Dynamic Trust Decay: Adaptive Profiling Mechanism for Blockchain Oracles

Decentralized blockchain oracles are critical for bridging on-chain smart contracts with off-chain real-world data. However, existing reputation systems often rely on static cumulative profiling, leading to a phenomenon we define as Reputation Inertia. In this state, an oracle's accumulated historical honesty acts as a buffer that masks potential malicious behavior (Whitewashing attacks). It simultaneously fails to account for benign disagreement during market volatility (Flash Crash scenarios). To address this dilemma, this paper proposes a novel graph-based profiling mechanism utilizing an Adaptive Exponential Weighted Moving Average (AEWMA). Unlike typical static models, our approach introduces a Dynamic Trust Decay factor that regulates the weight of historical reputation based on real-time network volatility. This allows the system to act in a highly sensitive manner to deviations during stable market conditions in order to rapidly detect sleeper cells, whilst also dampening penalty mechanisms during high-volatility events to prevent false positives. We provide a complete Solidity smart contract implementation demonstrating the on-chain feasibility of the AEWMA mechanism with $O(1)$ storage per edge. We validated the proposed method through discrete-event simulations using historical cryptocurrency market data. Experimental results demonstrate that our dynamic approach reduces the Time-to-Detection (TtD) for whitewashing attacks by approximately \textbf{81\%} compared to static baseline (from 9.0 to 1.7 rounds). It also maintained a False Positive Rate (FPR) of near \textbf{0.4\%} even under extreme volatility conditions ($5\times$ standard deviation), compared to 7.1\% for the static approach. These findings suggest that volatility-aware profiling significantly enhances the security and incentive compatibility of decentralized oracle networks.