One of the main challenges we faced was ensuring the scalability of real-time security monitoring across multiple blockchain platforms like Ethereum, Polkadot, and Arbitrum. The need to process vast amounts of on-chain data without latency required optimizing algorithms and leveraging decentralized computation. Another hurdle was designing a fraud detection model adaptable to evolving attack patterns. We overcame this by integrating machine learning models trained on historical blockchain data and implementing dynamic updates to adapt to new threats. Collaboration with security experts also helped refine our approach to on-chain forensics.