Detection of Low Embedding Rate QIM Steganography Based on Multi-Stage Adjacent Reliance Leaching

Covert communication can be readily achieved by Quantization Index Modulation (QIM) steganography in Voice over Internet Protocol (VoIP) streams, which presents a significant threat to cyber security. The VoIP streams have theoretically unlimited length. The secrecy of steganography can be enhanced by lengthening the carrier and lowering the embedding rate. In the case of low embedding rates (1 % −9 %), the steganography algorithm has little impact on the original VoIP stream. The existing steganalysis methods fail to capture this change effectively and therefore the detection performances are poor. To address this potential risk, this paper focuses on the local change of codeword reliance and and presents a steganalysis approach based on Multi-stage Adjacent Reliance Leaching, termed MARL. The experimental results indicate that our approach is more effective than existing steganalysis methods when the embedding rate is low.