Parallel Encrypted Transmission Method of Big Data on Financial Risk Hazards Based on Federated Learning

Financial risk big data contains key information such as corporate financial status and market trends, as well as sensitive business secrets and personal privacy. Once the data is leaked or tampered with during transmission, the company will suffer huge economic losses and may even cause a social trust crisis. Therefore, a parallel encrypted transmission method for financial risk big data based on federated learning is proposed. The authentication vector and the permutation matrix are defined to improve the data anti-interference ability, thereby authenticating the real-time transmission network of financial risk big data. The concept of federated learning is integrated to build a secure encryption framework, and the federated learning optimization quintuple model is defined to maximize the long-term cumulative reward and determine the optimal encryption action. By calculating the secure transmission index value and spacing, an encryption transmission model for financial risk big data is constructed to achieve multi-standard homomorphic encryption. The column obfuscation preprocessing, byte substitution encryption and cyclic encryption of financial risk big data are performed, and a specific code source identifier is assigned to the encryption transmission model to achieve parallel encrypted transmission of financial risk big data. Experimental results show that the designed method is highly consistent with the standard case for financial risk big data, and the average throughput under different concurrent connections reaches 425Mbps, which is much higher than other literature, proving that it can improve the security of data transmission.