Research and Implementation of Secure Dual Algorithm-Based Transmission System for Sensitive Data of Large Models in an Open Environment

With the rapid deployment of sensitive data of large models in open environments (SDLMIOE), ensuring secure and reliable transmission has become increasingly vital. This study suggests a secure transmission basis and framework that integrates Reinforcement Learning (RL) for dynamic encryption-parameter optimization and the Cubature Kalman Filter (CKF) for robust state estimation and anomaly detection in SDLMIOE. The RL module adaptively adjusts encryption levels, density strategies, and transmission priorities based on simulated network conditions, while the CKF well filters noisy signals and identifies potential interfering or injection attacks in real time. The complete system is modeled and evaluated using MATLAB simulations, focusing on scenarios linking large model file transfers (up to 1.2 GB) over variable network conditions. Simulation results show that the proposed method attains a 31.5% reduction in average latency, 27.8% increase in throughput, and 99.96% data integrity holding. The CKF-based anomaly detection module achieves a 98.9% detection accuracy with an incorrect alarm rate of 1.3%, and the RL-based optimization leads to a 22% improvement in encryption efficiency compared to fixed-policy systems. These results confirm the effectiveness of the proposed system in protecting sensitive model data during transmission in open and potentially combative environments.