1 Introduction

Owing to modern and latest technological enhancement, it is essential to secure the audio and speech signal from eavesdropper over a public wired and wireless network. Desecration may happen in these types of insecure networks for imperative information stealing, dishonest cash, funny activity challenges, terrorism, etc., In order to safeguard the speech data transmitted over a public network, various speech encryption algorithms have been developed such as International Data Encryption Algorithm (IDEA), Advanced Encryption Standard (AES), Data Encryption Standard (DES), Blind Source Separation (BSS) method, Chaotic system based algorithms, etc., [2, 6, 11, 19, 20, 24, 36]. DNA code design based on block quantum algorithm has used in reducing the error in hybridization of large chaotic random number generation. The segmented speech samples have combined and encoded as a DNA’s sequences [14, 27]. Quasi groups method is a dominant method wherein this supplements a larger size of samples by changing the amplitudes and this method is appropriate for high volume speech [38]. The sampled speech signal divided into four segments of equal time slot have permuted and substituted using four different chaotic maps [31, 35]. A cyclic elliptic curve (CEC) method and chaotic system generates PRN sequences based on a piecewise nonlinear chaotic map and are mixed with the key sequences derived from CEC points. Asymmetric key cryptography El-Gamal algorithm has implemented for speech signals encryption and decryption [15, 45]. Field Programmable Gate Array (FPGA) hardware realization for all integer, fractional, and mixed-order new Wang chaotic system has proposed to achieve a reasonable reduction in the hardware resources of chaotic systems implementation [8, 18]. Encryption system for speech signals based on circular shifts in row and column using three secret keys has developed [45]. A self-generated true random number based audio encryption system has implemented [41]. To improve the security level for speech signal, X-Tea and RC-6 algorithms have implemented in pipelined manner in terms of memory/area chip, operations used, encryption time etc. [9]. Compared with conventional encryption algorithms, chaotic system based algorithms are highly analogous in various aspects such as robustness of encryption process and decryption quality. Chaotic system based algorithms are in trend for past few decades to achieve secure communication [3, 12, 49]. The main objective of chaos systems in speech algorithms is to generate pseudo-random sequence to scramble the input speech signal [17]. Chaotic systems are more sensitive to its initial conditions and controlling variables. A tiny change in initial conditions and controlling variables makes huge variation in random number generated. These changing parameters have considered as keys in speech encryption algorithms [3, 22, 37].

Chaotic systems are highly preferable for its great permutation and diffusion properties. Some development of research work on chaos based encryption for image, audio, speech and video have discussed in [4, 32, 33, 39, 47, 48]. Substitution-boxes (S-box) based encryption method has constructed using hybrid chaotic maps and Linear Fractional Transform (LFT) to encrypt the necessary parts of sensitive information in Lifting-Wavelet Transform (LWT) frequency domain [13, 23, 25, 29]. Hyper-chaos has proposed for image encryption since hyper-chaos has more than one positive Lyapunov exponents and more complex dynamical behavior than normal chaos. [5, 10, 21, 28, 50] Color image encryption scheme, based on skew tent map and hyper-chaotic system of 6th-order CNN has developed [16, 40] to confuse and diffuse the pixels. Most of the hyper-chaotic systems have effectively designed for image encryption technique. Some researchers have worked on speech encryption using hyper-chaotic system and implemented on hardware [7, 21]. A Speech scrambling method based on a hybrid chaotic system has developed using Logistic and Henon maps [30].

In this paper, a synchronized hybrid and hyper-chaotic system for speech encryption has introduced. A hybrid chaotic system developed by using Zaslavskey map and Zigzag map have used to generate pseudo random numbers for permuting the speech samples. A five dimensional hyper-chaotic system that is a chaotic oscillator used to produce pseudo random numbers for substituting the scrambled speech samples. A masking sequence has developed from the hyper-chaotic oscillator for masking the encrypted sequence. Existing speech encryption algorithms fail to provide enough security level, lodge extra bandwidth usage and keep noise in the decrypted signal [1, 43, 44, 46]. To overcome these flaws, a novel hybrid-hyper chaotic system and binary masking technique has developed to encrypt the input speech. This speech encryption algorithm provides better security to endure security attacks. The exclusivity of this proposed speech encryption algorithm are:

  1. (i).

    to scramble the speech signal with high confusion and diffusion

  2. (ii).

    to protection to cryptanalytic attack

  3. (iii).

    to reconstruct the input speech with high quality, and

  4. (iv).

    to minimize the computational complication

The practical applications of the proposed speech encryption algorithm are: Hot links, Military voice communication, Voice over IP, Personal mobile communication and telecommunication services, etc.

