Research Article  Open Access
Wei Feng, Jing Zhang, Zhentao Qin, "A Secure and Efficient Image Transmission Scheme Based on Two Chaotic Maps", Complexity, vol. 2021, Article ID 1898998, 19 pages, 2021. https://doi.org/10.1155/2021/1898998
A Secure and Efficient Image Transmission Scheme Based on Two Chaotic Maps
Abstract
The application of multimedia sensors is widespread, and people need to transmit images more securely and efficiently. In this paper, an image transmission scheme based on two chaotic maps is proposed. The proposed scheme consists of two parts, secure image transmission between sensor nodes and sink nodes (SITSS) and secure image transmission between sensor nodes and receivers (SITSR). For resourceconstrained environments, SITSS utilizes TentLogistic Map (TLM) to generate chaotic sequences and adopts TLMDriven permutation and transformation to confuse image pixels. Then the cipher image is obtained through TLMDriven twodimensional compressed sensing. Compared with existing schemes, the secret key design of SITSS is more reasonable and requires fewer hardware resources. When sampling ratio is greater than 0.6, its image reconstruction quality has obvious advantages. For environments with huge security threats, SITSR adopts dynamic permutation and confusion based on discrete logarithms to confuse the image and exploits dynamic diffusion based on discrete logarithms to generate final cipher image. Similarly, compared with some existing schemes, the design of SITSR is more practical, and the statistical characteristics of the cipher image are better. Finally, extensive simulation tests confirm the superiority of the proposed scheme.
1. Introduction
Nowadays, the application of multimedia sensors is increasingly widespread in many fields, such as medicine, transportation, industry, education, and military. In these application scenarios, flexibly deployed sensors need to transmit massive images, such as medical and military images [1, 2]. Since it involves privacy protection, commercial and military security, etc., efficient and secure protection needs to be provided for these images. However, image data has several significant characteristics that are different from text data, such as large volume and strong pixel correlation [3]. And the hardware resources of sensors are limited. Therefore, traditional encryption schemes such as Advanced Encryption Standard (AES) are generally not suitable for heterogeneous application environments [4–7]. In order to continuously improve the efficiency and security of image transmission, researchers have been committed to designing new schemes based on emerging techniques and methods [3–28]. Among these new schemes, the ones based on compressed sensing (CS) and chaotic systems are favored by more and more researchers [11–13, 16–28].
CS [29, 30] is a breakthrough signal acquiring paradigm, which can effectively capture and recover a signal with fewer nonadaptive samples. Once introduced, CS is quickly applied to image related information security applications [4–7, 10, 11, 31–37]. In the past decade, researchers have gradually introduced CS into information security applications in resourceconstrained environments. In [4], a scheme called DiffieHellmanHashCompression was proposed. This scheme uses Semitensor Product (STP) CS to encrypt images of different dimensions and adopts hash algorithm and permutation operations to ensure secure image transmission. Taking into account the high privacy sensitivity and redundancy of medical images, Wang et al. [5] constructed a CS based medical image encryption scheme. This scheme carries out image encryption between sensor nodes by using a measurement matrix as the secret key and can realize image compression, privacy protection, and data aggregation simultaneously. In order to overcome the resource constraints of sensor nodes and ensure the security of data transmission, an image encryption system was exploited [6]. While enhancing the security of transmission process by integrating the quantization and diffusion operations, the system uses a new CS model and parallel reconstruction algorithm to shorten the encryption/decryption time. In [7], a flexible and secure data encryption system based on CS was proposed. The plain image is first sparsely represented through discrete wavelet transform and then permutated by Arnold scrambling. Finally, after CS and logistic chaotic permutation, the cipher image is obtained. Utilizing structurally random matrices, Unde et al. [10] presented an efficient scheme based on CS. In their scheme, artificial noise is injected into quantized CS measurements, thereby enhancing the ability to resist ChosenPlaintext Attacks (CPAs).
Chaotic systems have several characteristics that are very suitable for designing cryptosystems [1, 2]. Consequently, more and more researchers leverage chaotic systems to design various image encryption schemes. In [16], an image encryption scheme using memristive chaotic system was provided. This scheme uses Secure Hash Algorithm (SHA) to generate the secret key and calculate the initial value of the chaotic system. And it also introduces a dynamic Deoxyribonucleic Acid (DNA) encoding method to generate two regular DNA matrices for encoding images. In order to protect medical images, Moafimadani et al. [23] presented an image encryption scheme based on a chaotic system, which uses a fast permutation operation to scramble the plain image and utilizes an adaptive diffusion operation to generate the cipher image. In [24], a chaotic image encryption scheme using a new symmetric key generation system was proposed. This scheme exploits blocklevel permutation and improved zigzag transformation to achieve the confusion effect and adopts pixel shuffling to complete the pixel diffusion operation. With the goal of improving the security and efficiency of image encryption, Zhu et al. [25] proposed an efficient and simple Sbox generation method using a new compound chaotic system and then introduced a new image encryption scheme based on double Sboxes. Based on dynamic DNA encoding and two chaotic systems, Zhou et al. [26] proposed an image encryption scheme with a tworound permutationdiffusion structure. This scheme exploits a twodimensional (2D) rectangular transformation to complete the permutation operation, and before the diffusion operation, the hamming distances of DNA matrices are used to update the initial values of the chaotic systems.
As can be seen from abovementioned works, in terms of designing image encryption schemes, reducing resource consumption and achieving higher security are key motivations. Although these schemes have advantages in some aspects, they all have room for further improvements. For example, the scheme proposed in [4] adopts SHA to resist CPAs. However, the implementation of SHA demands considerable hardware resources and would hinder the applicability of this scheme in resourceconstrained environments. In addition, some encryption schemes adopt onetime pad secret key. When a large number of images need to be encrypted, such design is not practical. Therefore, while further improving the efficiency and security of image encryption, to overcome the shortcomings of these schemes, an image transmission scheme based on two chaotic maps, 2DCS, dynamic perturbation, and discrete logarithms (ITSCDD) is proposed. The proposed scheme consists of two parts, secure image transmission between sensor nodes and sink nodes (SITSS) and secure image transmission between sensor nodes and receivers (SITSR). Compared with some existing schemes, ITSCDD has contributions summarized as follows:(1)SITSS is designed for resourceconstrained environment, whereas SITSR is designed for environments with huge security threats. Therefore, the applicability and practicability of ITSCDD are higher.(2)Dynamic perturbation parameters (DPPs) derived from system times and last encryption times are designed. So, ITSCDD not only guarantees the diversity of equivalent key streams, but also does not rely on external algorithms.(3)The secret key design of SITSS is more practical and requires fewer hardware resources.(4)2DCS based on lightweight chaotic map can reduce resource overhead.(5)Discrete logarithms under finite multiplicative group are introduced to ensure higher security.
