Generative AI-based privacy protection and security management mechanism for distributed financial systems

Abstract

In distributed financial systems, cross-institutional data sharing faces significant security challenges, including sensitive data leakage risks, unauthorized access, and inadequate privacy protection mechanisms. To address these issues, this paper proposes a generative AI-based privacy protection and security management mechanism for distributed financial systems. The proposed method employs a Gradient Penalty Generative Adversarial Network (GP-GAN) to generate privacy-preserving synthetic data that retains the statistical distribution characteristics of original financial data without exposing sensitive information such as account details and transaction records. A blockchain-based fine-grained access control model is constructed to allocate data access permissions according to user roles (ordinary users, financial institution administrators, regulators) and standardize access processes through smart contracts. The Probabilistic Neural Network (PNN) is restructured from risk prediction to abnormal attack detection, enabling real-time identification of side-channel attacks and data injection attacks by analyzing node data transmission characteristics. Furthermore, a game-theoretic weighting method balances multi-node security collaboration and data sharing efficiency. Experimental results demonstrate that the proposed mechanism achieves a privacy protection strength of 92.6%, an access control accuracy of 96.3%, an attack detection recall rate of 89.7%, and an edge node processing latency below 50 ms, effectively ensuring data security in distributed financial systems.