IEEE TRANSACTIONS ON SPEECH AND AUDIO PROCESSING, VOL. 10, NO. 8, NOVEMBER 2002 637 Perception-Based Partial Encryption of Compressed Speech Antonio Servetti, Student Member, IEEE, and Juan Carlos De Martin, Member, IEEE Abstract—Mobile multimedia applications, the focus of many forthcoming wireless services, increasingly demand low-power techniques implementing content protection and customer privacy. In this paper low complexity perception-based partial en- cryption schemes for speech are presented. Speech compressed by a widely-used speech coding algorithm, the ITU-T G.729 standard at 8 kb/s, is partitioned in two classes, one, the most perceptually relevant, to be encrypted, the other, to be left unprotected. Two partial-encryption techniques are developed, a low-protection scheme, aimed at preventing most kinds of eavesdropping and a high-protection scheme, based on the encryption of a larger share of perceptually important bits and meant to perform as well as full encryption of the compressed bitstream. The high-protection scheme, based on the encryption of about 45% of the bitstream, achieves content protection comparable to that obtained by full encryption, as verified by both objective measures and formal listening tests. For the low-protection scheme, encryption of as little as 30% of the bitstream virtually eliminates intelligibility as well as most of the remaining perceptual information. Low-power, portable devices could therefore achieve very high levels of speech-content protection at only 30–45% of the computational load of current techniques, freeing resources for other tasks and enabling longer battery life. Index Terms—Encryption, low-power, multimedia security, speech compression, speech perception, speech transmission. I. INTRODUCTION T HE increasing relevance of multimedia applications is placing a great demand on content protection and cus- tomer privacy. Communications can be intercepted, especially over wireless links. Since encryption can effectively prevent eavesdropping, its use is widely advocated. Unfortunately, encryption and decryption are computationally demanding, a severe problem in mobile, portable devices, where power consumption needs to be reduced as much as possible. The need for encryption in wireless systems has led to intense activity aimed at reducing the complexity of encryption algorithms [1]. One solution to the problem of introducing encryption into power-constrained, real-time multimedia applications is partial encryption. Instead of encrypting multimedia signals in their en- tirety, only a subset of the bitstream is protected. The subset is chosen to cause, after encryption, the desired degree of degra- dation after decoding. Proposed partial encryption (sometimes Manuscript received August 7, 2001. This work was supported in part by CERCOM, the Center for Wireless Multimedia Communications, Torino, Italy; http://www.cercom.polito.it. The associate editor coordinating the review of this manuscript and approving it for publication was Dr. Harry Printz. A. Servetti is with the Dipartimento di Automatica e Informatica, Politecnico di Torino, 10129 Torino, Italy (e-mail: servetti@polito.it). J. C. De Martin is with the IEIIT-CNR, Politecnico di Torino, 10129 Torino, Italy (e-mail: demartin@polito.it). Digital Object Identifier 10.1109/TSA.2002.804300 referred to also as selective encryption) techniques have been developed for compressed image and video data. Efficient en- cryption of MPEG compressed video was proposed in [2]–[4]. More recently, the approach was extended to image compres- sion [5], [6]. Speech and audio signals are an essential component of most multimedia applications. Not only speech services are the basis of the huge wireless telephony industry, but speech is also the most important component of advanced audiovisual services such as videoconferencing and news broadcasting. The benefits of partial encryption of speech signals could thus be very signifi- cant. We present partial encryption of speech compressed by the widely used algebraic code-excited linear-prediction technique. The proposed schemes deliver extremely effective content pro- tection and can be straightforwardly extended to a number of international telephony standards based on the same algorithm. The paper is organized as follows. Partial encryption of mul- timedia data is presented in Section II. The first part of Sec- tion III, after a brief review of speech compression techniques, describes partial encryption of G.729 compressed speech. The rest of Section III is devoted to the analysis of effective ways to evaluate performance. Section IV presents the experimental results obtained in the form of objective distortion measures as well as results from formal listening tests to demonstrate the ef- fectiveness of the proposed partial encryption techniques. II. PARTIAL ENCRYPTION OF MULTIMEDIA DATA Digital speech, audio, images and video bitstreams are char- acterized by nonuniform perceptual importance: the effects of errors can be much more pronounced for some bits than for others. So far, perhaps the most significant application of the nonuniform sensitivity of multimedia signals to bit errors has been Unequal Error Protection (UEP) schemes for multimedia transmission over wireless channels. The same principle can also be applied to aid the introduc- tion of content encryption in low-power, wireless multimedia scenarios. Instead of encrypting the multimedia stream (voice, audio, image or video) in its entirety, only a perceptually relevant fraction of the stream is subject to encryption, while the remaining part is transmitted unprotected. Fig. 1 shows the two alternative approaches: full content encryption is shown in Fig. 1(a), while perception-based partial encryption is shown in Fig. 1(b). Encryption of only a fraction of the bitstream lowers the com- putational load, thereby freeing resources for other tasks or, in the case of portable devices, extending battery life. With mul- timedia mobile applications quickly becoming the focus of up- coming wireless services, the possibility of delivering content 1063-6676/02$17.00 © 2002 IEEE