communications are increasingly used in a way that they are scrambling and the combination of scrambling frequency and time domain.

Transcription

1 Journal of Advances n Computer Research Quarterly pissn: x eissn: Sar Branch, Islamc Azad Unversty, Sar, I.R.Iran (Vol. 7, No. 3, August 2016), Pages: Audo Sgnal Encrypton Based on Permutaton Relatons and Resd due Number System Fereshteh Ghanbar Adv, Mohammad Mehrna M.Sc. Computer Systems Archtecture, Dezfoul Branch, Islamc Azad Unversty, Dezfoul, Iran qermezkon@gmal.comm Receved: 2016/02/28; Accepted: 2016/05/14 Abstract Ths paper presentss a new audo encrypton algorthm based on permutaton and resdue number system. In the current approach, sgnal samples are combned wth resdue number system through permutaton; therefore, a specfc nose s generated. Then, the audo sgnal s combned wth the nose, generated by the algorthm, to produce a slencee sgnal. For ths purpose, the algorthm uses a predefned permutaton format and a modular set n the form of {2 n -1, 2 n, 2 n +1} to encrypt the man sgnal. Consequently, the proposed algorthm converts the nput sgnal nto a slence sgnal and ncreases ts securty. The sgnal-to-nose very low values of rato (SNR) was used to evaluate the algorthm. The results ndcated SNR pertanng to audo sgnals n the proposed method. Keywords: Encrypton, Decrypton, Permutaton SNR, Resdue Number System, Audo Sgnal, 1. Introducton Voce s the ablty to speak or the act of speakng. An audo sgnal s comprsed of a number of bts. Ths sgnal contans postve and negatve values. Audo communcatons are ncreasngly used n a way that they are becomng more vulnerable; therefore, t s really necessary to provde them wth more securty. Varous encrypton algorthms are used to provde the securty of audo sgnals. In ths regard, there are fve general encrypton methods such as scrambl lng tme doman, scramblng ampltude, 2D scramblng and the combnaton of scramblng frequency and tme doman. An audo sgnal s encrypted n dgtal or analog formats [1]. Generally, the objectve of speech encrypton s to perform some reversble operatons on a pece of speech n a way that t wll be totally ncomprehensble to an unautho orzed lstener. Threee mportant crtera for ratng encrypton algorthms are as follows: The ablty of algorthm to generate an encrypted speech whose comprehensble resdue s lttle. The effectveness of encrypton and decrypton retreved nformaton on the sde of the desred recpent. The securty of algorthm aganst cryptanalyss attacks processs on the qualty of 67

2 Audo Sgnal Encrypton Based on F. Ghanbar Adv, M. Mehrna The problem that encrypton experts are faced wth s to desgn encrypton systems that dstort speech sgnals havng so much redundancy that t would be mpossble to retreve useful nformaton from them. Usng a predefned permutaton mechansm and a resdue number system wth modules n such a format as {2 n -1, 2 n, 2 n +1}, the proposed approach encrypts audo sgnals. The encrypted sgnal wll be lke a slence sgnal wth no loss of qualty durng decrypton. Employng the proposed algorthm, audo fles can be encrypted n order to save audo messages and transmt them on the Internet. Ths system has two promnent advantages. Frst, the encrypted sgnal wll be heard lke a slence sgnal. The second advantage s that relocatng modules, changng ther values, and changng the permutaton pattern can result n so many states n the encrypton algorthm. Usng SNR, the proposed algorthm was evaluated. It presented a hgh level of securty wth approprate qualty. The current paper s organzed as follows: Secton 2 presents an ntroducton to the resdue number system. In Secton 3, some nstances of the prevous algorthms are presented on voce encrypton. The encrypton process s explaned n Secton 4. The results of analyzng and evaluatng the performance of the algorthm are presented n Secton 5. Fnally, the concluson s dscussed n Secton Resdue Number System The resdue number system s a non-weghted and unconventonal number system that supports parallel computatons, the lmted dstrbuton of carry, low power consumpton, and secure communcatons. It s used n applcatons employng addton, multplcaton, subtracton, and multplcaton by a range of numbers. The resdue number system s characterzed wth a set of modules. If all the modules are co-prme, the system s optmal. The hybrd of other systems can be used to optmze ths system so that the selecton of modular sets would become easer, and the dsplay range extended. For ths purpose, the hybrd of the resdue number system, multlevel resdue number system, and the Van Hunt resdue number system can be employed. Moreover, t s suggested that the redundant resdue system be used to establsh and ncrease system securty. The resdue number system s characterzed wth a set of modules lke {m 1, m 2,, m n } n whch all the modules are postve ntegers. If all n modules are co-prme, the system has the largest possble dsplay range equal to [α, α+m) n whch α s an nteger and M s calculated as follows: = n M m (1) =1 The nteger X (α X < α + M) s a unque representaton n the resdue number system. It s ndcated wth a set of remanders lke (x 1, x 2, x 3,, x n ) n a way that: x = X mod m, = 1,2,..., n (2) Usng the Chnese Remander Theorem, X s obtaned on the remander set {m 1, m 2,,m n } as follows: n n = 1 ( ) X = ( xn M ) (3) m m 68

