Complexity Reduced Multi-tier Perceptual Based Partial Encryption for Secure Speech Communication R. H. Laskar, F. A. Talukdar, B. Bora, K. S. P. Fernando, J. Anthony and L. Doley Department of Electronics & Communication Engineering National Institute of Technology, Silchar, Silchar-10, Assam, India Email: {rabul18, josephnits}@yahoo.com; {fatalukadar, bhargovnits, kspfernando, lokhinits207}@gmail.com, Abstract—Present day secure speech communication mainly deals with providing maximum security at the cost of minimum complexity. While doing so, they are mainly looking forward to the speech coders for compressing the speech signal thus minimizing the complexity. In this paper an algorithm is proposed that aims at exploiting the basic characteristics of speech signal, while designing such system with reduced complexity. As observed from different speech signals, silent zones are present which comprises of about 30% to 40% of the whole speech. These zones do not bear any speech information. Processing these parts leads to significant increase in the computational complexity and also needs more frames to be transmitted. These can be avoided by removing such zones from the original speech signal. In case of secure speech communication, modern day encryption system does not provide the user, full flexibility in choosing the level of security. Those encryption algorithms mainly consist of single level of encryption. A scheme is proposed where up to three tiers of security levels can be achieved and the user is given the full flexibility in choosing the level of security one desires to impart to the speech communication system. Removal of the silent zones reduces the complexity which may circumvent the increasing computational complexity while implementing the multi-tier encryption algorithm. Keywords- speech characteristics; silent zones; encryption; speech transmission; secure speech communication. I. I NTRODUCTION The increasing demand of multimedia applications in com- munication system has paved the way for secure communica- tion. This is essential to overcome unauthorized modifications and unwanted disclosure while transmitting speech and other data, especially in wireless channels [1]. But unfortunately encryption and decryption techniques are computationally de- manding, a severe problem in mobile, portable devices where power consumption needs to be reduced as much as possible. Therefore a technology which uses less power in real-time application and also provides maximum security is highly anticipated. One solution to this problem of incorporating encryption in power constrained, real time multimedia appli- cation is to provide a perceptual based partial encryption [1] which is realized by encrypting the subset of the speech signal which is perceptually more relevant whereas the other part having less significant is left unprotected. To reduce power consumption a speech coder [2] is used to compress the speech signal first then partial encryption of the compressed speech signal is done. Compression [3] improves bandwidth utilization which is also a major concern in wireless communication as the number of end users utilizing the same channel is increasing rapidly day by day. In this paper a scheme is proposed where further optimization is done in both power consumption as well as time complexity by pre-processing the speech signal prior to feeding it to the speech coder. During this pre-processing phase, the speech signal is analyzed in order to find the silent zones. About 30% to 40% of the input speech signal consist of the silent zones which only carry information such as background noise and the noise generated by the systems [4] present in the transmitter side. So by segregating these zones from the original speech, the computational load on the coder is reduced significantly. In order to keep the naturalness of the speech signal intact at the receiver side, a few silent frames are transmitted and are replicated at the receiver side according to the positions of the silent zones in the original speech signal. A multi-tier encryption system is also proposed in order to maximize the security aspects and the user is given full flexibility while choosing the level of encryption as per his/her application demands. Multi-tier encryption will provide rigid security but at the cost of higher complexity. Because of which con- ventional encryption systems [5] are mainly of single level. Reduction in computational complexity by removing silent zones helps in realizing multi-tier encryption system for real time applications. The paper is organized as follows. The block diagram of the proposed model for secure speech communication is given in the Section II. Pre-processing of the speech signal to remove silent zones is discussed in Section III. In the first part of Sec- tion IV, a brief introduction on speech compression by speech coder (GSM FR) is presented and rest of the section deals with the implementation of the multi-tier encryption system which provides a rigid security to the speech communication. The experimental results that demonstrates the effectiveness of the proposed secure speech communication system is also presented in this section. Section V summarizes the work carried out and shows the future direction of work. II. PROPOSED ENCRYPTION SYSTEM The conventional perceptual based partial encryption system is discussed in [1]. In this method no pre-processing or post- processing of the speech signal is done and the offered security is minimal as it uses only a single level of encryption. The 978–1–4244–4547–9/09/$26.00 c  2009 IEEE TENCON 2009 1