BiteBack

BITEBACK is an on-chain consumer-protection and collective-resolution system. A user reports one confirmed suspicious payment by signing its transaction details. BITEBACK verifies the report, applies a deterministic merchant policy, and uses indexed blockchain data to find every wallet affected by the same violation. Instead of forcing each victim to open a separate ticket, the protocol receives one evidence-backed collective case with the affected wallets, valid charges, excess charges, and total compensation. The merchant must explicitly approve the frozen case before funds can move. BITEBACK also monitors Base, Optimism, and Arbitrum for unusual fees, TPS, failure rates, gas utilization, sender activity, and block timing using chain-isolated Bayesian models. These probabilistic alerts can warn wallets but can never create disputes, claims, decisions, or payouts without confirmed transaction evidence.

BITEBACK is a TypeScript application built with Hono, ethers, the Hedera SDK, The Graph Substreams, 0G, and a browser-based operator dashboard deployed on Vercel. The Graph supplies compact, finalized blockchain data for transaction discovery and five-minute anomaly metrics. Deterministic rules group repeated-payment violations by merchant, token, amount, wallet, and UTC window. 0G Compute converts published merchant terms into an inspectable policy, while 0G Storage preserves the canonical evidence package and allows its SHA-256 hash to be independently verified. After reviewing the evidence, the merchant signs an explicit acceptance for the exact incident, evidence hash, and payout amount. Hedera then distributes the authorized compensation and records the evidence, decision, and settlement audit trail through HCS, without deploying Solidity contracts on Hedera. The independent anomaly engine uses Normal–Inverse-Gamma posterior prediction, seasonal weekday/weekend baselines, conformal calibration, false-discovery correction, quarantine, and RPC heartbeat agreement. It includes a prequential yearly backtesting framework designed around precision-recall and detection latency rather than inflated point-adjusted scores.