The organization of this paper is as follows: Section-2 discusses the proposed hybrid and hyper-chaotic mathematical models, section-3 elaborates the proposed architecture of hybrid-hyper speech encryption technique and masking method of encrypted signal, section-4 presents results and performance analysis, and finally the conclusion is has presented in section-5.

2 Proposed method

The proposed work comprises three levels of encryption schemes. In first level, a hybrid chaotic system has developed by using Zaslavskey and Zigzag maps. The output of the Zigzag map has given to the Zaslavskey as input [46]. The resultant sequence has permuted with the DCT output of the speech signal [20]. In second level, a five dimensional hyper-chaotic system has used to diffuse the output of first level encryption. In this process, a reference speech has used as reference signal to substitution. To generate the reference speech signal a HMM speech synthesizer is used. The hyper-chaotic system has constructed by five variables as five keys. The reference speech has permuted using the hyper-chaotic system and the output has used for substitution with the output of first level encryption. From the 5D hyper-chaotic system, a masking sequence has developed for masking the entire encrypted sequence. As third level of encryption, the masking sequence has used to mask the encrypted sequence.

2.1 Hybrid chaotic system

The hybrid chaotic system has designed using two or more different independent chaotic systems, which may be discrete and/or continuous chaotic system. The hybrid chaotic system increases the complexity in the random number generation and hence becomes difficult to analyze the sequence of random numbers. This is a better approach for a cryptosystem, which has used for secure speech in public network. The advantages of this chaotic system are strong secure key generation, key distribution, authentication, etc. In this paper, we have used two discrete chaotic maps, Zaslavsky map and Zigzag map to achieve better chaotic behavior.

2.1.1 Zaslavsky map

The Zaslavsky Map is a discrete-time nonlinear dynamical system exhibits deterministic dynamic behavior. This map is suitable to build hybrid chaotic map due to its flexible and dynamic characteristics. Equation (1) shows the Zaslavsky map:

$${\displaystyle \begin{array}{c}y=\operatorname{mod}\left({y}_{n-1}+v\left(1+\mu {z}_{n-1}\right)+\varepsilon v\mu \cos \left(2\pi {y}_{n-1}\right),1\right)\\ {}z={e}^{-r}\left({z}_{n-1}+\cos \left(2\pi {y}_{n-1}\right)\right)\end{array}}$$
(1)

where \(\mu =\frac{1-{e}^{-r}}{r}\); r = 2–5; v = 100–150 and ε = 0–1.

Initial values for random key stream generation are y0 = 0.587201561347 and z0 = −0.28432144902. Equation (1) has iterated for Ns times where Ns is bit stream limit. The pre-processing of generating the key stream is given by eq. (2):

$${x}_n=\operatorname{mod}\left( abs\left(\operatorname{int}\left({z}_n\times {10}^9\right)\right),1\right)$$
(2)

The key stream is normalized by (3):

$$\overline{x}=\frac{x-{x}_{\mathrm{min}}}{x_{\mathrm{max}}-{x}_{\mathrm{min}}}$$
(3)

where xmin – minimum value of generated key and xmax - minimum value of generated key.

2.1.2 Zigzag map

The output values of the zigzag map will alternate between positive and negative values. The absolute value of output of zigzag map is equal to the tent map due to symmetry. The alternation between positive and negative values provides unintelligibility and is helpful in being there execution of non-idealities. The generalized zigzag map has given in the eq. (4):

$${x}_{n+1}=\left\{\begin{array}{c}-m\left({x}_n+\frac{2}{\left|m\right|}\right),-1<{x}_n\le -\frac{1}{\left|m\right|}\\ {}{mx}_{n,}-\frac{1}{\left|m\right|}<{x}_n\le \frac{1}{\left|m\right|}\\ {}-m\left({x}_n-\frac{2}{\left|m\right|}\right),-\frac{1}{\left|m\right|}<{x}_n\le 1\end{array}\right.$$
(4)

