Fraud Protection

Security measures, fraud detection mechanisms, penalty tiers, and the appeal process for the Data Bounty Board.

Overview

The Data Bounty Board employs multiple layers of fraud protection to ensure fair payouts for legitimate contributors and protect bounty creators from low-quality or fraudulent submissions.

Detection Mechanisms

Automated Detection

┌─────────────────────────────────────────────────────────────────┐
│                    Fraud Detection Pipeline                      │
├─────────────────────────────────────────────────────────────────┤
│                                                                 │
│  Submission                                                     │
│      │                                                          │
│      ▼                                                          │
│  ┌─────────────────┐                                           │
│  │  Hash Check     │ ← Detect exact duplicates                 │
│  └────────┬────────┘                                           │
│           │                                                     │
│           ▼                                                     │
│  ┌─────────────────┐                                           │
│  │  Perceptual     │ ← Detect near-duplicates                  │
│  │  Hash Check     │   (cropped, resized, filtered)            │
│  └────────┬────────┘                                           │
│           │                                                     │
│           ▼                                                     │
│  ┌─────────────────┐                                           │
│  │  AI Detection   │ ← Detect AI-generated content             │
│  └────────┬────────┘                                           │
│           │                                                     │
│           ▼                                                     │
│  ┌─────────────────┐                                           │
│  │  Metadata       │ ← Detect spoofed/fake metadata            │
│  │  Analysis       │                                           │
│  └────────┬────────┘                                           │
│           │                                                     │
│           ▼                                                     │
│  ┌─────────────────┐                                           │
│  │  Behavioral     │ ← Detect suspicious submission patterns   │
│  │  Analysis       │                                           │
│  └────────┬────────┘                                           │
│           │                                                     │
│           ▼                                                     │
│  Pass / Flag / Block                                           │
│                                                                 │
└─────────────────────────────────────────────────────────────────┘

Hash-Based Detection

SHA-256 Exact Match

Every submission is hashed using SHA-256. Identical files are instantly detected:

// How files are hashed
async function hashFile(file: ArrayBuffer): Promise<string> {
  const hashBuffer = await crypto.subtle.digest('SHA-256', file);
  const hashArray = Array.from(new Uint8Array(hashBuffer));
  return hashArray.map(b => b.toString(16).padStart(2, '0')).join('');
}

Perceptual Hashing

For images and videos, perceptual hashes detect near-duplicates:

Modification Detected	Example
Cropping	Trimming edges
Resizing	Upscaling/downscaling
Rotation	90/180/270 degrees
Flipping	Mirror images
Color adjustments	Brightness, contrast
Minor filters	Instagram-style filters
Compression	Re-saving as lower quality

AI Content Detection

Multiple AI detection systems analyze submissions:

System	Detection Type
Statistical analysis	Pixel distribution patterns
Artifact detection	AI-specific generation artifacts
Model fingerprinting	Signatures of known AI models
Metadata analysis	Missing/fake EXIF data
Style analysis	Characteristic AI styles

AI detection continues to improve. Content that passes today may be flagged in future scans.

Behavioral Analysis

The system monitors submission patterns:

Pattern	Suspicion Level
Submissions faster than humanly possible	High
Identical timestamps across submissions	High
Same IP with multiple accounts	High
Unusual geographic patterns	Medium
Consistent quality across unrelated topics	Medium
Batch submissions with sequential filenames	Low-Medium

Cross-Platform Checks

Reverse image/video searches detect:

Stock photos
Previously published content
Content from other data platforms
Stolen social media content

Penalty Tiers

Tier Overview

Tier	Reputation Loss	Cooldown	Earnings Impact	Review Required
Warning	0	None	None	No
Minor	-5	1 day	None	No
Moderate	-20	7 days	Forfeited	No
Severe	-50	30 days	Forfeited	Yes
Permanent Ban	-100	Forever	Forfeited	Yes

Warning

Trigger: First minor quality issue

Consequences:

Submission rejected
No reputation penalty
Educational notification sent

Example violations:

Slightly below quality threshold
Missing optional metadata
Minor format issues

Minor Penalty

Trigger: Repeated quality issues, minor rule violations

Consequences:

-5 reputation points
24-hour cooldown from submissions
Submission rejected

Example violations:

Multiple quality rejections
Incorrect metadata (unintentional)
Submitting to wrong category

Moderate Penalty

Trigger: Significant violations, pattern of issues

Consequences:

-20 reputation points
7-day cooldown
Pending earnings forfeited
Account flagged for monitoring

Example violations:

PII in submission
Misrepresenting content
Attempting duplicate submissions
Minor consent violations

Severe Penalty

Trigger: Serious rule violations, fraud attempts

Consequences:

-50 reputation points
30-day suspension
All pending earnings forfeited
Manual review required to return

Example violations:

AI content with deception
Stolen content
Multiple accounts
Hate speech or explicit content
Serious consent violations

Permanent Ban

Trigger: Critical violations, repeated severe violations

Consequences:

Account terminated
All earnings forfeited
Wallet address blacklisted
Cannot create new accounts

Example violations:

CSAM (reported to authorities)
Organized fraud
Identity theft
Threats or harassment
2+ severe violations

Wallet Blacklist

How It Works

Wallets associated with banned accounts are permanently blocked:

Banned Account
    │
    ├── Primary wallet → Blacklisted
    │
    ├── Associated wallets → Blacklisted
    │   (from same IP, linked transactions)
    │
    └── Future attempts → Blocked on connect

What Gets Blacklisted

Wallet Relationship	Blacklisted
Primary connected wallet	Yes
Wallets used from same IP	Yes
Wallets with direct transfers to/from primary	Flagged for review
ENS names pointing to blacklisted addresses	Yes

Blacklist Persistence

Blacklist is permanent
Cannot be removed through appeal (for permanent bans)
Shared across all PRIV Protocol services
Used by partner platforms (optional data sharing)

IP Tracking

What We Track

Data Point	Purpose	Retention
IP address (hashed)	Fraud prevention	90 days
Geolocation (country only)	Geographic restrictions	90 days
Connection timestamps	Pattern analysis	30 days
Device fingerprint	Multi-account detection	30 days

Privacy Safeguards

IP addresses are immediately hashed. We cannot and do not see your actual IP address, only a one-way hash used for fraud detection.