The remainder of this paper is organized as follows. 2DCS, discrete logarithms, and two chaotic systems are introduced in Section 2. ITSCDD is described in Section 3. Simulation tests and theoretical analyses are carried out in Section 4. Finally, conclusions are drawn in Section 5.
2. Fundamental Knowledge
In SITSS, 2DCS is introduced to realize image data compression and encryption. Discrete logarithms are used to enhance the security of SITSR. Two chaotic systems called TentLogistic Map (TLM) [38] and 2D LogisticSineCoupling Map (2DLSCM) [13] are adopted to generate the chaotic sequences.
2.1. 2DCS
In terms of computational complexity and storage space, 2DCS has obvious advantages over traditional CS [39, 40]. Assuming that and are random matrices, they both have the size of . Then, one can obtain the 2D measurements of an image . Specifically,where and operate on the rows and columns of , respectively.
When decoding, one can regularize the image signal recovery by using signal prior information in the form of penalty:where is the regularization parameter, is a cost function which is used to handle the illposed problem, and is the datafidelity term. Moreover, researchers have proposed many 2DCS reconstruction algorithms to solve the optimization problem mentioned above. In this paper, 2D projected gradient with embedding decryption (2DPGED) [12] algorithm is adopted.
2.2. Discrete Logarithms
Discrete logarithm calculation is a complex nonlinear calculation. In the encryption process, the use of discrete logarithms can improve its nonlinearity [14]. For the prime 257 and its corresponding finite multiplicative group , one can define the discrete logarithms as follows: if satisfies , then is said to be the discrete logarithm of . Since has 128 generators, we can use them to enhance the diversity of equivalent key streams. To avoid complex discrete logarithm calculation, we calculate the discrete logarithm values under different generators in advance and save them to the 2D matrix with the size of . Consequently, in ITSCDD, discrete logarithm values can be obtained by directly accessing . If one wants to calculate the discrete logarithm value of 107 under the generator 3, namely, calculating , one can access to obtain the discrete logarithm value 31. Table 1 shows the discrete logarithm values of 101 to 107 under the first eight generators.

2.3. TLM and 2DLSCM
To save hardware resources, TLM is adopted in SITSS, which is easy to implement and has good chaotic performance. TLM can be defined aswhere is generated by the th iteration, is the input of the th iteration, is the initial state, and are the control parameters. Figure 1 shows the 2D bifurcation diagram and Lyapunov Exponents (LE) diagrams of TLM.
(a)
(b)
(c)
Compared with TLM, 2DLSCM has better chaotic performance, but its structure is more complex, so it is more suitable for environments with more hardware resources. 2DLSCM can be defined aswhere is the system state generated by the th iteration, is the input of the th iteration, is the initial state, and is the control parameter. The value ranges of all these parameters are . Figure 2 shows the 2D bifurcation diagram and LE diagram of 2DLSCM.
(a)
(b)
3. Proposed Image Transmission Scheme
Different from some existing schemes, ITSCDD consists of two parts, secure image transmission between sensor nodes and sink nodes (SITSS) and secure image transmission between sensor nodes and receivers (SITSR). Figure 3 shows the secure image transmission between sensor nodes and sink nodes.
Compared with the existing schemes, SITSS has two main innovations. One is introducing TLM to save the hardware resources of sensors, and the other is introducing DPPs to enhance the ability to resist CPAs. Figure 4 shows the secure image transmission between sink nodes and receivers.
Considering that sink nodes have more resources, there are huge security threats in the process of transmitting images to receivers through the media cloud. We have adopted some measures to improve the security of image transmission, such as the adoption of 2DLSCM with better chaotic performance and the introduction of discrete logarithms.
3.1. Transmission between Sensor Nodes and Sink Nodes
To save space, in this subsection, we mainly introduce the improvements to 2DCSETC [12].
3.1.1. DPP Generation
According to previous cryptanalysis works, the main reason why some schemes cannot resist CPAs is that equivalent key streams only depend on the secret key [41–47]. Therefore, some researchers use the hash value of the plain image to ensure the diversity of equivalent key streams. However, the implementation of hash algorithm is not suitable for sensor nodes with limited resources. Considering that system times and last encryption times are constantly changing and would be affected by many factors, they are used to generate DPPs. The specific generation process of DPPs is as follows:(i)Step 1: obtain the system time in milliseconds.(ii)Step 2: get the time spent in the last encryption process in milliseconds. If it is the first time to encrypt, set to an initial value .(iii)Step 3: one DPP is obtained by .(iv)Step 4: repeat Step 1 through Step 3 until 32 DPPs are obtained, namely, .
In this way, we can obtain a set of DPPs. Like the hash value, DPPs can ensure that the equivalent key streams used when encrypting different images are different, thereby effectively resisting CPAs. More importantly, no complicated calculations are required to obtain DPPs, and even if the same plain image is encrypted, different equivalent key streams would be generated.