where m is a real number and m ϵ (−3, 3). In this eq. (5), m = −2 represents the zigzag map. Generating of bits from this map is simple. |xn| < 1/2 represents 0 and |xn| > 1/2 represents 1. Without loss of generality, the initial state of the system has been assumed as a slightly positive number, i.e., ×0 = 0+. Critical point falls at m = 3 that provides larger value. This case is at the average of instability, since the system has driven out of the map. In 2 < m < 3, the system is chaotic and stable. Chaos can be observed for 1 < m < 2. Since the initial state of the system has assumed as positive, the output value has confined to positive values. If the system gets out of the positive region and goes into the negative region, it will trap in the negative region. For |m| < 1, no chaotic behavior has observed. The output of the system will ultimately settle in zero. −2 < m < −1 is also a chaotic region for the system. The output in this region alternates between positive and negative values. There is no asymptotic density distribution when the output values are in the positive and negative regions for odd and even time steps respectively. In other words, limfn(x) does not exist as n → ∞. The bifurcation diagram of the Zaslavzky and Zigzag maps have given in Figure 1. These diagrams show the random behavior of the Zaslavzky and Zigzag chaotic maps.

Fig. 1
figure 1

Bifurcation diagram chaotic maps. (a) Zaslavzky Map, (b) Zigzag Map

Full size image

2.2 Hyper chaotic system

Chaos is a highly used system to generate pseudo random numbers for various deterministic non-linear applications. Chaotic systems provide excessive sensitivity to initial conditions and have random like behaviors. Hyper-chaos is a unique chaotic system which has multi direction instability when chaos makes instability in only one direction. Lyapunov exponent plays an effective role in the chaotic property of nonlinear dynamic system. If any one Lyapunov exponent in a nonlinear dynamic system is non-zero and positive then the system becomes a chaotic behavior. Normally the dimension of the nonlinear dynamic system equals to the number of Lyapunov exponent and at least. For example, a 3-dimensional nonlinear dynamic system will have three or more Lyapunov exponent. If three or more Lyapunov exponents are positive and non-zero then the nonlinear dynamic system will produce higher instability. Higher dimensional hyper-chaotic system has higher key space, more complex in randomness and more sensitive. This behavior of higher dimension hyper chaos leads to research in cryptographic applications.

In this paper, 5-dimensional hyper-chaotic system has used for pseudo random number generation to substitute the speech signal. The 5-D chaotic system is presented in eq. (5):

$$\left.\begin{array}{c}{x}_1^{\prime }=-{ax}_1+{x}_2{x}_3\\ {}{x}_2^{\prime }=-{bx}_2+{fx}_5\\ {}{x}_3^{\prime }=-{cx}_3+{gx}_4+{x}_1{x}_2\\ {}{x}_4^{\prime }={dx}_4-{hx}_1\\ {}{x}_5^{\prime }={ex}_5-{x}_2{x}_1^2\end{array}\right\}$$
(5)

Where x1 to x5 are state variables and a to h are constant parameters. Nonlinear terms are exist in \({x}_1^{\prime }\), \({x}_3^{\prime }\), and \({x}_5^{\prime }\) of this dynamical system (1) i.e. x1x2, x2x3, and \({x}_2{x}_1^2\). We have chosen the initial value for all the five state variables from [0, 0,0, 0, 0] to [1] and constant parameters as a = 4.3 to 20, b = 10.5 to 50, c = 10 to 30, d = 0 to 2, e = 15 to 25, f = 20 to 40 and h = 1 to 10.

Different dynamical periodic orbit has obtained by varying the constant parameters in a range in addition to hyper-chaotic. The variety of Lyapunov exponents of the system have obtained from the rage of different constant parameters. For simple and efficient mechanism, the constant parameter ‘b’ has chosen randomly. The constant parameter ‘b’ rage has taken between [10.5, 50], there are two positive integers have appeared so the system is hyper-chaotic. The bifurcation diagram of the hyper-chaotic and hybrid chaotic maps are illustrated in Figure 2. These diagrams show the random behavior of the hyper-chaotic and hybrid chaotic maps.

Fig. 2
figure 2

Bifurcation diagram of hyper chaotic Map and Hybrid Chaotic Map. (a) Hyper Chaotic Map, (b) Hybrid Chaotic Map

Full size image

3 Proposed encryption and decryption algorithm

The Figure 3 shows the architecture of hybrid-hyper speech encryption technique based on permutation and substitution processes and masking of encrypted sequence developed from hyper-chaotic system.

