Attestia — Concepts
How Attestia Works
Attestia turns media verification into an open, incentive-driven trust layer. Every submission gets an immediate authenticity signal from Attestia's built-in deepfake detector; independent attesters then add their own reviews, and the weighted result is anchored on-chain where anyone can verify it.
Protocol flow — hover each node for details
01
Submitter
Submit Media Content
pays: Attestation Credits (AC)
Uses Attestation Credits (AC) to request verification of a media item.
02
Decentralized Storage
Store & Preserve Content
tamper-proof · always available
The media is stored in a decentralized network — no single party controls it, and no one can alter or delete it.
Attester Network
03
Attestia
Deepfake detection
instant · bootstrap attester
Attestia’s built-in deepfake detector acts as the bootstrap attester: it runs immediately when media is stored—no review window—and supplies the first authenticity signal while independent attesters join.
03 · A1
Independent Attester
Retrieve & score
window: 30 min | sealed
03 · A2
Independent Attester
Retrieve & score
attesters: N ≥ 3 | sealed
Independent attesters retrieve the media from decentralized storage, run their own analysis within a 30-minute window, and submit sealed off-chain scores. Scores stay hidden until the window closes to prevent imitation.
04
Score Aggregator
Aggregate & Anchor Consensus
ZK proof → on-chain
All attester scores are aggregated via verifiable computation with a ZK proof. Consensus result anchored on-chain as a final immutable attestation.
05
Smart Contracts
Slash / Reward
aligned: rewarded | outliers: slashed
Attesters whose scores align with consensus are rewarded. Outliers are penalised. Submitter receives the final on-chain verdict.
01Submitter
Submit Media Content
pays: Attestation Credits (AC)
02Decentralized Storage
Store & Preserve Content
tamper-proof · always available
03Attester Network
Attestia
Deepfake detection
instant · bootstrap attester
Independent Attester
Retrieve & score
window: 30 min | sealed
Independent Attester
Retrieve & score
attesters: N ≥ 3 | sealed
04Score Aggregator
Aggregate & Anchor Consensus
ZK proof → on-chain
05Smart Contracts
Slash / Reward
aligned: rewarded | outliers: slashed
Design principles
Why this architecture?
Every structural choice in Attestia answers a specific failure mode of existing verification systems.
Off-chain scores, on-chain anchors
Attester evaluations stay sealed until the window closes — this prevents later participants from copying earlier ones (the "herding" problem). Only the final aggregate goes on-chain via EAS: immutable, public, and composable with any external application.
Cryptoeconomic incentives
Staking, rewards, and slashing are not optional extras. They make honest participation rational and manipulation expensive — by design. Expected loss from dishonest behavior always exceeds the expected gain from any single task.
A market, not a committee
No single model or organization defines ground truth. Attestia combines an initial AI authenticity signal with independent attester reviews, then aggregates them.
As more attesters contribute, the system becomes increasingly decentralized: high-quality contributors gain weight and rewards, while unreliable signals lose influence over time.
As more attesters contribute, the system becomes increasingly decentralized: high-quality contributors gain weight and rewards, while unreliable signals lose influence over time.
Roles & mechanics
Who participates — and how.
Submitters pay for verification; the native attester and independent attesters supply scores under different rules. Wallets register as submitters or independent attesters — not both.
ROLE · 01
Submitters
Media organizations, platforms, or individuals who register assets for verification. They pay using Attestation Credits (AC) — a cost that varies by content type and acts as friction against spam, giving economic weight to each verification request.
pays: Attestation Credits (AC)receives: on-chain verdictROLE · 02
Attestia Native Attester
Attestia's built-in deepfake detector scores every submission immediately as a baseline input to aggregation — not a final verdict. Its weight in consensus is highest when few independent attesters participate and decreases as N grows. It does not stake, earn attester rewards, or face slashing.
runs: on submissionweight: decreases with NROLE · 03
Independent attesters
Decentralized evaluators who retrieve media, run their own analysis — ML models, heuristics, domain expertise — and submit a sealed authenticity score within the time window. They stake collateral: alignment with the combined consensus earns rewards; significant deviation may trigger slashing.
stakes: ETH collateralearns: ETH + ATTArisk: slashingMECHANISM · 04
Off-chain attestations via EAS
Each evaluation is a structured EAS attestation — signed data binding the content hash, asset ID, authenticity score, deepfake-risk score, and a chain-aligned timestamp. Keeping scores off-chain protects attester privacy and prevents herding; the cryptographic commitment still makes them tamper-evident and independently verifiable.
format: EAS off-chainMECHANISM · 05
Native + independent aggregation
Consensus combines the native score xA (weight wA(N) from the participation schedule) with a reputation-weighted average of independent scores. As N increases, wA falls — trust shifts from bootstrap signal to decentralized market judgment.