3.1.2. TLMDriven Global Permutation
Obviously, confusion is the requirement that must be considered when designing modern cryptosystems. Confusion means that each bit of the secret key should affect as many cipher image bits as possible [48]. Permutation operations are commonly used to achieve confusion, but permutationonly image encryption schemes have been proven to be insecure [49]. Therefore, SITSS introduces DPPs in the permutation process. This makes the permutation process not only dependent on the secret key, but also dependent on the DPPs that will inevitably change every time the plain image is encrypted. Compared with 2DCSETC using the random permutation matrix to complete the permutation and treat it as secret key, we use TLM and DPPs to complete the permutation. This can not only reduce the resource overhead of sensor nodes, but also improve the ability to resist CPAs. The specific process of TLMDriven Global Permutation (GP) is as follows:(i)Step 1: use the parameters to iterate TLM times. In order to avoid negative effects, discard the first chaotic state values.(ii)Step 2: convert the obtained chaotic sequence of length into the integer sequence where , returns the integer part of an operand.(iii)Step 3: stretch the plain image of size into the 1D sequence .(iv)Step 4: calculate the index of 32 DPPs and the permutation position where . Swap two pixels of according to .
3.1.3. TLMDriven NegativePositive Transformation
A nonlinear operation called NegativePositive Transformation (NPT) is introduced by 2DCSETC to improve security. Similarly, we use TLM and DPPs to complete NPT instead of using a random matrix in the form of secret key. This can further reduce the resource overhead of sensor nodes and improve the ability to resist CPAs.(i)Step 1: use parameter to iterate TLM times. In order to avoid negative effects, discard the first chaotic state values.(ii)Step 2: convert the obtained chaotic sequence of length into the bit sequence where .(iii)Step 3: according to , perform the following NPT operation on . where .(iv)Step 4: reshape into the 2D cipher image.
3.1.4. TLMDriven 2DCS
If the chaotic sequence generated by the chaotic system is assembled into a complete measurement matrix, its performance is usually almost the same as other commonly used random matrices [11]. Moreover, compared with directly using a random matrix and treating it as secret key, the chaotic measurement matrix can significantly save the resource overhead of sensor nodes. In SITSS, TLM is used to generate the measurement matrices required for 2DCS. Suppose the size of the measurement matrices and to be created is ; the specific process of TLMDriven 2DCS is as follows:(i)Step 1: use the parameters to iterate TLM times. In order to avoid negative effects, discard the first chaotic state values.(ii)Step 2: arrange the obtained chaotic sequence into the square matrix of size .(iii)Step 3: take rows from the orthogonal basis of as the measurement matrix .(iv)Step 4: repeat Step 1 through Step 3; create the measurement matrix in a similar manner.(v)Step 5: use and to obtain the 2D measurements of the cipher image .
In addition to the improvements made above, the other steps of SITSS are basically the same as those of 2DCSETC, which are not repeated here. Since we have introduced TLM and DPPs in SITSS, the security of image transmission between sensor nodes and sink nodes has become higher, and the resource requirements for sensors are also lower. Significantly, SITSS still maintains the advantages of 2DCSETC, which is demonstrated and discussed in Section 4.1. To save hardware resources, we directly use as the secret key of SITSS.
3.2. Transmission between Sink Nodes and Receivers
In SITSR, we use 2DLSCM [13] which has better chaotic performance to generate chaotic sequences. Moreover, discrete logarithms and DPPs are introduced to achieve secure image transmission between sink nodes and receivers. It should be noted that through the use of discrete logarithms and our targeted design, DPPs can be directly sent out in plaintext form by sink nodes. When decrypting, receivers can directly use DPPs that arrived in plaintext form. In other words, DPPs are not onetime pad secret keys, nor are they secret parameters. Next, we introduce the specific process of SITSR, as shown in Figure 5.
3.2.1. Secret Key and Chaotic System Parameters
In order to avoid the secret key issues pointed out in some cryptanalysis works and simplify the generation process of chaotic system parameters [14, 41, 42], we set the secret key in this stage as a binary sequence with the length of 270 bits. Namely, . In specific implementation, we directly use nine 32bit unsigned integers , , , , , , , , to generate three sets of parameters , , for 2DLSCM. As shown in equation (11), this means that the bits of correspond to the 30 bits of each unsigned integer, respectively.where
Besides, these three sets of chaotic system parameters , , are used to generate chaotic matrices for the encryption process.
3.2.2. DPP Generation
The generation process of DPPs in SITSR is exactly the same as SITSS. And we mark 32 DPPs used in SITSR as .
3.2.3. Dynamic Permutation and Confusion Based on Discrete Logarithms
As mentioned above, discrete logarithms and DPPs are introduced in dynamic permutation and confusion based on discrete logarithms (DPCD), so as to enhance the security of image transmission. Specifically, compared with some existing permutation operations, DPCD has the following advantages:(1)Use to further perturb the permutation results and adopt different perturbation strategies for the row index and column index. Therefore, the permutation results depend not only on the secret key, but also on .(2)Based on discrete logarithms, and the sorting results of the chaotic matrix are used to nonlinearly transform the pixel value of each plain image pixel, thereby further improving the security of image transmission.
In order to better describe the specific steps of DPCD, an algorithm is provided in Algorithm 1.

3.3. Dynamic Diffusion Based on Discrete Logarithms
To further improve security, dynamic diffusion based on discrete logarithms (DDD) also adopts discrete logarithms and . Specifically, compared with some existing diffusion operations, DDD has the following advantages:(1)Considering that multipixel diffusion is of little significance, singlepixel diffusion is adopted, thereby reducing the amount of computation(2)The nonlinearity of the diffusion process is improved by introducing discrete logarithms; thus the security of image transmission is further improved
In order to better describe the specific steps of DDD, an algorithm is provided in Algorithm 2.

Since a symmetric encryption structure is adopted in SITSR, the decryption process is actually constituted by the corresponding inverse operations of the encryption operations. With the received DPPs and the agreed secret key , receivers can decrypt the plain image from the cipher image. To save space, these inverse operations are not repeated here.
4. Simulation Tests and Analyses
In this section, extensive simulation tests are performed to demonstrate the superiority of ITSCDD. ITSCDD is an image transmission scheme composed of two parts, and the resource conditions and design goals of each part are different. Therefore, SITSS is compared with 2DCSETC for resourceconstrained environments, whereas SITSR is compared with more versatile schemes for general application environments. Without loss of generality, randomly generated secret keys are used to complete the tests. Table 2 lists the hardware and software configurations used in the tests.

