Proof of Contribution (PoC)

This page documents exactly how YKYR validates every browsing-data upload, assigns a trust score, and commits a tamper-proof proof on-chain. It satisfies VRC-5 (PoC transparency) and VRC-14 (metric-based rewards).

At a glance: an encrypted ZIP → decrypt & hash-check → multi-stage evaluation → score normalisation → on-chain ProofResponse.

1 · End-to-End Flow

2 · Key Algorithms

2.1 Authenticity Check (Honesty Verification)

Purpose – Verify that the client’s self-reported summary matches what the raw data really shows.
Concept – Re-derive three simple ground-truth metrics from the decrypted browsing entries:

Count of URLs visited
Exact list of per-page seconds spent
Points total based on a public formula (URLs + minutes) × bonus

The server then compares these freshly calculated numbers to the ones the client claimed. If every single value lines up, authenticity is perfect (1.0). Any mismatch—no matter how small—drops authenticity below perfect, signalling possible tampering or error.

2.2 Correctness Evaluation

Purpose – Decide whether a batch of raw browsing records is usable at all.
Concept – Treat each record like a quick inspection at the factory gate:

Does it have the essential labels? (a URL and a time-spent value)
Is the URL even a real URL?

If at least half of the incoming boxes pass those two checks, the whole shipment is declared “correct”; otherwise it’s rejected. Empty shipments are automatically rejected.

2.3 Quality Evaluation

Purpose – Put a 0-to-1 quality grade on a browsing session that already passed the basic correctness test.
Concept – Think of two lenses, each looking for signs of genuine, attentive browsing:

Time behaviour (60 % weight).
Rewards a mix of reasonable visit lengths, penalises a flood of ultra-short clicks, suspiciously identical visit times, or one-page marathons.
Bot detection (40 % weight).
Flags mechanical patterns—e.g., three quick, tiny visits in a row or nearly identical time-spent values—that humans rarely produce.

Each lens produces a partial score. Weighted, summed, clamped to 0-100, then normalised to 0-1, the result expresses “How human-like and meaningful was this set of visits?”

2.4 Sigmoid Normalisation

def sigmoid(x, k=constants.K, x0=constants.X0):
    """
    Maps raw 0-1 quality into a curve that emphasises mid-range differences.
    Default: k=6, x0=0.5 (check constants.py for actual values)
    """
    z = k * (x - x0)
    return 1 / (1 + math.exp(-z))

Threshold VALIDITY_THRESHOLD marks moderate authenticity (MODERATE_AUTHENTICITY_THRESHOLD).

3 · ProofResponse Structure

{
  "dlp_id": 10,
  "ownership": 1.0,
  "authenticity": 1.0,
  "integrity": true,
  "correctness": true,
  "quality": 0.82,
  "attributes": {
    "label": "high"
  },
  "score": 0.91,
  "valid": true,
  "metadata": {
    "dlp_id": 10,
    "valid": true
  }
}

Only valid==true files accrue rewards; otherwise contributors can resubmit.

4 · Valid-Data Thresholds

Valid Calculation:

# From proof.py line 60-61
valid = ownership and authenticity and integrity and correctness and (final_score >= constants.MODERATE_AUTHENTICITY_THRESHOLD)
# where MODERATE_AUTHENTICITY_THRESHOLD should be VALIDITY_THRESHOLD

5 · Schema Requirements

Input Data Structure

{
  "author": "0x...",
  "signature": "0x...",
  "random_string": "0x...",
  "data_hash": "0x...",
  "data": {
    "browsingDataArray": [
      {
        "url": "https://example.com",
        "timeSpent": 5000  // milliseconds
      }
    ],
    "evaluationMetrics": {
      "url_count": 10,
      "timeSpent": [5, 3, 7, ...],  // seconds
      "points": 123
    }
  }
}

Required Fields (from constants.py)

• Each browsing entry must have: url and timeSpent
• URLs must be valid (checked by is_valid_url())
• Time values in milliseconds, converted to seconds for metrics

6 · How to Reproduce Locally

# Prepare input data
# Place your encrypted ZIP file in the input directory
# The ZIP should contain the JSON structure shown in section 5
 
# Build & run proof container
$ docker build -t ykyr-proof .
$ docker run -v $PWD/input:/input -v $PWD/output:/output \
           -e DLP_ID=51 \
           -e SIGNED_MESSAGE=0x... \
           ykyr-proof
 
# Inspect proof
$ jq . output/proof_response.json

Note: The input directory should contain a .zip file with the encrypted data structure. The proof system will automatically find and process the first ZIP file in the directory.

The JSON can then be submitted with addRefinementWithPermission() to testnet (Moksha) or mainnet.

7 · Label Classification

Based on the sigmoid score (from label_browsing_behavior() function):

• Low trust: score < 0.4
• Moderate trust: 0.4 ≤ score < 0.8
• High trust: score ≥ 0.8

These labels are stored in attributes.label and can be used for filtering or reputation systems.