3 Journal of Advances n Computer Research (Vol. 7, No. 3, August 2016) = n M m (4) =1 M 1 M = N = ( M ) = 123 K n m m In whch s defned as the multplcatve nverse of M on module m. Gven the characterstcs of the resdue number system, t s used n computatonal applcatons such as dgtal sgnal processng system, dgtal flters, codng, encrypton systems, dgtal communcatons, ad hoc networks, nformaton storage and retreval, and error detecton and correcton. Moreover, snce the computatons are done separately on remanders n ths system, f an error occurs on one of the remanders, others are mmune to ts mpact. 3. Prevous Works Encrypton may be performed n ether a dgtal or analog way. In analog encrypton, whch s also called speech scramblng, the operatons are done on the speech samples. In analog scramblng, there s no need for a modem or speech compresson to transmt. The qualty of retreved speech does not depend on ts language. Usng an approprate nterface, these scramblng methods can be easly connected to the avalable analog channels such as telephone, satellte, and moble communcaton lnks. The dgtal encrypton dgtalzes the nput speech sgnal frst. Then the dgtalzed sgnal s compressed n order to generate a bt sequence wth an approprate bt rate. Usng modem, ths bt sequence s encrypted and transmtted to the channel. Analog scramblng s dvded nto several classes: - Tme scramblng - Frequency scramblng - Scramblng wth pseudo-nose sequences - Scramblng wth maxmal-length sequences - Scramblng wth gold sequences - Scramblng wth pseudo-gold sequences - Scramblng wth Barker sequences - Scramblng wth pseudo-barker sequences - Scramblng wth Kasam sequences There are also 7 dfferent styles defned for the cryptography of audo sgnals. 1- There s no need for cryptography. 2- The qualty of speech s not hgh whle the requred securty level s so low. 3- The requred securty s low whle power consumpton and tme delay are very low. (5) 69

4 Audo Sgnal Encrypton Based on F. Ghanbar Adv, M. Mehrna 4- The requred securty s hgher than Case 3 whle more power and tme are consumed. 5- The requred securty s average, and the consumed tme and power are hgh. 6- The securty s more than Case The securty and tme delay are hgher. 8- The sender emphaszes on the hgh securty of ts audo streamng wthout consderng other factors (power consumpton and tme delay). Two nstances of audo encrypton algorthms based on the abovementoned mechansms are explaned n what follows. In [1], a four-level method was presented for audo sgnal encrypton. In ths method, the output of each level s used as the nput of the next. In algorthm [1], changng the formaton of bts at the frst level, usng a random generator at the second level, usng t agan at the thrd level, and, fnally, usng the ascendng formaton of the ampltude feld at the fourth level result n the generaton of an encrypted sgnal lke a contnuous beep. authors n [3], propose a smple and effectve detecton algorthm whch detects the tme-frequency (TF) ponts occuped by only a sngle source for each source. The detecton algorthm dentfes the sngle source ponts by comparng the normalzed real and magnary parts of the TF coeffcent vectors of the mxed sgnals, whch s smpler than prevously reported algorthms. Then we propose a modfed smlarty-based robust clusterng method (MSCM) to estmate the number of sources and the mxng matrx usng these detected sngle source ponts. Expermental results show the effcency of the proposed algorthm, especally n the cases where the number of sources s unknown. In [2], the Independent Component Analyss (ICA) of the audo sgnal was used to propose a method for encrypton. In ths method, the audo sgnal s dvded nto several parts. The sgnal dvson process s n a way that each part has an equal number of samples that wll be ndependent from each other so that the successful use of ICA would be guaranteed. In the next step, the correlaton matrx s generated, and then each part s encrypted through the process of non-correlaton, named whtenng n ths paper. The whtenng process uses the Prncple Component Analyss (PCA) of the dmensonalty reducton algorthm. the authors n [11], present an RNS algorthm resolvng the Closest Vector Problem (CVP). Ths algorthm s partcularly effcent for a certan class of lattce bass. It provdes a full RNS Baba round-off procedure wthout any costly converson nto alternatve postonal number system such as Mxed Radx System (MRS). An optmzed Cox-Rower archtecture adapted to the proposed algorthm s also presented. The man modfcatons resde n the Rower unt whose feature s to use only one multpler. Ths allows freeng two out of three multplers from the Rower unt by reusng the same one wth an overhead of 3 more cycles per nner reducton. An analyss of feasblty of mplementaton wthn FPGA s also gven. The authors n [12] demonstrate the effcency of approach usng 32 chosen egenvalues n the key generaton algorthm. The egenvalues are derved from a reference mage. Throughout the processes on the mage data, we use modular arthmetc to ensure that computatons wth the resultng RNS become very effcent. Moreover, the approach consders mage n a dvded matrx doman and fnally combnes all ndependent cryptographc operatons as encrypton s a one-to-one 70

