1 Introduction

Alongside the development in communication technology, information exchange in public networks has increased. It is essential to secure private content and this issue has been considered has been the subject of a significant body of researcher over several decades. Cryptography is the process of encrypting sensitive information into scrambled messages [19]. However, the form of encrypted messages makes them conspicuous and attracts attacker into attempting to decrypt the message. Steganography is defined as the art of hiding confidential messages within media, in the form of images, text, video, audio, and protocols [1, 9, 11, 17]. This method provides protection in the form of confidentiality, information availability, and integrity.

The main difference between cryptography and steganalysis lies in the cryptography, where the confidential message is encrypted to prevent third partiesfromunderstanding it. By doing that, the existence of the secret message is recognized, but cannot be undertood without using decryption methods. In steganography, the existence of a secret message is concealed in the media container. The combination of steganography and cryptography can increase the data security [8, 9, 17]. According to the mechanism usedfor data embedding, steganography can be divided into spatial domain and frequency domain. According to Alsarayreh [5], the spatial domain provides more hiding capabilities and offers lower computational time.

A Steganography system consist of an embedding process and an extracting process. A stego image is produced in the embedding process while a cover image and secret message will be required as input. To hide the secret information, the pixel of the cover image will be secretely chosen with a stego key. After that, the secret message can be extracted as shown as in the Fig. 1.

Fig. 1
figure 1

General Steganography System

Full size image

According to Kanan and Nazeri [11], there are 3 main features in steganography that make it suitable for information hiding, which are:

  1. 1.

    Payload Capacity: Determined by the number of secret bits embedded in each cover image pixel. A higher number indicates more data can be embedded into a cover image.

  2. 2.

    Imperceptibility: usually calculates the stego image’s quality by using a peak signal-to-noise ratio (PSNR). The number will be small if the image quality is poor and high if the image quality is good.

  3. 3.

    Robustness: Prevents the secret data from being stolen or manipulated.

The three main features of steganography is shown in Fig. 2.

Fig. 2
figure 2

Trade-off triangle of Steganography features

Full size image

This paper is organized as follows: In Section 2, we provide brief information of the various techniques of steganography and examine their strength and weaknesses. In Section 3, we give the details of the proposed method. Section 4 shows the experimental results and analysis, and the conclusion is given in section 5.

2 Related work

2.1 Least significant bit (LSB)

The LSB method is the most popular technique used in spatial domain steganography. Muhammad, Ahmad, Rehman, Jan, and Sajjad [14] implemented two level encryption algorithms alongside a multi-level encryption to increase the security of the LSB method. Although this method increased the security of the secret message, the payload capacity and image quality are not increased; it is also needed longer computational time.

According to Sabeti, Samavi, and Shirani [18], calculated pixel complexity and setting the threshold value before using the LSB method to embed the secret message. This can be improved the produced stego-image. This method also secured againts analytical steganalysis attack. The drawback of this method is when the data embedding is high, the stego-image is weak againts visual attack. Similar results have also beenfound by Muhammad, Sajjad, Mehmood, Rho, and Baik [13].

2.2 Multi-bit embedding

Multi bit embedding is the improvement of the LSB method which in this method will use more than one bit for embedding instead of using only the last bit of image’s pixel value. Chen, Chang, and Le [6] improved the payload capacity by combining Canny edge detection and Fuzzy edge detection to increase the accuracy for finding edge areas in the images. This method increased the payload capacity, but the disadvantage represented in the image quality, whichis reduced when embedding includes large data.

Furthermore, in recent years, multi-bit embedding has also been used to increase the payload capacity. Several uses of steganography based on multi-bit embedding have been reported [3, 4, 7, 10, 12, 19, 20]. Thismethod, which focused on increasing the payload capacity, generally has a problem with the quality of the produced image.

2.3 Block-data hiding

The block-data hiding is a steganography method where, before embedding, the image is divided into blocks. Nguyen, Arch-int and Arch-int [15, 16] proposed a method to increase the payload capacity, image quality and security by using multi-bit plane block data hiding, adaptive region selection, and AES encryption. The stego image produced through this method is very good and secured against Human Visual Attack. The drawback is that the computational workload is very high.

According to Nguyen, Arch-int, and Arch-int [16], Cellular automata techniques and XOR encoding can be used to improve the security of the block-data hiding methods. This system has good security, but the computational workload is high.

