Saltar al contenido principal

Proof of Personhood Systems

The Unique Human Problem

Challenge: Prove you're a unique human without revealing who you are.

Requirements:

1. Uniqueness: One proof per human (no Sybils)
2. Privacy: Don't reveal identity
3. Decentralization: No central authority
4. Accessibility: Available to everyone
5. Security: Hard to fake/transfer

Paradox: These requirements are contradictory!

For uniqueness: Need to verify each person is distinct
But: Verification requires some identifying information
And: Information can compromise privacy

For privacy: Minimize information collected
But: With minimal info, can't ensure uniqueness
Result: Potential Sybils

For decentralization: No trusted party
But: Someone must verify humanness
Who?: Trusted parties (contradiction)

For accessibility: No barriers to entry
But: Verification has costs (time, money, technology)
Result: Some people excluded

For security: Hard to fake
But: Biometrics can be faked, spoofed
Social verification can be gamed
Technical verification can be bypassed

No perfect solution exists. All systems trade off these properties.

Worldcoin: Biometric Verification

Approach: Iris scanning for uniqueness

How it works:

1. Visit an "Orb" (scanning device)
2. Iris scan captures unique pattern
3. Generate hash of iris pattern
4. Check hash against database (no duplicates)
5. If unique, issue World ID (credential)
6. World ID links to crypto address
7. User can now prove humanness with World ID

Technical design:

Iris scan → Iris Code (2048 bits)

Hash to 256 bits

Store only hash (not original scan)

Check for duplicates in database

If unique: Issue credential

Credential stored on-chain or device

Privacy claims:

Worldcoin says:
- Iris scan deleted after hash generation
- Only hash stored (can't reverse to image)
- No personal info collected (except iris)
- Can prove uniqueness without revealing identity

Verifier sees:
"This World ID belongs to a unique human"

Verifier doesn't see:
- Who the human is
- Where they registered
- Iris scan or hash

Criticisms:

1. Centralization:
- Worldcoin Foundation controls Orbs
- Database of hashes is centralized
- Can be shut down, censored
- Single point of failure

2. Privacy concerns:
- Biometric data inherently identifying
- Hash reversibility unclear (potential future attacks)
- Location data from Orb visits
- Surveillance implications

3. Accessibility:
- Limited Orb availability (mostly wealthy countries)
- Physical visit required
- Excludes remote areas
- Barrier for disabled individuals

4. Coercion risk:
- Biometrics can't be revoked
- If compromised, permanent
- Potential for forced scanning
- Black market for World IDs

5. Trust assumptions:
- Must trust Worldcoin deletes scans
- Must trust hash can't be reversed
- Must trust no backdoors
- Unverifiable claims

Scale:

Launched: July 2023
Users: 5M+ verified (as of mid-2024)
Orbs: 1,500+ locations globally
Token price: $2-5 per WLD
Funding: $250M from a16z and others

Proof of Humanity: Video Verification

Approach: Social verification with video proof

How it works:

1. Submit profile:
- Name (or pseudonym)
- Photo
- Video of yourself saying "I certify I am a real human"
- Ethereum address

2. Deposit stake:
- Lock ETH as anti-spam measure
- Currently: 0.125 ETH (~$250)

3. Verification phase:
- Existing verified humans review submission
- Check: Is video genuine? Is person real?
- Vouch: Existing users can vouch for you
- Challenge: Anyone can challenge (costs deposit)

4. If approved:
- Profile becomes verified
- Address gets Proof of Humanity status
- Can now claim UBI token (periodic distribution)
- Can participate in Sybil-resistant systems

5. Renewal:
- Must re-verify periodically (annually)
- Prevents dead profiles

Game theory:

Vouching:
- If you vouch for fake profile, you lose stake
- If you vouch for real profile, you might earn rewards
- Incentive: Only vouch for people you're confident in

Challenging:
- Cost: Deposit to challenge
- If challenge succeeds: Earn stake from fake profile
- If challenge fails: Lose your deposit
- Incentive: Challenge obvious fakes, avoid borderline

Verifying:
- Jurors earn fees for correct decisions
- Jurors lose stake for incorrect decisions
- Incentive: Review carefully, vote honestly

Challenges:

1. Video can be deepfaked:
- AI-generated videos improving rapidly
- May fool human reviewers
- Arms race: Detection vs generation

