Paper Title
Blockchain Security, Cross-Chain Systems, Graph Neural Networks, Double-Spending, Reputation Systems

Abstract
Cross-chain systems enable blockchain interoperability but introduce significant security vulnerabilities, particularly double-spending attacks that exploit asynchronous validation and inconsistent consensus mechanisms. This paper presents a novel hybrid framework combining Graph Neural Networks (GNNs) for real-time detection with a Dynamic Cross-Chain Penalty and Reputation Sharding (DCPRS) mechanism for prevention. Our approach leverages Graph Attention Networks to analyze transaction patterns and implements economic disincentives through cross-chain smart contracts. Experimental validation on a simulated cross-chain environment demonstrates 93.2% detection accuracy with 4.2% false positive rate, significantly outperforming traditional consensus-based approaches. The DCPRS framework reduces repeat attack attempts by 78% through adaptive penalty mechanisms and reputation-based node isolation. Keywords - Blockchain Security, Cross-Chain Systems, Graph Neural Networks, Double-Spending, Reputation Systems