4.1. Simulation Tests for SITSS
Since reducing the resource consumption of sensors and improving the security of image transmission is our goal in designing SITSS, the analysis and verification of SITSS are mainly focused on these two aspects. The test images used are eight images used in [12].
4.1.1. Encryption and Decryption
Four plain images Lena, Boats, House, and Parrots are shown in Figure 6. Their corresponding cipher images and decrypted images generated in SITSS are also provided. As can be seen from these images, the cipher images are similar to noise, attackers cannot obtain useful information from them, and there are no significant visual differences between the decrypted images and corresponding plain images.
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)
(j)
(k)
(l)
(m)
(n)
(o)
(p)
4.1.2. Reconstruction Quality
Researchers often use Peak SignaltoNoise Ratio (PSNR) to evaluate image reconstruction quality. Generally, a higher PSNR value indicates a better reconstruction quality. The definition of PSNR is as follows:where is the size of the reconstructed image and original image . PSNR versus sampling ratio for 2DCSETC and SITSS is listed in Table 3. As can be seen from Table 3, SITSS can achieve the same or slightly different PSNR values as 2DCSETC. And when sampling ratio is greater than 0.6, its image reconstruction quality has obvious advantages.
 
Each bold value means that one of the two compared schemes has a higher PSNR value than the other. 
4.1.3. Secret Key
In 2DCSETC, the random permutation matrix and random binary integer matrix are used as the secret key, thereby obtaining a huge key space. However, in resourceconstrained environments, it is not suitable to use such secret key that requires a large amount of storage space. For example, if the size of the plain image is , the secret key used would be at least 2 228 224 bytes in length. In addition, in the encryption and decryption process, the generation and storage of two measurement matrices also bring significant resource requirements. However, in SITSS, we only need to store six floatingpoint numbers used as the secret key. Meanwhile, SITSS also enjoys a large enough key space, which is about . Apparently, such a large key space is sufficient to resist brute force attacks.
4.1.4. ChosenPlaintext Attack
As we know, CPAs are the reasons why some encryption schemes are cracked. It is generally believed that a secure encryption scheme should be able to resist CPAs. Derived from system times and last encryption times, DPPs always change dynamically and will be affected by many factors, thereby ensuring the diversity and unpredictability of equivalent key streams. Without relying on external algorithms such as hash algorithms, the diversity of equivalent key streams brings excellent resistance to CPAs.
4.2. Simulation Tests for SITSR
The simulation tests presented in this subsection are to demonstrate the superiority of SITSR in terms of security and encryption efficiency. The test images used for SITSR are from The USCSIPI Image Database (https://sipi.usc.edu/database/).
4.2.1. Encryption and Decryption of Different Types of Images
For different types of images, we uniformly treat them as 8bit grayscale images in ITSCDD. Specifically, for images with a pixel depth of less than 8 bits, we directly process them as 8bit grayscale images, whereas for images with pixel depth greater than 8 bits, we encrypt them in groups of 8 bits. For example, if we need to encrypt an image with a pixel depth of 16 bits, we can encrypt the lower 8 bits and higher 8 bits of each plain image pixel separately. Figure 7 shows the encryption and decryption effects of SITSR for different types of images. One can see that SITSR has excellent encryption effects for different types of images. The generated cipher images are very similar to noise images, and attackers cannot directly obtain any valuable information from these cipher images.
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)
(j)
(k)
(l)
4.2.2. Key Space and Key Sensitivity
Since the key space would affect the ability to resist brute force attacks, a secure encryption scheme should have a sufficiently large key space [50]. We carefully design the secret key, which not only solves the issues about equivalent secret keys, but also expands the key space to . Therefore, SITSR is excellent in terms of the ability to resist brute force attacks.
It is well known that a secure encryption scheme should be able to achieve superior confusion effect [50, 51]. That is, one smallest change in the secret key should make the cipher image change significantly. To evaluate the performance of SITSR in this regard, we randomly generated the secret key
Using , we encrypted elaine.512.tiff to obtain the corresponding cipher image . Next, we changed one bit of to get two secret keys , with single smallest changes. These two changed secret keys also were used to encrypt elaine.512.tiff, thus obtaining the corresponding cipher images , . Finally, the difference images between , , and were calculated. As can be seen from Figure 8, the difference images between , , and look similar to an ordinary cipher image. This means that the key sensitivity of SITSR in the encryption process is extraordinary.
(a)
(b)
(c)
(d)
(e)
(f)
Similarly, in order to verify the key sensitivity of SITSR in the decryption process, and were adopted to decrypt . The test results are shown in Figure 9. Once again, judging from the difference image between the wrongly decrypted images, the key sensitivity of SITSR in the decryption process is excellent.
(a)
(b)
(c)
(d)
(e)
For measuring the degree of changes between images, NPCR (Number of Pixel Change Ratio) and UACI (Unified Average Change in Intensity) are commonly used indicators [13]. Among them,refers to the ratio of the pixels that change, whereasrefers to the average intensity of the pixel value changes. In equations (14) and (15), is the size of the images, , , and is the difference matrix between the image before the change and the image after the change. If , then . Otherwise, . In order to quantitatively analyze the key sensitivity of SITSR, we calculated the NPCR and UACI values between the cipher images before and after the secret key changes. As one can see from Table 4, all the test results are very close to the ideal values, which means that SITSR does have extremely high key sensitivity.
 
The bold values are the ideal values. 
4.2.3. Pixel Value Distribution
Unlike plain images, cipher images must have extremely uniform pixel value distributions; otherwise attackers will have the opportunity to conduct statistics based attacks [50, 51]. In order to visually demonstrate the pixel value distribution characteristics of the plain images and the cipher images generated by SITSR, the pixel value distribution histograms of these images are plotted. As can be seen from Figure 10, the pixel distributions of the plain images are relatively concentrated, whereas the pixels of the cipher images are extremely uniformly distributed throughout the entire value range.
