Abstract
Recent widespread utilization of Voice over Internet Protocol (VoIP) has made it a common medium for covert communication via steganography. In order to counter these security threats, several steganalysis approaches have been developed, among which the integration of signal processing and machine learning techniques has enabled the creation of highly accurate steganalyzers. This research study proposes a hybrid method that combines speech signal processing techniques with Artificial Intelligence (AI). Data preprocessing is first applied to audio signals compressed in G.729 codec, which effectively extracts intra-frame features and inter-frame correlations. The resulting data are fed into a deep learning network for the training model to distinguish between cover data and stego data. The evaluation of the implementation findings demonstrates significant improvements in both detection accuracy and computational efficiency. The present technique is assessed for two steganography families, namely Quantization Index Modulation (QIM) and Pitch Modulation Steganography (PMS), as well as their combined application, Heterogeneous Parallel Steganography (HPS). Furthermore, all three techniques: QIM, PMS, and HPS show a superiority in accuracy when compared to conventional methods. Furthermore, during the steganalysis testing phase for 1000 ms audio files, the response test time was less than 5ms with 98.85%, 96.94% and 91.90% accuracy for QIM. PMS and HPS methods respectively. It highlights the high speed of the suggested model in the testing step which is utilized for real-time steganalysis applications.