IoT Security Situation Prediction Based on AGWOOptimized BiGRU-ATTN

Abstract

To address the complexity and variability of Internet of Things (IoT) security situation prediction, this paper proposes an IoT security situation prediction model based on an improved Grey Wolf Optimizer (AGWO) optimized Bidirectional Gated Recurrent Unit with an attention mechanism (BiGRU-ATTN). Aiming at the shortcomings of the standard Grey Wolf Optimizer, such as slow convergence and susceptibility to local optima, the algorithm is enhanced through chaotic mapping-based population initialization, a nonlinear adaptive convergence factor, and a fitness-weighted position updating strategy, thereby improving the global search capability and convergence speed. Moreover, a BiGRU network is employed to capture complex temporal correlations in security situation sequences, while an attention mechanism dynamically assigns different weights to key features. Finally, the improved grey wolf optimizer is used to optimize the hyperparameters of the BiGRU-ATTN network. Experimental results demonstrate that, compared with traditional methods, the proposed model achieves superior fitting performance and faster convergence.