Fig. 3
figure 3

Block diagram of proposed system

Full size image

3.1 Encryption algorithm

  • Step-1: The input speech signal has compressed by DCT to make unintelligibility. Input sequence (SD) is permuted with the random numbers (Zk) generated by hybrid chaotic system where the hybrid chaotic system is designed by Zaslavskey and zigzag maps. x0 of the Zigzag map in equation-4 is given as input to the initial state of the Zaslavskey map in equation-1. Input speech samples have permuted with the random number sequence generated by Zaslavsky map Zk:

$${S}_{DP}={S}_D\oplus {Z}_k$$
(6)

where SDP and SD are encrypted and input speech signal at first level

  • Step-2: In second level of encryption, a reference speech signal which is generated by HMM speech synthesizer has been used as reference for substitution process. The reference speech have permuted by using the hyper-chaotic system and used for substitution process. Thus the first level encrypted sequence SDP is substituted by the binary sequence generated by the 5-D hyper-chaotic system SRP represented in equation-5 and obtained SE. Here the output of hyper-chaotic system have permuted using a reference speech SR that is generated by a speech synthesizer.

  • Step-3: A masking sequence has developed for masking the encrypted signal by applying three combinations of different logical operation:

$$\mathrm{Ms}=\left(\left(\mathrm{x}1|\mathrm{x}3\right)\&\left(\mathrm{x}2\&\mathrm{x}4\right)\right)\mid \left(\mathrm{x}5!\right)$$
(7)

  • Step-4: In third level of encryption, the masking sequence Ms. is XORed with second level encryption sequence SE. Finally the encrypted signal TS is transmitted over the public channel.

3.2 Decryption algorithm

The reverse process of encryption method is the decryption method. Here the original speech signal has recovered without any loss of information.

4 Experimental results and analysis

The performance of proposed crypto algorithm has been measured and analyzed by key space analysis, key sensitivity analysis, correlation coefficient analysis, SNR, Peak Signal to Noise Ratio (PSNR), segmental SNR (SNRseg), frequency-weighed segmental SNR (fwSNRseg) analysis, computational complexity measure, subjective evaluation of speech quality, PESQ analysis, NSCR and UACI analysis. The proposed system has implemented and tested in Matlab. 10 sample speech signals are taken randomly from TIMIT database and are sampled at 8 kHz with length of 3 Sec to 8 Sec and 8000 samples per frame. R2013a in Intel Core-i3 CPU @ 3.30GHZ speed, 4GB RAM, 64bit operating system. The acceptable range of different SNR metrics as per NIST statistical test suite [33] are listed in the Table 1.

Table 1 Acceptable range of SNR as per NIST statistical test suite
Full size table

4.1 Histogram analysis

Histogram is a pictorial representation (Figure 4) consisting of spikes whose area is proportional to the frequency of a variable and whose width is equal to the class interval. Histogram(X) creates a histogram plot of X. The histogram function uses automatic binning algorithm that returns the bin with uniform width, chosen to cover the range of elements in X. In a more general mathematical sense, a histogram is a function mi that counts the number of observations that fall into each of the disjoint categories known as bins.

$$n=\sum_{l=1}^k{m}_i$$
(8)

where, n is the total number of observations and k be the total number of bins and mi the histogram.

Fig. 4
figure 4

Plot and histogram of original, encrypted and decrypted speech. a. Original Speech, b. Encrypted Speech, c. Decrypted Speech, d. Histogram of Original Speech, e. Histogram of Encrypted Speech, f. Histogram of Decrypted Speech

Full size image

Variance of histogram is analysed as follows:

$$Var(x)=\frac{1}{n^2}\sum_{i=1}^n\sum_{j=1}^n\frac{1}{2}{\left({x}_i-{x}_j\right)}^2$$
(9)

where x is the vector of histogram values and xi and xj are the number of samples equal to i and j respectively. This quantity analysis of each key employ variances of histograms of encrypted speech signals to evaluate the uniformity. If the variance is low means, uniformity of encrypted speech is high and vice versa. Two variances of encrypted speech have calculated by two different secret keys. By varying the secret keys if we have obtained lower uniformity of encrypted speech indicates strong encryption.

In the Table 2, the variances obtained by two different secret keys x1 and k and the variances obtained by modifying one parameter of k, x1, x2, x3, x4, x5, and x5. By changing any of the secret key, 10 to 20% fluctuations have obtained. If all the secret keys have changed, then 80–100% fluctuations is possible. It shows the proposed algorithm has good key sensitivity.

Table 2 Histogram variance analysis
Full size table

4.2 Spectrogram analysis

The spectrograms of the original, encrypted and decrypted speech signals have presented in Figure 5. The spectrograms of the original and encrypted speech signals show that these two signals are completely different with higher encryption quality. The spectrograms of the original and decrypted speech signals are identical to each other. Thus higher decryption quality has proved and this hold good for all the spectrogram analyses.