2. Identity rental:
- Real person creates profile
- Sells access to scammer
- Scammer controls address
- Hard to detect remotely

3. Accessibility:
- Requires camera
- Requires internet
- Requires existing verified user to vouch
- Barriers for some populations

4. Privacy trade-off:
- Video and photo are public
- Name (even pseudonym) is public
- Less private than Worldcoin
- Face is strongly identifying

5. Scalability:
- Human reviewers don't scale
- Long verification times (days/weeks)
- Expensive (stake + fees)

Scale:

Launched: March 2021
Verified profiles: ~15,000 (as of 2024)
UBI distributed: $1M+ to verified humans
Success rate: ~70% of submissions approved
Cost: ~$300 (stake + gas) per verification

BrightID: Social Graph Verification

Approach: Prove humanness through social connections

How it works:

1. Download BrightID app
2. Make connections:
- Video call with people you know
- They verify you in-app
- Mutual verification (both parties confirm)

3. Join groups:
- Verification parties (scheduled events)
- Existing communities
- Build connection graph

4. Achieve verification:
- Need N connections (varies by application)
- Connections must be diverse (not all from same cluster)
- Achieve "sponsorship" level
- Apps can set own thresholds

5. Use BrightID:
- Prove to apps you're human
- Without revealing your identity
- Just prove: Sufficient verified connections

Social graph analysis:

Honest user's graph:
Alice
/ | \
/ | \
Bob Carol Dave
| \ | / |
| \ | / |
Emma Frank Grace

Properties:
- Diverse connections
- Multiple communities
- Mutual friendships
- Gradual growth

Sybil attacker's graph:
Attacker
/ | \ \
/ | \ \
S1 S2 S3 S4 ... S1000

Properties:
- Star pattern (one center)
- All controlled by attacker
- No mutual connections
- Sudden creation

Detection:
- Graph analysis algorithms
- SybilRank, SybilInfer
- Identify star patterns
- Flag suspicious clusters

Verification levels:

Level 1: Seedling
- 3+ connections
- Basic protection
- Can use some apps

Level 2: Sapling
- 10+ connections
- Multiple groups
- More trust

Level 3: Verified
- Sponsored by app
- Sufficient diversity
- Full access

Advantages:

+ No biometrics (privacy-friendly)
+ No central authority (decentralized)
+ Free (no monetary cost)
+ Accessible (just needs app)
+ Revocable (can burn identity and restart)

Challenges:

1. Sybil attacks via social engineering:
- Attacker makes real connections
- Slowly builds legitimacy
- Eventually controls multiple verified IDs
- Detection is statistical, not certain

2. Exclusion:
- Need existing connections
- Cold start problem for new users
- Communities might exclude outsiders
- Network effects favor early adopters

3. Graph analysis complexity:
- Hard to detect sophisticated attacks
- Privacy vs accuracy trade-off
- Requires centralized graph analysis?

4. Adoption:
- Requires network effect
- Chicken-and-egg problem
- Low adoption = easy to game

Scale:

Launched: 2017
Users: ~100,000 (as of 2024)
Connections: 1M+ mutual verifications
Apps integrated: ~20
Unique humans: ~50,000 (estimated verified)

Comparison of Proof of Personhood Systems

Property           Worldcoin    Proof of Humanity    BrightID
─────────────────────────────────────────────────────────────────
Uniqueness ★★★★★ ★★★★☆ ★★★☆☆
Privacy ★★☆☆☆ ★☆☆☆☆ ★★★★☆
Decentralization ★★☆☆☆ ★★★☆☆ ★★★★☆
Accessibility ★★☆☆☆ ★★☆☆☆ ★★★★★
Security ★★★★☆ ★★★☆☆ ★★★☆☆
Cost Free $300 Free
Time Minutes Days/Weeks Hours
Adoption 5M+ 15k 100k
Sybil Resistance Very High High Medium

No winner:

  • Worldcoin: Best uniqueness, worst privacy
  • Proof of Humanity: Balanced, but expensive
  • BrightID: Best privacy, hardest to bootstrap

Ideal (doesn't exist yet):

  • Uniqueness of Worldcoin
  • Privacy of BrightID
  • Decentralization of all three
  • Accessibility of BrightID
  • Security better than all