5 Journal of Advances n Computer Research (Vol. 7, No. 3, August 2016) mappng. Ths deals wth the possblty of havng any pxel value ll-stored or wrongly receved at the recever end, wthout affectng the decrypton process. However, the fnal recovered mage wll dffer by a neglgble amount. the authors n [13], present a speech encrypton algorthm, whch can penetrate compresson encoder. The algorthm scrambles the speech sgnal n frequency doman based on FFT transformaton doman n the sender, and t decrypts and reconstructs the sgnal n the recevng end. The encrypted speech sgnal stll has the characterstcs of speech, the resdual ntellgblty of the speech s low enough and the speech bandwdth does not be wdened. Scramble matrx can meet the needs of speech transmsson securty and acheve one key for one secret communcaton flexbly. End-to-end communcaton can avod the common eavesdroppng and theft, and has a relatvely broad prospects for development. In [14], a robust Audo steganography technque s proposed by randomzng and dynamcally changng the embeddng sequence. Advance Encrypton standards s used for provdng addtonal securty and robustness to the algorthm and tested for the 30 speech fles. The addton of cryptography n steganography ncreases the robustness and ntroduces a hgher level of securty snce the key s requred to decrypt the secret message. To evaluate the qualty of the stego fles SNR and Correlaton coeffcents are calculated and an expermental test s also performed over 10 lsteners to dentfy the change n the orgnal audo and cover audo. the authors n [15], presents an analyss of the tradeoff between securty level and compresson rato n real-tme voce communcaton. It s hypotheszed that the combnaton of varable btrate compresson and same length encrypton wll nduce vulnerablty to traffc analyss. The varaton of packet szes can leak nformaton about the conversaton startng wth language dentfcaton, dentfyng certan phrases, and reconstructng phonemes. The proposed soluton to ths problem s to rely on constant btrate compresson or to pad the sent frames to a multple of 16, 32, or 64 bytes. In [16], a symmetrc key s developed whch conssts of reshufflng and secret arrangement of secret sgnal data bts n cover sgnal data bts. In ths paper the authors have performed the encrypton process on secret speech sgnal data bts-level to acheve greater strength of encrypton whch s hdden nsde the cover mage. The encrypton algorthm appled wth embeddng method s the robust secure method for data hdng. 4. The Proposed Algorthm Ths part presents the audo sgnal encrypton based on the permutaton of the resdue number system as the proposed algorthm. Ths algorthm can be executed on both stereo and mono sgnals for encrypton. Three dfferent encrypton keys are defned n the current approach. The frst encrypton ncludes the values of modular components; the second encrypton s the formaton order of modules; and the thrd formaton key s the permutaton pattern. The relatons of permutaton pattern are multples of the number of modules. It s obvous that the more the number of modules are, the more the permutaton relatons become; therefore, the coeffcent of sgnal securty soars. However, f the number of modules and the permutaton relatons of the sgnal ncrease, the overhead resultng from ther applcaton to the sgnal becomes heaver; thus, the output sgnal volume ncreases. In the proposed algorthm, the permutaton relatons scramble the elements composng sgnal. They are scrambled n an organzed way due to the predefned relatons. Havng the encrypton keys, the encrypted sgnal can be decrypted. 71

