files)
python3 -m pytest -m llm_output_xss --llm-endpoint=https://api.example.com/chat       # Output XSS injection (LLM05)
python3 -m pytest -m llm_output_sqli --llm-endpoint=https://api.example.com/chat      # Output SQL injection (LLM05)
python3 -m pytest -m llm_agency --llm-endpoint=https://api.example.com/chat           # Excessive Agency (LLM06)
python3 -m pytest -m llm_agency_hitl --llm-endpoint=https://api.example.com/chat      # HITL bypass (LLM06)
python3 -m pytest -m llm_mcp_poisoning --llm-endpoint=https://api.example.com/chat    # MCP Tool Poisoning (CVE-2025-6514)
python3 -m pytest -m llm_shadow_escape --llm-endpoint=https://api.example.com/chat    # Shadow Escape
python3 -m pytest -m llm_confused_deputy --llm-endpoint=https://api.example.com/chat  # Confused Deputy
python3 -m pytest -m llm_react2shell --llm-endpoint=https://api.example.com/chat      # ReAct2Shell (CVE-2025-55182)
python3 -m pytest -m llm_langchain_injection --llm-endpoint=https://api.example.com/chat # LangChain Injection
python3 -m pytest -m llm_alignment_faking --llm-endpoint=https://api.example.com/chat # Alignment Faking
python3 -m pytest -m llm_reward_hacking --llm-endpoint=https://api.example.com/chat   # Reward Hacking
python3 -m pytest -m llm_sycophancy --llm-endpoint=https://api.example.com/chat       # Sycophancy Exploitation
python3 -m pytest -m llm_emoji_attack --llm-endpoint=https://api.example.com/chat     # Emoji Attack
python3 -m pytest -m llm_multilingual_safety --llm-endpoint=https://api.example.com/chat # Multilingual Safety
python3 -m pytest -m llm_pleak --llm-endpoint=https://api.example.com/chat            # PLeak Extraction
python3 -m pytest -m llm_whisper_leak --llm-endpoint=https://api.example.com/chat     # Whisper Leak
python3 -m pytest -m llm_toxicity --llm-endpoint=https://api.example.com/chat         # Toxicity Scoring
python3 -m pytest -m llm_adversarial_opt --llm-endpoint=https://api.example.com/chat  # Adversarial Optimization
python3 -m pytest -m llm_orchestration --llm-endpoint=https://api.example.com/chat    # Red-Team Orchestration
python3 -m pytest -m llm_crescendo_stateful --llm-endpoint=https://api.example.com/chat # Stateful Multi-Turn

Test Intensity Levels

Level Duration Coverage Use Case
light 5-10 min Basic Quick assessment
medium 15-25 min Balanced Regular testing (default)
aggressive 30-60 min Thorough Comprehensive audit

Project Structure

siteiq/
├── config.py              # Configuration management
├── conftest.py            # Pytest fixtures
├── requirements.txt       # Dependencies
├── README.md              # This file
├── DEPLOYMENT.md          # Deployment guide
├── geo.txt                # SEO/GEO implementation plan
├── payloads/              # Attack payloads
│   ├── sql_injection.py
│   ├── xss.py
│   ├── directory_traversal.py
│   ├── wordpress.py
│   ├── seo.py             # SEO test data & thresholds
│   ├── geo.py             # GEO test data & regions
│   ├── llm.py             # LLM attack payloads
│   ├── llm_emerging.py    # Emerging LLM threats (2024-2025)
│   ├── api_security.py    # API security payloads
│   ├── secrets.py         # Secret patterns & paths
│   ├── ssti.py            # Template injection payloads
│   └── xxe.py             # XXE injection payloads
├── utils/
│   └── scanner.py         # Core scanner utilities
├── tests/                 # Test modules
│   ├── test_sql_injection.py
│   ├── test_xss.py
│   ├── test_security_headers.py
│   ├── test_ssl_tls.py
│   ├── test_authentication.py
│   ├── test_wordpress.py
│   ├── test_directory_traversal.py
│   ├── test_csrf_owasp.py
│   ├── test_seo.py        # SEO analysis tests
│   ├── test_geo.py        # GEO testing
│   ├── test_llm.py        # LLM security tests
│   ├── test_api_security.py    # API security tests
│   ├── test_secrets.py         # Secrets detection
│   ├── test_ssti.py            # Template injection
│   ├── test_subdomain_takeover.py  # Takeover tests
│   └── test_xxe.py             # XXE injection tests
├── webapp/                # Web application
│   ├── app.py
│   └── templates/
│       ├── index.html     # Dashboard
│       ├── results.html   # Results page
│       └── help.html      # User guide
└── reports/               # Generated reports (JSON)

Security Severity Levels

Level Description Action
CRITICAL Immediate exploitable risk (SQL injection, RCE) Fix immediately
HIGH Serious vulnerability (XSS, auth bypass) Fix within 24-48 hours
MEDIUM Moderate risk (missing headers, weak SSL) Fix within 1-2 weeks
LOW Minor issue (version disclosure) Fix when possible
INFO Informational (potential attack surface) Review and consider

Reports

JSON reports are generated in the reports/ directory:

{
  "target": "https://example.com",
  "timestamp": "2024-01-15T10:30:00",
  "total_findings": 5,
  "findings_by_severity": {
    "critical": 1,
    "high": 2,
    "medium": 1,
    "low": 1,
    "info": 0
  },
  "findings": [...]
}

Legal Disclaimer

IMPORTANT: This tool is intended for authorized security testing only.

  • Only test systems you own or have explicit written permission to test
  • Unauthorized testing may violate laws and regulations
  • The authors are not responsible for misuse of this tool

Documentation

License

MIT License - See LICENSE file for details.