An Improved InceptionTime Model for Pattern Synchronization in Covert Communication

The rapid development of signal recognition technology has brought security risks to communication. How to counter various types of signal recognition has always been a key issue in the field of covert communication. With the iteration of blind detection technology, traditional signal concealment schemes can no longer meet the need for anti-interception. Therefore, this paper implements a transform domain modulation of signals based on butterfly operation transformation, mapping signals to the transform domain and making the signals monitored by third parties noise. Based on this type of data set, this paper proposes a time series classification network based on InceptionTime and Omni-Scale, which can use convolution kernels of different sizes to extract multi-scale receptive fields and successfully reduce the error rate of signal classification from 1.2% to 0.7%, improving the reliability of covert communication.