6 Audo Sgnal Encrypton Based on F. Ghanbar Adv, M. Mehrna A set of modules lke {2 n -1, 2 n, 2 n +1} was used n the current approach. Gven the fact that there are 3 members n ths set, the permutaton relatons wll have 9 rules. They are closely assocated wth the number of members n the set of modules. These relatons can be the number of modular sets by the power of 2 so that the sgnal encrypton process could be based on them. The relatons pertanng to a modular set lke {2 n -1, 2 n, 2 n +1} are demonstrated n the followng table. Table 1: Permutaton Relatons of the Proposed Algorthm Output nput Y[n] Y[n+1] Y[n+2] Y[n+3] Y[n+4] Y[n+5] Y[n+6] Y[n+7] Y[n+8] The defnton of the permutaton relatons should also have specal relatonshps so that the encrypton process would be done free of errors. These relatonshps are as follows: 1- The number of relatons should be the number of modules by the power of The remander of each element of the nput sgnal on all the modules should exst n permutaton relatons. In Table 1, the values of x[n] refer to the nput sgnal samples, and the values of Y[n] represent the output sgnal samples resultng from applyng the permutaton relatons and the resdue number system. The relatons are appled n a way that the frst sample and the last three samples were not changed, and the relatons were appled to the rest of the samples. That the frst sample and the last three samples were not changed resulted n the correct executon of relaton. The encrypted sgnal s defned accordng to [ [ ] [ ] [ ] [ ] [ ] [ EncryptedSgnal = x 1 Y 1 Y n x end 3 x end 2 x end 1] ] (6) Gven the regulated scramblng of audo sgnal samples, sgnal decrypton based on the permutaton relatons, shown n Table 1, and the modular set wll be a quck, effcent process. An nstance of sgnal form change based on the proposed encrypton sgnal can be seen n Fgures (1) and (2). 72

7 Journal of Advances n Computer Research (Vol. 7, No. 3, August 2016) Fgure 1. The Man Sgnal (not Encrypted Sgnal) Fgure 2. The Encrypted Sgnal: Ths Sgnal s Encrypted by Algorthm and Convert to Slent Sgnal wth same frequency Accordng to Fgures (1) and (2), there s no sgn of the man sgnal samples n the encrypted sgnal whch s lke a sgnal wth a constant frequency. Snce dfferent encrypton parameters are used n the algorthm, t can adapt to every condton related to the transfer medum. Increasng or decreasng the number of modules and changng the number of permutaton relatons or the type of relatons, t can also ncrease the securty coeffcent or decrease the overhead resultng from the encrypton process. However, t s necessary to establsh a balance between the number of modules and permutaton relatons and the type of relatons n order to propose an optmal algorthm. The decrypton of audo sgnals s another problem of audo sgnal encrypton algorthms. Such an algorthm s consdered effcent when the decrypted sgnal has the same qualty as the man sgnal, or the qualty s close to the man one. The qualty of decrypted sgnal s very mportant n the proposed algorthm; therefore, efforts were made to decrypt the sgnal wth the best qualty. Consderng the fact that the resdue number system was used to decrypt the audo sgnal, t s necessary that the remanng value of modules be changed to the ntal values after permutaton through the Chnese Remander Theorem. The sgnal decrypton process has two general steps: - Fndng the multplcatve nverson for the module used n the encrypton algorthm based on equ. 3, equ. 4 and equ. 5 - Usng Relaton equ. 3 and the permutaton relatons defned n Table 1 to decrypt the encrypted sgnal 73

8 Audo Sgnal Encrypton Based on F. Ghanbar Adv, M. Mehrna 5. Performance Evaluaton There are no absolute measures to evaluate nose, and all noses are measured n a relatve way. The sgnal-to-nose rato (SNR) s used to measure the sgnal nose. The values of SNR that are smaller than 12 ndcate that the sgnal has much nose; however, the values above 30 mean that the sgnal has very lttle nose. SNR = 10log 2 ( n) N = 10 2 $ s s( n ) n( n) Matlab was used to smulate the sgnal encrypton and decrypton algorthm. The pseudo-code pertanng to the encrypton algorthm can be seen n Fgure 3: [Y,Fs]= [Read Audo Sgnal] Y=Y(:,1)+Y(:,2); Y=Y'; Resultn=1; for n=2:numel(y)-3 result(resultn)=mod(y(n+1),5) + 5; result(resultn+1)=mod(y(n-1),3) + 3; result(resultn+2)=mod(y(n),4) + 4; result(resultn+3)=mod(y(n+1),3) + 3; result(resultn+4)=mod(y(n-1),5) + 5; result(resultn+5)=mod(y(n),3) + 3; result(resultn+6)=mod(y(n-1),4) + 4; result(resultn+7)=mod(y(n+1),4) + 4; result(resultn+8)=mod(y(n),5) + 5; Resultn = Resultn + (3^2); n=n+3; end result = [Y(1) result Y(end-3) Y(end-2) Y(end-1)]; Fgure 3: Algorthm Source Code n Matlab In Table 2, SNR was measured for 7 dfferent audo sgnals that were converted nto an encrypted sgnal through the encrypton algorthm. Accordng to ths table, the values of SNR are very low, a fact that ndcates the hgh power of the proposed encrypton algorthm. Ths table can also lead to the nference that the proposed algorthm can deal wth every number of samples correctly and encrypts the sgnal at the hghest level of securty. (7) Dfferent Sgnals Sgnal 1 Sgnal 2 Sgnal 3 Sgnal 4 Sgnal 5 Sgnal 6 Sgnal 7 Table 2: SNR after Encryptng Audo Sgnals The Number of Samples SNR e e e e e e-08 74

