CNN-Enhanced ADPCM Encryption Method for Secured Wireless Communication

The modern long-distance wireless communication systems, suffer of Encryption Cracking (EC) attack, high Signal to Noise Ratio (SNR), transmission latency, Multipath Propagation, Fading and noise Interference. One of the most significant of them is EC attack or man-in-the-middle (MitM) attacks or denial of service (DoS) attacks. As a remedy to the problem, the Adaptive Differential Pulse-Code Modulation (ADPCM) is among the most eminent approaches, to provide data security, narrow band functionality and signal integrity maintenance. This paper proposed the improved adaptation of the ADPCM method for data encryption by adding 4-bit Convolutional Neural Network (CNN) keys. The ADPCM encoder firstly predicts the input signal samples, secondly quantizes the difference between the predicted value and the actual value and finally the step size is dynamically adjusted based on the signal's characteristics. The addition of CNN layer in the feedback loop is used for quantized step detection, pooling layers sampling for 4-bit random number key addition and connecting the samples for a complete sample. The CNN fed the output samples to ADPCM as feedback to help prediction and reconstruct the signal. CNN-ADPCM shows better performance over traditional encoder like ADPCM, DPCM and PCM by a higher compressing rate, slightly better EC resilience and signal reconstruction. The nobility of this research is to improve ADPCM by adding CNN based deep learning approach to increase protection over EC/ MitM/ DoS attacks.