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)
(j)
(k)
(l)
In addition, the histogram variance analysis and chisquare test analysis are also performed on the cipher images to qualitatively analyze the ability of SITSR to resist statistical attacks. In general, if the histogram variance of a cipher image is smaller, then the uniformity of the cipher image is higher. The histogram variance of an 8bit grayscale image can be defined as follows:where ; and are the numbers of the pixels whose grayscale values are equal to and . Table 5 lists the histogram variance test results of some images. These images include one random image, the 8bit grayscale image Lena, and its cipher images generated by different schemes.
From Table 5, one can see that the histogram variance of the plain image is very large, which means that its pixel value distribution is extremely uneven, whereas among the cipher images generated by the four image encryption schemes, the cipher image of SITSR has the smallest histogram variance, which indicates that this cipher image has the most uniform pixel value distribution and is closest to the random image.
Another way to quantitatively analyze the uniformity of a cipher image is the chisquare test. The chisquare value of a cipher image can be calculated as follows:where is the height and width of the cipher image, is the number of pixels whose pixel value is , is the number of all possible pixel values (for 8bit grayscale images, ), and . Next, the critical value of the chisquare test at the significant level can be determined, which is 293.2478. If a cipher image has a chisquare value less than 293.2478, then this image can be considered to have passed the chisquare test; that is, its pixel value distribution is statistically uniform. Consequently, the smaller the chisquare value of a cipher image is, the better its uniformity is. As can be seen from Table 6, for some commonly used test images, the cipher images generated by SITSR have all passed the chisquare test. And in most cases, SITSR performs better than another scheme.

4.2.4. Plain Image Sensitivity
When the plain image changes, the corresponding change degree of the cipher image is plain image sensitivity. Generally speaking, to effectively resist differential attacks, an image encryption scheme must have excellent plain image sensitivity. To intuitively show the plain image sensitivity of SITSR, we first encrypted some commonly used test images. Next, after changing one pixel bit for each plain image, the plain images with the smallest changes were encrypted. At last, we calculated the difference images between the cipher images obtained before and after the smallest changes. The relevant test results are shown in Figure 11. As one can see from Figure 11, each plain image has undergone only one smallest change, but almost all cipher pixels have changed. In addition to that, these significant differences are independent of where the plain images change and are very close to random images.
UPCR and UACI are also utilized to qualitatively analyze the plain image sensitivity of SITSR. The UPCR and UACI values between the cipher images obtained before and after the smallest changes of 15 common test plain images are shown in Table 7. Judging from the test results, SITSR has excellent plain image sensitivity. The test results of SITSR are closer to the ideal values and and perform better in terms of stability.
 
The bold values here emphasize that SITSR has better performance than the other scheme. 
4.2.5. Information Entropy
Information entropy is an indicator that can be used to measure the randomness or disorder of an information source. If the information entropy of the information source is higher, it can be considered that the information source has higher randomness or disorder [18–20]. When it comes to an 8bit grayscale image, the information entropy of the grayscale image can be calculated as follows:where is the total number of symbols ; is the occurrence probability of symbol . For the 8bit grayscale cipher images, the ideal value of the information entropy is 8 [18–20]. From Table 8, one can see that the information entropy of each cipher image generated by SITSR is very close to the ideal value 8. As shown in Table 9, compared with several image encryption schemes, the information entropy of the Lena cipher image generated by SITSR is closest to the ideal value 8. Therefore, SITSR performs best in terms of the information entropy.

In order to measure the randomness of cipher images more comprehensively, a measure named Local Shannon Entropy (LSE) is proposed [52]. This measure is increasingly adopted to verify the randomness of cipher images [13]. Mathematically, LSE can be defined as follows:where are randomly selected nonoverlapping image blocks, is the number of pixels in each block, and is the information entropy of . According to the test method suggested in [52], we carried out the LSE test on the cipher images generated by SITSR, and the relevant test results are shown in Table 10. Compared with two recent image encryption schemes, SITSR has the best performance in terms of standard deviation and pass rate.
 
The bold values indicate that compared with the other two schemes, SITSR has the best performance in terms of standard deviation and pass rate. 
4.2.6. Pixel Correlation
The extremely high correlation between adjacent pixels is one of the salient features of plain images and also one of the important reasons why traditional encryption schemes are not suitable for image encryption [50]. Therefore, a secure image encryption scheme should eliminate the correlation between adjacent pixels as much as possible. CC (correlation coefficient) is an effective indicator to measure the correlation between adjacent pixels, and its definition is as follows:where and are the grayscale values of two adjacent pixels; and are the expectation and variance of the grayscale value . In order to verify the performance of SITSR in terms of the pixel correlation, for the horizontal, vertical, and diagonal directions, we have randomly selected 20,000 pairs of adjacent pixels from each plain image and its corresponding cipher image to calculate the CCs. The relevant test results are shown in Table 11.

From Table 11, one can see that there are very high correlations between adjacent pixels of the plain images; that is, the absolute values of CCs are extremely high, whereas in the cipher images generated by SITSR, there is almost no correlation between adjacent pixels; that is, the absolute values of CCs are extremely low .
In addition, in order to more intuitively show the correlation changes between adjacent pixels caused by the encryption of SITSR, the correlation distribution charts of the plain image elaine.512.tiff and its corresponding cipher image are drawn. As can be seen from Figure 12, after the encryption processing of SITSR, there is almost no correlation between adjacent pixels in each direction.
(a)
(b)
(c)
(d)
(e)
(f)
4.2.7. ChosenPlaintext Attack
In fact, almost all simulation tests related to security analysis can only ensure the security of image encryption schemes under ciphertextonly attacks (COAs) [51, 53]. This is exactly why some image encryption schemes have been broken. Among the four types of attacks, which are COAs, KnownPlaintext Attacks (KPAs), CPAs, and ChosenCiphertext Attacks (CCAs), CCAs are the most threatening ones, but the attack conditions required by them are practically meaningless [3, 50]. If attackers can choose cipher images arbitrarily, then they do not need to crack at all, because for any cipher image, they can directly recover its plain image. Therefore, it is generally believed that CPAs are the most threatening ones among common practical attacks. Actually, in the cryptanalysis works about image encryption, the vast majority of them adopt CPAs [3, 51]. Next, from the perspective of attackers, the ability of SITSR to resist CPAs is analyzed.