Fig. 5
figure 5

Spectrogram of original, encrypted and decrypted speech. (a). Spectrogram of Original Speech, (b). Spectrogram of Encrypted Speech, (c). Spectrogram of Decrypted Speech

Full size image

4.3 Correlation analysis

The auto-correlation function can identify the chaotic system that produces a strong encryption. A useful measure to assess the encryption quality of any cryptosystem is the correlation coefficient between similar segments in the clear signal and the cipher signal. It is calculated as:

$${\mathrm{r}}_{\mathrm{xk}}=\frac{\mathrm{C}\left(\mathrm{x},\mathrm{k}\right)}{\sqrt{\mathrm{V}\left(\mathrm{x}\right)}\sqrt{\mathrm{V}\left(\mathrm{k}\right)}}$$
(10)

where C(x, k) is the covariance between the original signal x and the encrypted signal k. V(x) and V(k) are the variances of the signals x and k. The variance V(x) is computed as:

$$V(x)=\frac{1}{N_s}\sum_{i=1}^{N_s}{\left(x(i)-E(x)\right)}^2$$
(11)
$$E(x)=\frac{1}{N_s}\sum_{i=1}^{N_s}\Big(x(i)$$
(12)
$$C\left(x,k\right)=\frac{1}{N_s}\sum_{i=1}^{N_s}\left(x(i)-E(x)\right)\left(k(i)-E(k)\right)$$
(13)

where Ns is the number of speech samples. The low value of correlation coefficient rxk shows a good encryption quality. The correlation coefficients for the three different encrypted speech samples with the chaotic maps illustrated in Fig. 1 and the encrypted speech signals with proposed method using five different chaotic random number generator are tabulated in Table 2.

It has observed that from these results, the proposed algorithm produces encrypted speech with low correlation between similar segments in the original speech and the encrypted speech, which means that the encryption method gives good encryption results. In this proposed method it is obtained that the correlation co-efficient as 0.998 shows the original speech signal has been permuted almost 100% in decryption process so it is tough to the eavesdroppers to hack the speech signal in channel during transmission.

4.4 SNR analysis

Signal to Noise Ratio (SNR) test is a good estimator for measuring the speech signal intelligibility. The popular time domain metric is the SNR, which has defined as the average of the SNR values of short segments of the output signal as calculated below eq. (14):

$$SNR=10{\mathit{\log}}_{10}\frac{\sum_{i=1}^{N_s}{x}^2(i)}{\sum_{i=1}^{N_s}{\left(x(i)-y(i)\right)}^2}$$
(14)

where y(i) is decrypted speech signal. If the SNR is closer to zero, the higher is the quality of the decrypted signal.

4.5 Peak signal to noise ratio (PSNR) analysis

Peak Signal to Noise Ratio (PSNR) is ratio between the maximum possible power of original speech signal and the power of encrypted signal [2]. PSNR is a calculation of encryption quality of the original signal [17]. Higher PSNR indicates that the encryption or reconstruction is of higher quality. The PSNR is obtained from eq. (15):

$$PSNR\;=\;10\log\frac{{nx}^2}{\left\|x-k\right\|^2}$$
(15)

The quality of decrypted signal of this algorithm is measured using SNR calculation as illustrated in Table 3. The average value of CC, SNR, PSNR, SNRseg and fwSNRseg for the ten decrypted speeches are 0.99299, 122.49 dB, 49.489 dB, 122.905 dB and 55.186 dB respectively with a difference of 0.0087, 2.72 dB, 3.42 dB, 2.72 dB and 2.34 dB. This algorithm brings about good range of SNR value and good quality of decrypted speech.

Table 3 Performance analysis of decryption process
Full size table

4.6 Robustness test

The robustness test has made for different input signals and the comparison for the number of samples change rate (NSCR) and the unified average changing intensity (UACI) in percentage have examined. The ciphered speech signals have compared by NSCR and UACI [39]. The NSCR is specified by,

$$NSCR\;=\;{\textstyle\sum_i}\frac{d_i}l\;\times\;100\%$$
(16)

and

$$UACI\;=\frac{1}{l}\left[\sum_i\frac{\left|{x}_i-{x_i}^{\prime}\right|}{65535}\right]$$
(17)

where x and x’ are the two ciphered speech signals whose corresponding original signal signals have only one-sample difference; the values of the samples at position i of x and x’ are respectively denoted by xi and xi’; l corresponds to the length of the speech vector. The ideal values for NSCR and UACI are 100% and 33.3%. In Table 3, the minimum, maximum and average values of NSCR and UACI, computed from the encryption of four different modified versions of each speech signal are given.

