In a [[Introducing Advanced Federated Learning|federated learning]] setup for fraud detection, banks could add noise to model updates through Substrate pallets, with levels adjusted based on transaction sensitivity. [[ZKP/ZKP Base Layer/Core Concepts/Zero-Knowledge Proofs|ZKPs]] would verify noise applications through Substrate's verification system, ensuring trust and compliance with financial privacy standards. For instance, a bank could contribute updates to a shared fraud detection model, with higher noise applied to updates [[ZKP/Research/Noise/Scalable Noise Addition|involving high-value transactions]] to prevent identification, while ZKPs confirm the noise application through both [[Ethereum Virtual Machine (EVM)]] pallet and native [[Substrate]] verification pallets, ensuring that the aggregated model remains accurate and compliant with privacy regulations. See also: [[Introducing Advanced Federated Learning]]