Apparently, attackers will encounter several problems when they try to break SITSR with CPAs. Firstly, we assume that they could obtain the equivalent key streams of the encryption process from chosen plain images and corresponding cipher images. However, because SITSR introduce DPPs in serval encryption steps, the equivalent key streams they obtained cannot be used to recover other ordinary plain images, which are encrypted under different DPPs. Secondly, SITSR also performs nonlinear transformations on the plain image pixels during the permutation process, so the common attack method that ignores the permutation process by the chosen plain images composed of singlevalue pixels cannot work. Thirdly, SITSR adopts a nonlinear diffusion structure; that is, it adopts the discrete logarithms based on two different generators, which makes the encryption process cannot be simplified by chosen plain images. To sum up, SITSR can effectively resist CPAs.
4.2.8. Encryption Efficiency
Improving encryption efficiency is one of the most important motivations to design new image encryption schemes. SITSR introduce DPPs and discrete logarithms, but in fact, discrete logarithms can be calculated in advance, and the calculation process of DPP is very simple, so the impact on encryption efficiency is very small. In addition, SITSR uses singlepixel diffusion and only performs three iterations in the encryption process. These also help to reduce the total number of primitive operations that need to be executed. Table 12 shows the average time required by SITSR and some other recent image encryption schemes to encrypt the 8bit grayscale image Lena . As can be seen from Table 12, although the time complexity of each image encryption scheme is , the scheme proposed in [13] requires the least number of primitive operations, so it has the highest encryption efficiency, whereas for SITSR, it adds a certain number of primitive operations to ensure the security, but it still maintains the significant advantage of high encryption efficiency. That is, in terms of encryption efficiency, SITSR is significantly better than the remaining four image encryption schemes.
5. Conclusions
In order to improve the efficiency and security of image transmission, an image transmission scheme based on two chaotic maps is proposed in this paper. The proposed scheme divides the image transmission from sensor nodes to receivers into two stages and carries out a targeted design, which can better adapt to heterogeneous application environments. For image transmission between sensor nodes and sink nodes, the proposed scheme reduces the requirements for hardware resources and improves the image reconstruction quality by introducing a lightweight chaotic map. Besides, the design of dynamic perturbation improves the security of image transmission at this stage, whereas for image transmission between sink nodes and receivers, the proposed scheme improves the security and efficiency of image transmission by introducing another chaotic map with better chaotic performance and discrete logarithms. In order to verify and demonstrate the excellent performance of the proposed scheme, extensive simulation tests and theoretical analyses are carried out. These tests and analyses show that, compared with some recent schemes, the proposed scheme has higher feasibility, security, and practicability. In the future, we will extend the proposed scheme to video transmission.
Data Availability
The figure data and table data used to support the findings of this study are included within the article.
Conflicts of Interest
The authors declare no conflicts of interest.
Authors’ Contributions
Wei Feng conceptualized the study, was responsible for methodology and software, performed formal analysis and investigation, prepared the original draft, and was involved in funding acquisition; Jing Zhang was involved in conceptualization, funding acquisition, and supervision, validated the data, and reviewed and edited the manuscript; Zhentao Qin was responsible for methodology and software, validated the data, and reviewed and edited the manuscript. All authors have read and agreed to the published version of the manuscript.
Acknowledgments
This research was supported by the Science and Technology Development Center Project of Chinese Ministry of Education (no. 2018A0105), the Natural Science Key Project of Education Bureau of Sichuan Province (no. 18ZA0288), the Guiding Science and Technology Plan Project of Panzhihua City (nos. 2019ZDG18 and 2020ZDS40), and the Doctoral Research Startup Foundation of Panzhihua University (no. 2020DOC019).
References
 Y. Zhang, Y. Xiang, L. Y. Zhang, Y. Rong, and S. Guo, “Secure wireless communications based on compressive sensing: a survey,” IEEE Communications Surveys and Tutorials, vol. 21, no. 2, pp. 1093–1111, 2019. View at: Publisher Site  Google Scholar
 Y. Zhang, Q. He, Y. Xiang et al., “Lowcost and confidentialitypreserving data acquisition for internet of multimedia things,” IEEE Internet of Things Journal, vol. 5, no. 5, pp. 3442–3451, 2018. View at: Publisher Site  Google Scholar
 C. Li, Y. Zhang, and E. Y. Xie, “When an attacker meets a cipherimage in 2018: a year in review,” Journal of Information Security and Applications, vol. 48, Article ID 102361, 2019. View at: Publisher Site  Google Scholar
 Z. Niu, M. Zheng, Y. Zhang, and T. Wang, “A new asymmetrical encryption algorithm based on semitensor compressed sensing in WBANs,” IEEE Internet of Things Journal, vol. 7, no. 1, pp. 734–750, 2020. View at: Publisher Site  Google Scholar
