Information security occupies a very important part of national security. Chaos communication can provide high-level physical layer security, but its harsh claims on the chaotic system parameters of the transmitter and the receiver resulting in reduced synchronization coefficient and more difficult consistent synchronization of point to multipoint networking. In this article, a chaotic synchronization and communication system based on reservoir computing (RC) has been proposed. In this scheme, the trained RC highly synchronized with the emitter acts as the receiver with simplified structure under the premise of ensuring safety. Simultaneously, the cross-prediction algorithm has been proposed to weaken the accumulation effect of prediction synchronization error of RC and facilitate the realization of long-term communication. Furthermore, the tolerance of the system performance to the signal-to-noise ratio with the variations of the mask coefficients has been investigated, and the optimal operation point under the condition of the adjustable number of nodes and leakage rate of RC has been numerically analyzed. The simulation results show that the normalized mean-square error of synchronization of 10-6 magnitude and the bit error rate of decryption at 10-8 level can be obtained. Finally, from the operational perspective, a 100-m short-distance experiment confirms that its communication performance is consistent with the simulation results. We strongly believe that the proposed system offers the opportunity of a new research direction in chaotic secure communications.

In the field of information security, the unpredictability of random numbers plays determinant role according to the security of cryptographic systems. However, limited by the capability of pattern recognition and data mining, statistical-based methods for random number security assessment can only detect whether there are obvious statistical flaws in random sequences. In recent years, some machine learning-based techniques such as deep neural networks and prediction-based methods applied to random number security have exhibited superior performance. Concurrently, the proposed deep learning models bring out issues of large number of parameters, high storage space occupation and complex computation. In this paper, for the challenge of random number security analysis: building high-performance predictive models, we propose an effective analysis method based on pruning and quantized deep neural network. Firstly, we train a temporal pattern attention-based long short-term memory (TPA-LSTM) model with complex structure and good prediction performance. Secondly, through pruning and quantization operations, the complexity and storage space occupation of the TPA-LSTM model were reduced. Finally, we retrain the network to find the best model and evaluate the effectiveness of this method using various simulated data sets with known min-entropy values. By comparing with related work, the TPA-LSTM model provides more accurate estimates: the relative error is less than 0.43%. In addition, the model weight parameters are reduced by more than 98% and quantized to 2 bits (compression over 175x) without accuracy loss.