The results are considerably close to the ideal values and independent on the position of the modified sample as compared with the ideal values for NSCR and UACI for robust test are 100% and 33.3% and from the equation (16) and (17), the performance analysis for robustness test has made. From Table 4, comparison between existing methods and proposed method, depict that the resultant values are significantly nearer.

Table 4 Robustness Test NSCR and UACI analysis
Full size table

4.7 Key sensitivity analysis

In this key sensitivity analysis, we have changed the key values for one bit. It demonstrates a dramatically cause of decryption error at the receiver side. The Table 5 clearly proves that a small change in each key provides more than 99% difference in key space. Thus it provides more complex to larger space [33].

Table 5 Key Sensitivity analysis
Full size table

4.8 Perceptual evaluation of speech quality (PESQ) analysis

The PESQ analysis is carried out and results are furnished in Table 6. As many as 50 speech samples in different varity from TIMIT database and 50 recorded speech samples in different environments (Total: 100 samples) are taken to find out the average correlation coefficient [33].

Table 6 Average and inappropriate correlation coefficient for representative and chosen 100 samples randomly
Full size table

4.9 Subjective evaluation

The quality of encrypted and decrypted speech have exposed to suitable evaluation by the known sources in different age population. They agreed that the decrypted speech is same as the original speech and the encrypted speech is noisy. The subjective evaluation [33] of encrypted and decrypted speech listed in Table 7.

Table 7 Subjective evaluation of encrypted and decrypted speech
Full size table

4.10 Types of classical attacks

Generally the cryptanalyst knows the whole thing about the cryptosystem other than the secret key and intending to attack the cryptosystem. In cryptanalyzing there are four types of classical attacks: (a) Cipher text only, (b) Known plain text, (c) Chosen plain text and (d) Chosen cipher text. Out of which, chosen plain text attack is the most dominant attack. If a crypto system can withstand this attack, then it can withstand all other attacks.

The proposed method is sensitive to the system parameters and initial key parameters of hybrid chaotic system developed by using Zaslavskey map and Zigzag map ×0, y0, z0 and m. If any one of the key values has changed, the entire output would be dramatically changed. Hence, the proposed method can withstand the chosen plain text attack. The experimental results proof that the proposed method provides a higher level of security to the speech communication models [42].

4.11 Overall performance comparison

The performance of the proposed speech encryption algorithm has compared with four existing methods. One common speech sample has taken randomly from TIMIT dataset, which has already used in result analysis study. The performance metrics used for this performance comparison are correlation coefficient, SNR, NSCR and UACI. This comparison presented in Table 8 shows that the proposed method provides better results compared with existing methods.

Table 8 Overall Performance Comparison
Full size table

Advanced Encryption Standard (AES) is a digital cryptosystem in which high degree of security can be attained but it is not often used in existing speech communication systems owing to the bandwidth expansion of the encrypted speech and the degradation of the Signal-to-Noise Ratio (SNR) performance. The original information can be easily detected by eavesdroppers from the residual intelligibility such as signal energy, talk spouts, and the original pitch which are present in the encrypted speech. This can degrade the security of the speech cryptosystem. The chaotic based speech scrambler is the best choice to remove the dashes of residual intelligibility in the encrypted speech and to withstand the frequency domain attacks. Chaotic systems have mixing, stretching, folding and sensitivity to key parameter properties. These confusion and diffusion properties make the chaotic systems a worthy choice for constructing a good cryptosystem. [26, 34].

5 Conclusions

In the proposed system, a complex algorithm for secured speech communication is proposed. The proposed algorithm comprising the two chaotic systems: one is hybrid chaotic system built by Zaslavskey and zigzag maps and another one is five-dimensional hyper-chaotic system in which a masking sequence is used for masking the encrypted signal and increases the complexity. Since from the result analysis, it, demonstrates large key space, good correlation, NSCR and UACI. Performance analysis reflects that the proposed system is highly complex, secure, and negligibly prone to noise and it can be applied in the real time communication. The system can ensure voice security and it is the finest mechanism. This work covers way for lots of openings, wherein this can be implemented in hardware.