Scrambler type recognition based on shallow neural networks

In this paper, we propose a scrambler type recognition method using shallow neural networks to identify the scrambling type under the misleading code condition. Firstly, the inter-correlation properties of scrambled code elements are derived based on the scrambling principle and the characteristics of error-correcting codes, then a biased autocorrelation function is composed, and the superiority of the biased autocorrelation function over the unbiased autocorrelation function is demonstrated, and the two are combined as the real part and the imaginary part respectively to form the set of correlated feature samples. To enhance the extraction capability of correlated features, a correlated feature extraction network is constructed, and the purpose of scrambler type identification is accomplished by learning and training the set of correlated feature samples. Simulations show that the algorithm has a wider range of application and can identify a variety of scrambler types, while the anti-bug performance is stronger, laying the foundation for further scrambler parameter identification.