# IOTA Web Guardian - Roadmap 2026

**Authors:** Paolo Alberti, Niccolò Normani  
**Last Updated:** April 28, 2026  
**Version:** 1.0

---

# Horizon 0 - 2026
## Medium Goal - Agent Detector Plugin
**Timeline:** June → August 2026  
**Focus:** Ship a free, production-ready WordPress plugin with AI bot detection (no blockchain, no payments)

### Core Functionality
The plugin focuses on **one job done extremely well**: detecting AI bots and giving creators control.

**Features:**
- **Detection Engine**
  - JA4 TLS fingerprinting (headless browsers vs real Chrome)
  - User-Agent heuristics with fallback rules
  - Behavioral signals (missing CSS/JS fetch, request rate patterns)
  - Honeypot link mechanism for trap-based detection

- **Creator Controls**
  - Per-bot granular control: block, allow, or log
  - Activity dashboard in WordPress admin
    - Which AI bots visited
    - Request frequency and page patterns
    - Detection confidence scores

- **Allowlist Management**
  - Built-in verified allowlist (Googlebot, Bingbot, Common Crawl via reverse DNS)
  - Publisher UI to extend allowlist
  - "I am human" accessibility path for false positives

- **Distribution & Onboarding**
  - Listed on WordPress.org Plugin Directory
  - 60-second setup: install → email confirm → first bot blocked
  - Zero blockchain jargon in UI
  - Works on shared hosting (no infrastructure changes required)

### Success Criteria
- **Target:** 100+ active installs by end of July 2026
- WordPress.org marketplace listing live
- Detection accuracy: >90% on known AI crawlers
- False positive rate: <2% (measured via telemetry)
- Average setup time: <60 seconds

### Distribution Channels
- WordPress Plugin Directory (organic search)
- ProductHunt launch
- HackerNews ("Show HN" post)
- IndieWeb, IndieHackers communities
- Direct outreach to WordPress creator newsletters
- Twitter/X threads targeting creators affected by AI scraping

### Technical Architecture
- WordPress-native (no DNS changes, no external services)
- Detection at plugin layer (PHP + optional sidecar for advanced fingerprinting)
- Opt-in telemetry for continuous improvement
- Structured logging with reason codes for every block/allow decision

---

## Maximum Goal - Non-Custodial IOTA Web Guardian
**Timeline:** August → December 2026  
**Focus:** Introduce optional payment layer on IOTA for creators who want to monetize AI access through a non custodial approach (not E2E yet, which would be horizon 1)

### New Capabilities
Building on Phase 1's detection foundation, add the economic layer:

**Monetization Features:**
- **x402 Payment Protocol**
  - AI agents receive 401 challenge on protected content
  - Compliant agents present Verifiable Presentation (VP) signed by DID
  - Payment processed on IOTA L1 → Guardian JWT issued → 200 OK content

- **Web2-First Onboarding (Custodial Mode)**
  - Email/Google signup → wallet provisioned by partner custodian (e.g., Magic Labs)
  - DID generation and binding happen invisibly orchestrated by our solution
    - Private Keys are managed through KMS integration (i.e. AWS, HashiVault Corp)
    - Verifiable Credentials (VCs) are stored by us
    - Verifiable Presentations are signed by us based on the corresponding DID Private Keys
    - The payment is performed with the hosted non-custodial wallet in Magic Labs
  - Publisher dashboard shows balance in EUR/USD, not IOTA tokens
  - Fiat off-ramp integrated (Transak, Ramp, MoonPay, or SEPA)
  - One-click withdrawal to PayPal/Stripe/bank account

- **Pricing UI**
  - Human language: "€0.50 per 1,000 scraper visits"
  - Suggested presets for blog/news/docs
  - Advanced view available on demand (DID, blockchain details)

### Security & Anti-Bypass Measures
- **VP Replay Prevention**
  - VP bound to request-specific nonce (timestamp + URL hash)
  - JWT bound to IP/UA/DID → re-auth on context change
  - Short TTL (1 hour), nonce rotation

- **Sybil Resistance**
  - Combined per-DID and per-origin rate limiting
  - Reputation scoring: new DIDs throttled, clean history = fast-lane

- **Detection Hardening**
  - Layered detection: TLS fingerprint + behavioral + optional PoW challenge
  - Progressive challenge system (invisible to humans, costly for mass scrapers)
  - Honeypot links, HTTP/2 settings frame fingerprint

### Trade-Off Considerations
Based on learnings from Phase 1, evaluate:
- **Custodial vs. Non-Custodial Default**
  - Custodial (recommended by E2E solution): Zero regulatory risk for us, better UX, partner dependency
  - Non-Custodial: More control, higher complexity for users, higher latency as centralised by a "Web2" approach