 L. Wang, L. Li, J. Li, J. Li, B. B. Gupta, and X. Liu, “Compressive sensing of medical images with confidentially homomorphic aggregations,” IEEE Internet of Things Journal, vol. 6, no. 2, pp. 1402–1409, 2019. View at: Publisher Site  Google Scholar
 L. Li, G. Wen, Z. Wang, and Y. Yang, “Efficient and secure image communication system based on compressed sensing for iot monitoring applications,” IEEE Transactions on Multimedia, vol. 22, no. 1, pp. 82–95, 2020. View at: Publisher Site  Google Scholar
 L. Li, L. Liu, H. Peng, Y. Yang, and S. Cheng, “Flexible and secure data transmission system based on semitensor compressive sensing in wireless body area networks,” IEEE Internet of Things Journal, vol. 6, no. 2, pp. 3212–3227, 2019. View at: Publisher Site  Google Scholar
 R. Zhao, Y. Zhang, X. Xiao, X. Ye, and R. Lan, “Tpe2: threepixel exact thumbnailpreserving image encryption,” Signal Processing, vol. 183, Article ID 108019, 2021. View at: Publisher Site  Google Scholar
 Y. Zhang, R. Zhao, X. Xiao, R. Lan, Z. Liu, and X. Zhang, “HFTPE: highfidelity thumbnailpreserving encryption,” IEEE Transactions on Circuits and Systems for Video Technology, p. 1, 2021. View at: Publisher Site  Google Scholar
 A. S. Unde and P. P. Deepthi, “Design and analysis of compressive sensingbased lightweight encryption scheme for multimedia iot,” IEEE Transactions on Circuits and Systems II: Express Briefs, vol. 67, no. 1, pp. 167–171, 2020. View at: Publisher Site  Google Scholar
 Y. Zhang, Q. He, G. Chen, X. Zhang, and Y. Xiang, “A lowoverhead, confidentialityassured, and authenticated data acquisition framework for iot,” IEEE Transactions on Industrial Informatics, vol. 16, no. 12, pp. 7566–7578, 2020. View at: Publisher Site  Google Scholar
 B. Zhang, D. Xiao, and Y. Xiang, “Robust coding of encrypted images via 2D compressed sensing,” IEEE Transactions on Multimedia, vol. 23, no. 99, pp. 2656–2671, 2021. View at: Publisher Site  Google Scholar
 Z. Hua, F. Jin, B. Xu, and H. Huang, “2D logisticsinecoupling map for image encryption,” Signal Processing, vol. 149, pp. 148–161, 2018. View at: Publisher Site  Google Scholar
 W. Feng, Y. He, H. Li, and C. Li, “A plainimagerelated chaotic image encryption algorithm based on DNA sequence operation and discrete logarithm,” IEEE Access, vol. 7, pp. 181589–181609, 2019. View at: Publisher Site  Google Scholar
 H. Li, T. Li, W. Feng et al., “A novel image encryption scheme based on nonadjacent parallelable permutation and dynamic dnalevel twoway diffusion,” Journal of Information Security and Applications, vol. 61, Article ID 102844, 2021. View at: Publisher Site  Google Scholar
 X. Chai, Z. Gan, K. Yang, Y. Chen, and X. Liu, “An image encryption algorithm based on the memristive hyperchaotic system, cellular automata and DNA sequence operations,” Signal Processing: Image Communication, vol. 52, pp. 6–19, 2017. View at: Publisher Site  Google Scholar
 X. Wang and D. Xu, “A novel image encryption scheme based on Brownian motion and PWLCM chaotic system,” Nonlinear Dynamics, vol. 75, no. 12, pp. 345–353, 2014. View at: Publisher Site  Google Scholar
 Q. Yin and C. Wang, “A new chaotic image encryption scheme using breadthfirst search and dynamic diffusion,” International Journal of Bifurcation and Chaos, vol. 28, no. 4, Article ID 1850047, 2018. View at: Publisher Site  Google Scholar
 X. Wu, D. Wang, J. Kurths, and H. Kan, “A novel lossless color image encryption scheme using 2D DWT and 6D hyperchaotic system,” Information Sciences, vol. 349350, pp. 349350, 2016. View at: Publisher Site  Google Scholar
 R. Zahmoul, R. Ejbali, and M. Zaied, “Image encryption based on new beta chaotic maps,” Optics and Lasers in Engineering, vol. 96, pp. 39–49, 2017. View at: Publisher Site  Google Scholar
 H. Zhu, X. Zhang, H. Yu, C. Zhao, and Z. Zhu, “An image encryption algorithm based on compound homogeneous hyperchaotic system,” Nonlinear Dynamics, vol. 89, no. 1, pp. 61–79, 2017. View at: Publisher Site  Google Scholar
 A.V. Diaconu, “Circular interintra pixels bitlevel permutation and chaosbased image encryption,” Information Sciences, vol. 355356, pp. 314–327, 2016. View at: Publisher Site  Google Scholar
 S. S. Moafimadani, Y. Chen, and C. Tang, “A new algorithm for medical color images encryption using chaotic systems,” Entropy, vol. 21, no. 6, Article ID 577, 2019. View at: Publisher Site  Google Scholar
 P. Ramasamy, V. Ranganathan, S. Kadry, R. Damaševičius, and T. Blažauskas, “An image encryption scheme based on block scrambling, modified zigzag transformation and key generation using enhanced logistictent map,” Entropy, vol. 21, no. 7, Article ID 656, 2019. View at: Publisher Site  Google Scholar
 S. Zhu, G. Wang, and C. Zhu, “A secure and fast image encryption scheme based on double chaotic sboxes,” Entropy, vol. 21, no. 8, Article ID 790, 2019. View at: Publisher Site  Google Scholar
 S. Zhou, P. He, and N. Kasabov, “A dynamic dna color image encryption method based on sha512,” Entropy, vol. 22, no. 10, Article ID 1091, 2020. View at: Publisher Site  Google Scholar
 L. Liu, D. Jiang, X. Wang, X. Rong, and R. Zhang, “2D logisticadjustedChebyshev map for visual color image encryption,” Journal of Information Security and Applications, vol. 60, Article ID 102854, 2021. View at: Publisher Site  Google Scholar
 D. Jiang, L. Liu, L. Zhu, X. Wang, X. Rong, and H. Chai, “Adaptive embedding: a novel meaningful image encryption scheme based on parallel compressive sensing and slant transform,” Signal Processing, vol. 188, Article ID 108220, 2021. View at: Publisher Site  Google Scholar
 E. J. Candes, J. Romberg, and T. Tao, “Robust uncertainty principles: exact signal reconstruction from highly incomplete frequency information,” IEEE Transactions on Information Theory, vol. 52, no. 2, pp. 489–509, 2006. View at: Publisher Site  Google Scholar