How IP hashing works:

// IP is hashed with a secret salt
const ipHash = await hashWithSalt(ipAddress, IP_HASH_SALT);

// Only the hash is stored
// Original IP cannot be recovered from hash
database.store({ ipHash, timestamp });

What we DON'T do:

Store raw IP addresses
Share IP data with bounty creators
Track your browsing outside PRIV
Sell location data

Appeal Process

When You Can Appeal

Violation	Appealable
Warning	No (no penalty to appeal)
Minor penalty	Yes
Moderate penalty	Yes
Severe penalty	Yes
Permanent ban	Limited cases only
CSAM violations	No (reported to authorities)

How to Appeal

Step 1: Review the Violation

Check your notification for:

Specific violation cited
Evidence provided
Deadline for appeal (7 days)

Step 2: Gather Evidence

Collect supporting documentation:

Original files (RAW, source files)
Proof of creation (timestamps, project files)
Receipts/licenses (if using licensed content)
Context for misunderstandings

Step 3: Submit Appeal

# API: Submit appeal
curl -X POST https://api.privlabs.io/v1/b2b/appeals \
  -H "Authorization: Bearer your_token" \
  -H "Content-Type: application/json" \
  -d '{
    "violationId": "vio_x1y2z3",
    "reason": "False positive - content is original",
    "explanation": "This photo was taken by me on January 15, 2026...",
    "evidenceUrls": [
      "https://example.com/raw_file.arw",
      "https://example.com/metadata_screenshot.png"
    ]
  }'

Step 4: Wait for Review

Appeal Type	Review Time
Minor penalty	24-48 hours
Moderate penalty	3-5 business days
Severe penalty	5-10 business days
Permanent ban	10-14 business days

Appeal Outcomes

Outcome	Result
Upheld	Original penalty stands
Reduced	Penalty lowered (e.g., severe -> moderate)
Overturned	Penalty removed, reputation restored
Escalated	Requires additional review

What Helps Appeals

Clear evidence of original creation
Timestamps that predate submission
RAW or source files
Logical explanation for misunderstanding
Clean account history

What Hurts Appeals

No supporting evidence
Contradictory claims
History of violations
Hostile or threatening language
Lies or fabrications (makes it worse)

Reporting Fraud

How to Report

If you suspect fraud by another user:

# API: Flag submission
curl -X POST https://api.privlabs.io/v1/b2b/submissions/sub_x1y2z3/flag \
  -H "Authorization: Bearer your_token" \
  -H "Content-Type: application/json" \
  -d '{
    "reason": "duplicate_content",
    "explanation": "This image appears to be from stock photo site XYZ...",
    "evidenceUrls": ["https://stockphoto.com/original-image-123"],
    "originalHash": "a1b2c3..."
  }'

Report Reasons

Reason	Use When
`duplicate_content`	Content copied from elsewhere
`inappropriate_content`	Rule violations
`low_quality`	Doesn't meet requirements
`wrong_format`	Mismatched content type
`consent_violation`	Includes non-consenting parties
`copyright_violation`	Stolen/copyrighted content
`fraud_suspected`	General fraud suspicion
`other`	Other issues

Whistleblower Rewards

Accurate fraud reports may earn rewards:

Report Outcome	Reward
Minor fraud confirmed	1 PRIV
Moderate fraud confirmed	5 PRIV
Major fraud ring uncovered	50+ PRIV
False report	Warning to reporter

Abusing the report system with false or malicious reports will result in penalties.

Security Best Practices

For Contributors

Use your own content - Never submit others' work
Keep originals - Store RAW/source files for proof
Document creation - Screenshots of creation process help
One account only - Multiple accounts will be detected
Honest metadata - Don't fake timestamps or locations

For Bounty Creators

Enable approver voting - Worth the premium for quality
Set appropriate requirements - Clear specs reduce fraud
Monitor submissions - Review samples regularly
Report issues quickly - Faster detection means less loss
Use test bounties - Verify system before large budgets

FAQ

Why was I flagged when my content is original?

False positives happen. Common reasons:

Similar content exists elsewhere
AI detection isn't perfect
Metadata appeared suspicious

Submit an appeal with your original files.

How long does a cooldown last?

Penalty	Cooldown
Minor	24 hours
Moderate	7 days
Severe	30 days

Cooldowns start when the penalty is applied, not when you see it.

Can I create a new account after a ban?

Permanent bans include wallet blacklisting. New accounts with the same wallet, IP pattern, or device fingerprint will be detected and banned.

What happens to my pending earnings?

Penalty Level	Earnings Status
Warning/Minor	Retained
Moderate+	Forfeited

Are false reports penalized?

Intentionally false reports result in warnings, then penalties. Honest mistakes are not penalized.

Next Steps

Content Rules - Avoid violations
Approver Voting - Help fight fraud as a staked approver
Pricing - Understand fee structure

Fraud Protection

On this page