- **Detection Accuracy vs. False Positives**
  - Default-permissive policy on first hit (async classification)
  - Always-available "I am human" path
  - Dashboard transparency: "X requests rejected, Y% likely human"

- **Pricing Model**
  - Free tier: Detection + blocking (always free)
  - Pro tier ($9/month): Advanced analytics, allowlist management, priority updates
  - Revenue share: % of payments processed through monetization layer

### Success Criteria
- **10 real pilot publishers** onboarded (non-friends/family)
- Each running for 30+ days with shared telemetry
- **3 written case studies** with real numbers:
  - Revenue captured
  - Scrapers blocked
  - Uptime and conversion friction
- **External validation:**
  - Public bug bounty program launched
  - Light external security audit completed (1-week scope)
  - Public whitepaper: "How We Distinguish Humans from AI Agents"
- **End-of-year metrics:**
  - Active installs, scrapers blocked, EUR redistributed to publishers
  - p50/p95 latency measurements
  - Material ready for next funding round

### Demand-Side Enablement
- Reference SDK for legitimate AI agents (Claude, OpenAI) calling x402-protected endpoints correctly
- Documentation: "How AI companies can pay for content access"
- Without credible demand side, system is just a firewall. With it, it becomes a market.

---

## Cross-Cutting Tracks (April → December 2026)

### Track A: Hardening & Measurement
**Owner:** Niccolò (lead)  
**Focus:** Performance, security, observability

- Continuous benchmarking pipeline
- Stress testing on IOTA Rebased testnet
- Public benchmark page (updated quarterly)
- SOC 2 compliance roadmap (target 2027 certification)

### Track B: Detection Robustness
**Owner:** Niccolò (with external review)  
**Focus:** Anti-bypass, accuracy

- Four attack vector mitigation:
  1. Scraper masquerading as browser
  2. False positives on humans
  3. VP replay/forging
  4. Sybil attacks on DIDs
- Layered detection: fingerprint + behavior + challenge
- Honest framing: "We make bypass non-profitable, not impossible"

### Track C: Web2 Onboarding
**Owner:** Paolo (lead)  
**Focus:** Non-crypto publisher UX

- Custodial mode by default (partner integration)
- Fiat-denominated pricing and withdrawals
- Zero blockchain jargon on first screen
- 60-second setup wizard
- Whitelabel UI (publisher branding, not ours)

### Track D: Distribution & Proof
**Owner:** Paolo + Niccolò (joint)  
**Focus:** Real publishers, case studies, marketplace

- WordPress.org submission (October)
- Pilot publisher recruitment (non-friends/family)
- Video tutorial (IT + EN, 5-10 min)
- FAQ and exit story documentation
- Public report: aggregate metrics at year-end

---

## Key Performance Indicators (End of December 2026)

| Metric | Target |
|--------|--------|
| Real publishers onboarded | 10+ |
| WordPress.org plugin listing | Live |
| Public benchmark (p50/p95/p99) | Published |
| Threat model document | Published |
| External security audit | Completed |
| Detection bypass mitigation | All known vectors non-economic |
| Custodial on-ramp partner | ≥1 integrated |
| Case studies with real numbers | 3 |
| Public whitepaper on detection | Published |

---

## Risks & Mitigation

| Risk | Likelihood | Impact | Mitigation |
|------|------------|--------|------------|
| Bypass discovered after public launch | High | Medium | Bug bounty, fast patching, layered detection |
| Custodial partner unavailable/expensive | Medium | High | Evaluate 2 partners in Q2; fallback to non-custodial-only |
| Legitimate AI agents don't adopt x402 | Medium | High | Demand-side SDK (Q4), Anthropic/OpenAI outreach |
| WordPress.org rejects plugin | Low | Medium | Pre-submission compliance review |
| Latency regression on IOTA testnet | Medium | Medium | Continuous benchmarking, fallback paths |
| Large publisher requires SOC 2 | Medium | Medium | Compliance roadmap published, 2027 certification target |

## Guiding Principles

1. **Detection as foundation** – If detection breaks, everything breaks. Continuous priority.
2. **Web2 first impression, Web3 underneath** – Publishers shouldn't know DIDs exist unless they want to
3. **Bug bounty as signal** – Even small programs signal openness to scrutiny
4. **Default-permissive on humans** – False positives destroy trust faster than bypass attempts

**Prepared by:** Paolo Alberti, Niccolò Normani  
**For:** IOTA Foundation, project advisors, investor discussions  
**Status:** Internal review → External sharing

# Horizon 1 - 2027 - Full IOTA Web Guardian

Full IOTA Web Guardian E2E integration as per hackaton proposal. All wallets and DIDs are custodial and privates.