9 Journal of Advances n Computer Research (Vol. 7, No. 3, August 2016) Concluson In the current approach, an algorthm based on permutaton relatons and the resdue number system was proposed for audo sgnal encrypton. The results ndcated that the proposed algorthm had a hgh level of securty and audo qualty n decrypton. Usng the resdue number system, a three-member modular set, and 9 permutaton relatons n ths algorthm, the ntal sgnal was encrypted. Therefore, the cryptanalyss was dffcult, and the securty of audo sgnal ncreased. For decrypton, t was necessary to take the permutaton relatons nto account frst n order to reconstruct the sgnal. Based on the Chnese Remander Theorem, the values were then converted nto the ntal samples of sgnal. Therefore, a good qualty was presented after decryptng the encrypted audo sgnal. References [1] Harjnder Kaur, A four level speech sgnal encrypton YCSC, vol3, pp [2] Sttar B. Sadkhan, Ndaa A. Abbas, 2014, A proposed voce encrypton System Base on off-lne ca Algorthm, Journal Babylon unversty/pure and Appled Scences/No (7) Vol (22) [3] Tanbo Dong, Yngke Le & Jnhshu Yang, 2012, An Algorthm for Underdetermned Mxng Extmaton, Elsever, Neurocomputng(pp-1-9) [4] Mosa, 2009, Random encrypton of speech sgnal, Computer Engneerng & Systems. ICCES Internatonal Conference [5] Mosa, 2009, Chaostc encrypton of speech sgnals n transform domans, Computer Engneerng & Systems, ICCES Internatonal Conference [6] Anwar, S.; Alam, S.B.; Rahman, K.M.S, 2010, Informaton Theory and Informaton Securty (ICITIS), IEEE Internatonal Conference on P [7] Basher, E.B.M., 2013, Speech scramblng based on chaotc maps and one-tme pad, Computng, Electrcal and Electroncs Engneerng (ICCEEE), Internatonal Conference on Pp [8] Patl, S.B., 2015, A robust encrypton method for speech data hdng n dgtal mages for optmzed securty, Pervasve Computng (ICPC), Internatonal Conference P.1-5 [9] Kohata, M, 2014, Secure Speech Encrypton System Usng Segments for Speech Synthess, 2014, Intellgent Informaton Hdng and Multmeda Sgnal Processng (IIH-MSP), Tenth Internatonal Conference, P [10] Radha, 2011, Comparatve analyss of compresson technques for Taml speech datasets, Recent Trends n Informaton Technology (ICRTIT), Internatonal Conference on P [11] J. C. Bajard, 2015, RNS Arthmetc Approach n Lattce-Based Cryptography: Acceleratng the "Roundng-off" Core Procedure, Computer Arthmetc (ARITH) IEEE 22nd [12] G. K. Armah, 2015, Applcaton of resdue number system (RNS) to mage processng usng orthogonal transformaton, 2015, Communcaton Software and Networks (ICCSN), IEEE Internatonal Conference [13] Yaoyao Chen, 2015, End-to-end speech encrypton algorthm based on speech scramblng n frequency doman, Thrd Internatonal Conference on Cyberspace Technology on Pp. 1 5 [14] A. Kanhe, 2015, Robust Audo steganography based on Advanced Encrypton standards n temporal doman, 2015, Advances n Computng, Communcatons and Informatcs (ICACCI), Internatonal Conference on P.p

10 Audo Sgnal Encrypton Based on F. Ghanbar Adv, M. Mehrna [15] Atte, 2015, Analyss of the tradeoff between compresson rato and securty level n real-tme voce communcaton, Appled Research n Computer Scence and Engneerng (ICAR), 2015 Internatonal Conference on, Pp. 1-6 [16] S. A. Kulkarn, 2015, A robust encrypton method for speech data hdng n dgtal mages for optmzed securty, Pervasve Computng (ICPC), 2015 Internatonal Conference on Pp