🎭 Dating Safety Module

Overview

Dating platforms are rife with catfishing, romance scams, hidden relationship status, and multi-platform identity games. Protected Outlet extends its scanning engine from websites → people.

Module goals:

• Detect stolen/stock photos in dating profiles (reverse image)
• Flag users in committed relationships hiding their status
• Identify romance scammer language patterns early
• Correlate identities across Tinder, Bumble, Hinge, OKCupid, Grindr
• Build community scammer blacklist (opt-in sharing)

🏗️ Architecture

The dating safety module sits alongside the website scanning engine in the same backend, with a separate /api/v1/scan/person endpoint. Extension content script detects dating platform URLs and routes accordingly.

Dating Site URL → Content Script → Background Worker
                               ↓
            Extract: name, username, photos, bio, location
                               ↓
            POST /api/v1/scan/person (with profile data)
                               ↓
            Parallel: reverse image + name correlation + scam language + relationship signals
                               ↓
            Return: { riskLevel, score, categories: { identity, relationship, catfish, scammer, multiProfile } }
                               ↓
            Floater updates color/icon; click opens dating report modal

🌐 Backend API

POST /api/v1/scan/person

{
  "platform": "tinder",          // tinder, bumble, hinge, okcupid, grindr
  "profileUrl": "https://tinder.com/p/123456",
  "profileData": {
    "name": "Sarah Johnson",
    "username": "sarah_j_89",
    "bio": "Travel enthusiast...",
    "photos": ["https://.../1.jpg", "https://.../2.jpg"],
    "location": "Austin, TX",
    "age": 32
  },
  "userId": "uuid"
}

Response (200):
{
  "personId": "hash_of_identity",
  "riskLevel": "low|medium|high|critical",
  "score": 72,
  "categories": {
    "identity": {
      "photoMatchCount": 2,
      "photoSources": ["instagram.com/user123", "google images result"],
      "nameConsistency": "good|conflicting|multiple",
      "usernameCorrelation": ["tinder:johnny_s", "bumble:jsmith89"]
    },
    "relationshipStatus": {
      "partnerMentioned": true,
      "socialSignals": ["photo with ring", "spouse mention in bio"],
      "confidence": "medium"
    },
    "catfishIndicators": {
      "stolenPhotos": true,
      "bioPlagiarism": false,
      "inconsistencies": 1,
      "severity": "medium"
    },
    "scammerPatterns": {
      "keywordHits": ["need money", "emergency"],
      "scammerDBMatch": false,
      "behavioralRedFlags": ["quick ILOVEYOU", "asks for personal info early"],
      "risk": "medium"
    },
    "multiProfile": {
      "profilesFound": 3,
      "platforms": ["tinder", "bumble", "poF"],
      "namesUsed": ["John Smith", "Johnny S.", "J. Smith"]
    }
  },
  "warnings": ["Multiple profiles detected", "Photo appears on Instagram with different name"],
  "recommendations": ["Request video call", "Do not send money", "Report to platform"]
}

🔍 Scanner Engine

1. Reverse Image Search

• Compute perceptual hash (pHash) of each photo
• Query TinEye API / Google Custom Search / Yandex Images
• Match against internal photo_blacklist (known scammer hashes)
• Return: list of URLs where photo appears + platform classification

2. Name/Username Correlation

• Normalize: "Sarah Johnson" → ["sarah j", "sjohnson", "sarah.j"]
• Google Custom Search: site:tinder.com "Sarah Johnson"
• Match same photo + different name = catfish

3. Scammer Language Detection

• Keywords: "need money", "gift cards", "bitcoin", "wire transfer", "hospital", "stranded", "passport lost", "love you" (within days)
• Patterns: rapid romance escalation, sob stories, financial requests
• NLP option: fine-tune small model on scammer vs genuine message dataset

4. Relationship Status Inference

• Public signals: spouse mentions in social bios, wedding ring emojis 💍, joint property records (US only)
• Behavioral: weekend-only availability, refuses video calls, late-night only
• Public records API (paid) for marriage/divorce verification

5. Multi-Profile Correlation

• Photo hash + name similarity across platforms
• Username patterns (same base name with variations)
• Location + age mismatches across platforms
• Build identity graph: person → platform profiles

6. Community Scammer Database

• User reports → moderation queue → verified blacklist
• Photo hashes of verified scammers auto-flag new profiles
• Shared blocklist (opt-in community protection)

📱 Extension UI: Dating-Mode Floater

Click floater → opens Dating Report Modal with sections: Photo Analysis, Identity Correlation, Relationship Status, Scammer Indicators, Multi-Profile Map, Behavioral Timeline, Verdict & Recommendations.

🛡️ Privacy & Legal Compliance

• Explicit consent: User must click scan; no automatic background scanning of every profile viewed
• Data minimization: Store only perceptual hashes (pHash), not full photos; auto-delete after 30 days
• Public data only: Never bypass login, respect robots.txt, rate limits
• User control: One-click delete all history
• Clear ToS: No harassment/stalking; informational only; always verify in person
• Legal review required: CFAA, GDPR Art. 9, defamation, platform ToS violations

🗺️ Implementation Roadmap

Sprints 6–7 (Person Identity MVP)

• Design person_scans DB table + GIN index on photo_hashes
• Integrate TinEye API (or Google Custom Search) for reverse image
• Implement pHash computation service (Python ImageHash / Node phash)
• Create /api/v1/scan/person endpoint
• Extend extension content script: detect dating site, extract profile data
• Background worker routes to person API
• Dating report modal on dashboard

Sprint 8 (Scammer Language)

• Compile scam keyword dictionary (FTC, Scamwatch sources)
• Build regex + scoring engine
• Optional: NLP with Compromise.js or basic TF-IDF
• User opt-in to analyze conversation snippets

Sprint 9 (Relationship + Behavior)

• Public records API integration (US, paid)
• Social signal detection (public Twitter/Instagram bios)
• Behavioral heuristics: video call refusals, meeting delays
• Privacy compliance review

Sprint 10 (Multi-Profile)

• Photo hash deduplication across platforms
• Name/username fuzzy matching (Levenshtein, soundex)
• Identity graph (Neo4j or recursive PG queries)
• Report: "This person also on Bumble as @X"

Sprint 11 (Community + Reputation)

• "Report Scammer" button in extension popup
• Admin moderation dashboard
• Photo blacklist auto-update on verified reports
• User reputation scoring (report quality)
• Opt-in shared blocklist

Sprint 12 (B2B + Launch)

• Dating platform API (white-label verification)
• SDK + documentation
• Pilot with small dating app
• Marketing: "Date safer with Protected Outlet"

💰 Monetization

B2B: License API to dating platforms (~$0.01 per MAU/month) to embed identity verification directly in their app.

📖 Related Documents

• Full Build Blueprint — Core website security module
• Dating Module Spec (Markdown) — Detailed technical spec
• Deployment Status — Frontend deployment info