Pseudo-Random Number Generators (PRNGs) driven by chaotic systems have been widely used in various fields such as security communication for their excellent nonlinear dynamics. However, in practical engineering applications, the digital realization of chaotic systems on finite-precision hardware platforms inevitably creates the problem of dynamics degradation, which in turn directly leads to the lack of stability of the PRNG in long-term operation, and seriously threatens the security and reliability of the system. This paper proposes a method for constructing a PRNG based on nondegenerate chaotic systems driven by explainable degradation detection. The proposed method first constructs a discriminative representation of chaotic sequences through multidimensional feature modeling and mutual information filtering. Then, it achieves high-precision discrimination of degenerative behavior using Extreme Gradient Boosting (XGBoost) and introduces SHapley Additive exPlanations (SHAP) to analyze the model’s decision basis. Finally, it drives dynamic adjustments of the system state based on the detection results, thereby enabling continuous nondegenerative evolution of the chaotic system. The experimental results show that under unbalanced data conditions with extremely scarce samples of degenerative behavior, the proposed method exhibits excellent performance in all metrics, particularly with an AUC as high as 0.9994. This indicates that the model has a very high degree of confidence and discrimination ability in distinguishing between degenerative and normal states. Meanwhile, SHAP explainability analysis reveals the key feature mechanism, enhancing the transparency and credibility of the model’s discrimination. In addition, this PRNG significantly improves the statistical properties of its pseudo-random number sequence and successfully passes the NIST SP800-22 test, which effectively guarantees the long-term stability and reliability of the sequence. This study not only provides new ideas for the design of highly reliable nondegenerate PRNG, but also lays a theoretical and practical foundation for the practical application of chaos theory in the field of secure communications.
Jiang et al. (Wed,) studied this question.