[New Cheat Sheet Proposal] AI-Powered Advertising Systems Security

Context

AI-powered advertising systems (programmatic bidding, real-time targeting, creative generation, audience segmentation) represent a massive and growing attack surface not covered by existing OWASP cheat sheets. Current AI-related sheets (AI Agent Security, RAG Security, MCP Security, LLM Prompt Injection Prevention) focus on LLM application security but do not address the unique risks of advertising infrastructure where AI makes financial decisions at scale.

This proposal was suggested by @bkimminich from the OWASP Project Committee as a more appropriate starting point after I submitted a standalone Top 10 proposal for this domain.

Why This Is Needed

• Programmatic advertising processes over $ multi-Billion+ globally, with AI making targeting and bidding decisions in under 100ms
• Ad fraud costs the industry $ Multi Billion + annually, increasingly driven by adversarial AI
• Privacy regulations (GDPR, DMA, CPRA, COPPA) impose specific requirements on ad targeting that differ from general data processing
• The intersection of real-time financial transactions, user behavioral profiling, and AI inference creates attack vectors not found in other application domains
• No existing OWASP resource covers this space

Proposed Outline

1. Privacy & Consent Enforcement

• Multi-signal privacy checks (do-not-target, do-not-sell, do-not-share, opt-out)
• Consent validation before ad auction (IAB TCF string parsing)
• Age-of-majority checks for COPPA compliance
• Real-time opt-out enforcement with cache-backed lookups
• Privacy filtering before bid requests reach third parties

2. AI Model Security in Advertising

• Training data poisoning in bidding and targeting models
• Model inference attacks exposing user segment membership
• Adversarial inputs manipulating ad recommendations and bid prices
• Two-tier content safety: fast rule-based filters + LLM semantic review
• Confidence thresholding to reduce false positives
• Rate limiting and deployment locks for AI generation endpoints

3. Ad Fraud & Invalid Traffic (IVT) Prevention

• Multi-tier detection: pre-bid filtering + post-impression validation
• Bot detection via behavioral analysis and user-agent fingerprinting
• Dynamic blocklist generation (daily refresh from fraud pattern analysis)
• Outlier detection thresholds (impression caps per device/day)
• MRC-compliant filtering pipelines for billing integrity

4. Real-Time Bidding (RTB) Protocol Security

• OpenRTB request validation and sanitization
• Strict timeout enforcement on DSP bid responses
• Deal qualification before bid requests (prevent unauthorized targeting)
• Floor price enforcement (contractual + ML-predicted)
• Auction transparency via immutable event logs
• Supply chain verification (ads.txt, sellers.json)

5. Input Validation for Ad Creatives

• File upload sanitization with extension allowlists and path traversal prevention
• Unicode normalization (NFC) to prevent homograph attacks in ad text
• Zero-width character removal to prevent obfuscation
• VAST/VPAID tag sandboxing to prevent XSS from third-party creatives
• Content Security Policy headers for ad rendering contexts

6. Authentication & Authorization

• JWT validation with issuer allowlist, signature verification, and expiry checks
• JWKS key rotation for service-to-service authentication
• API key isolation per deployment
• Role-based access control for campaign management APIs

7. Beacon & Impression Verification

• Tamper-resistant impression tracking
• Duplicate beacon detection via lease tables
• Time-series storage for event recovery and reconciliation

8. Data Lineage & Audit Trails

• Append-only event logs for all ad decisions
• SOX-compliant data pipelines with immutable storage
• Revenue reconciliation pipelines for billing disputes

9. Rate Limiting & Resilience

• Circuit breakers for external service calls
• Exponential backoff with jitter for retries
• Graceful degradation under partial system failure

About Me

Principal Software Engineer at a Fortune 50 media company where I've spent 7+ years building and securing advertising systems that serve billions of ad decisions daily across streaming, digital, and programmatic channels. I've worked directly on ad serving, targeting, measurement, and compliance systems at scale.

References

• IAB Tech Lab OpenRTB Specification
• IAB ads.txt / sellers.json / app-ads.txt standards
• MRC Invalid Traffic Detection Guidelines
• TAG (Trustworthy Accountability Group) Certified Against Fraud Guidelines
• IAB Transparency & Consent Framework (TCF)
• NIST AI Risk Management Framework (AI 100-1)