 D. L. Donoho, “Compressed sensing,” IEEE Transactions on Information Theory, vol. 52, no. 4, pp. 1289–1306, 2006. View at: Publisher Site  Google Scholar
 Y. Zhang, P. Wang, H. Huang, Y. Zhu, D. Xiao, and Y. Xiang, “Privacyassured FOGCS: chaotic compressive sensing for secure industrial big image data processing in fog computing,” IEEE Transactions on Industrial Informatics, vol. 17, no. 5, pp. 3401–3411, 2021. View at: Publisher Site  Google Scholar
 Y. Zhang, P. Wang, L. Fang, X. He, H. Han, and B. Chen, “Secure transmission of compressed sampling data using edge clouds,” IEEE Transactions on Industrial Informatics, vol. 16, no. 10, pp. 6641–6651, 2020. View at: Publisher Site  Google Scholar
 L. Y. Zhang, K.W. Wong, Y. Zhang, and J. Zhou, “Bilevel protected compressive sampling,” IEEE Transactions on Multimedia, vol. 18, no. 9, pp. 1720–1732, 2016. View at: Publisher Site  Google Scholar
 J. Wang, L. Y. Zhang, J. Chen, G. Hua, Y. Zhang, and Y. Xiang, “Compressed sensing based selective encryption with data hiding capability,” IEEE Transactions on Industrial Informatics, vol. 15, no. 12, pp. 6560–6571, 2019. View at: Publisher Site  Google Scholar
 H. Peng, Y. Mi, L. Li, H. E. Stanley, and Y. Yang, “Ptensor product in compressed sensing,” IEEE Internet of Things Journal, vol. 6, no. 2, pp. 3492–3511, 2019. View at: Publisher Site  Google Scholar
 H. Gan, S. Xiao, T. Zhang, and F. Liu, “Bipolar measurement matrix using chaotic sequence,” Communications in Nonlinear Science and Numerical Simulation, vol. 72, pp. 139–151, 2019. View at: Publisher Site  Google Scholar
 L. Li, Y. Fang, L. Liu, H. Peng, J. Kurths, and Y. Yang, “Overview of compressed sensing: sensing model, reconstruction algorithm, and its applications,” Applied Sciences, vol. 10, no. 17, Article ID 5909, 2020. View at: Publisher Site  Google Scholar
 Y. Yicong Zhou, Z. Zhongyun Hua, C. ChiMan Pun, and C. L. P. Chen, “Cascade chaotic system with applications,” IEEE Transactions on Cybernetics, vol. 45, no. 9, pp. 2001–2012, 2015. View at: Publisher Site  Google Scholar
 A. Eftekhari, M. BabaieZadeh, and H. Abrishami Moghaddam, “Twodimensional random projection,” Signal Processing, vol. 91, no. 7, pp. 1589–1603, 2011. View at: Publisher Site  Google Scholar
 G. Chen, D. Li, and J. Zhang, “Iterative gradient projection algorithm for twodimensional compressive sensing sparse image reconstruction,” Signal Processing, vol. 104, pp. 15–26, 2014. View at: Publisher Site  Google Scholar
 C. Li, D. Lin, B. Feng, J. Lü, and F. Hao, “Cryptanalysis of a chaotic image encryption algorithm based on information entropy,” IEEE Access, vol. 6, pp. 75834–75842, 2018. View at: Publisher Site  Google Scholar
 W. Feng, Y. He, H. Li, and C. Li, “Cryptanalysis and improvement of the image encryption scheme based on 2D logisticadjustedsine map,” IEEE Access, vol. 7, pp. 12584–12597, 2019. View at: Publisher Site  Google Scholar
 W. Feng and Y.G. He, “Cryptanalysis and improvement of the hyperchaotic image encryption scheme based on DNA encoding and scrambling,” IEEE Photonics Journal, vol. 10, no. 6, Article ID 7909215, pp. 1–15, 2018. View at: Publisher Site  Google Scholar
 L. Y. Zhang, Y. Liu, F. Pareschi et al., “On the security of a class of diffusion mechanisms for image encryption,” IEEE transactions on cybernetics, vol. 48, no. 4, pp. 1163–1175, 2018. View at: Publisher Site  Google Scholar
 M. Li, D. Lu, W. Wen, H. Ren, and Y. Zhang, “Cryptanalyzing a color image encryption scheme based on hybrid hyperchaotic system and cellular automata,” IEEE Access, vol. 6, pp. 47102–47111, 2018. View at: Publisher Site  Google Scholar
 C. Li, D. Lin, and J. Lü, “Cryptanalyzing an imagescrambling encryption algorithm of pixel bits,” IEEE MultiMedia, vol. 24, no. 3, pp. 64–71, 2017. View at: Publisher Site  Google Scholar
 C. Li, D. Lin, J. Lü, and F. Hao, “Cryptanalyzing an image encryption algorithm based on autoblocking and electrocardiography,” IEEE MultiMedia, vol. 25, no. 4, pp. 46–56, 2018. View at: Publisher Site  Google Scholar
 M. Li, H. Fan, Y. Xiang, Y. Li, and Y. Zhang, “Cryptanalysis and improvement of a chaotic image encryption by firstorder timedelay system,” IEEE MultiMedia, vol. 25, no. 3, pp. 92–101, 2018. View at: Publisher Site  Google Scholar
 A. Jolfaei, X.W. Wu, and V. Muthukkumarasamy, “On the security of permutationonly image encryption schemes,” IEEE Transactions on Information Forensics and Security, vol. 11, no. 2, pp. 235–246, 2016. View at: Publisher Site  Google Scholar
 G. Alvarez and S. Li, “Some basic cryptographic requirements for chaosbased cryptosystems,” International Journal of Bifurcation and Chaos, vol. 16, no. 8, pp. 2129–2151, 2006. View at: Publisher Site  Google Scholar
 F. Özkaynak, “Brief review on application of nonlinear dynamics in image encryption,” Nonlinear Dynamics, vol. 92, no. 2, pp. 305–313, 2018. View at: Publisher Site  Google Scholar
 Y. Wu, Y. Zhou, G. Saveriades, S. Agaian, J. P. Noonan, and P. Natarajan, “Local Shannon entropy measure with statistical tests for image randomness,” Information Sciences, vol. 222, pp. 323–342, 2013. View at: Publisher Site  Google Scholar
 M. Preishuber, T. Hütter, S. Katzenbeisser, and A. Uhl, “Depreciating motivation and empirical security analysis of chaosbased image and video encryption,” IEEE Transactions on Information Forensics and Security, vol. 13, no. 9, pp. 2137–2150, 2018. View at: Publisher Site  Google Scholar
Copyright
Copyright © 2021 Wei Feng et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.