2.4 Bit-matching

The bit-matching method works by finding the matching bit between the secret message bit value and image’s pixel bit value to generate a key to point to the location of the pixel bit and the sequence of the secret message. According to Alamsyah, Muslim, and Prasetyo [2], to combine bit-matching and the DES method can increase the security of the bit-matching technique. This method has high image quality and security, but the drawback when it is vulnerable in to any direct image processing.

Alsarayreh, Alia, and Maria [5] worked on a the bit-matching method which focused on matching values between the ASCII value of the secret message and the pixel value from blue channel. This method provides an excellent good stego image, but security is lacking and weak when subjected to the image processing.

2.5 Current best solution

The current best solution is proposed by Alsarayreh [15]. This solution uses bit-matching methods to generate a key. In this solution the matching values between secret images and pixel’s colour values location, a key was created and sent to the receiver to reconstruct the secret images from the key. In order to get the matching value, the secret message was converted into their ASCII value to find the matches with the pixel colour values of the blue channel.

The reason for choosing this approach as the best solution was because it needs low computational workload and simple to implement. The image quality was not reduced from the method so it’s secure against any type of attacks.

This method doesn’t consider securing the key generated from the exact matching. Any attacker who gets the key can reconstruct the secret message from the key and the image. The limitation also exists when the author only consider to use the blue channel to find the matches since it will reduce the possibilities to broaden the matching values, this method is also not secure against image processing.

3 Proposed method

In this section, the proposed method takes the advantages of the bit-matching algorithm which is presented in section 2.4. The proposed solution incorporates the advantages from the model with the exact matching algorithm to produce undistorted stego images with high embedding capacity. Furthermore, AES security features will also be adopted to improve the security.

The fundamental idea of the proposed system is to use all color channels from images and locate all the Exact Mathces (EM) between the ASCII values of the secret message and the color values of image pixels. A key dependent data will be created and encrypted as a part of the separating procedure. This method overcomes the LSB limitation of hiding capacity and also overcomes the high computational complexity of frequency domain steganography. The process of finding the Exact Matching and the key dependent data will be shown in Section 3.1, and the key creation will be explained in Section 3.2. The extraction of the secret information will be demonstrated in Section 3.3.

3.1 Exact matching (EM) between images and secret message

This section will explain a method of getting the match between cover image and secret message. This method starts with converting the secret message into decimal ASCII value, after which it starts reading the cover image color starting from Blue channel, Green channel and finally to Red channel as decimal value. The system will try to discover the matches between image’s pixel color and the secret message. This method will not regenerate the location of any duplicate character from the secret message. It will only use one pixel location and use it to regenerate the duplicate character. After finished all matching between secret messages and images, all locations (x,y) of the matching pixel will be saved and used to generate the key for sending to the receiver. Table 1 shows a sample of a secret message “test sample” that has been used to demonstrate how the system works. In the secret message, the characters (t) and (e) are both repeated twice, but the system will only use the first matched location value and use it for the construction of the key.

Table 1 Matched values from a sample secret message
Full size table

The algorithm below shows how the proposed method works to find the exact matches between the cover image and secret message:

Algorithm 1: Extraction exact matches pixel locations

Input: Cover image and secret message.

Output: Array of matches pixel locations and character position in secret message.

Begin:

Step 1: Convert the secret message into ASCII value and read the cover image color value.

Step 2: Store the pixel location that matches the cover image’s color value and ASCII value from the secret message starting from the blue color channel, and then green color channel, and finally red color channel.

End

3.2 Key-dependent data creation process

The key is created from the location of the pixel in the cover image and the location of the character in the secret message. However, the key must have three values for each character. The first value from creating the key is the x-axis (row) and y-axis (column) value from the pixel. The second value is the location of the character in the secret message. The third value is the special character that will be used to reconstruct the secret message from the key. The (,) will be used for divide the row and column of the pixel location. The ($) character will be used to separate the pixel location and the character location in the secret message. The (^) character is used to divide each character from secret message from others, and the (*) character will be used to divide the color channels used. The system will only use one pixel location for any duplicate character from the secret message to maintain computational simplicity. Table 2 shows an example for the key creation for each character from the secret message “test sample”.

Table 2 The Key Structure for Each Character
Full size table

After having found all matches and created the key, the last step is to combine all the key values from each character and encrypt it with an Advanced Encryption System with 128-bit key size to form the key that will be sent to the receiver. Figure 3 shows the key created from the secret message “test sample” before encryption. Notice that the character p cannot be found in the blue channel but it can be found in the green channel instead. So, during the key creation process, the key value of character p is after the special character (*) to show that the pixel value is from the next color channel, which is the green color channel. Figure 4 shows the key after AES encryption has been used.

Fig. 3
figure 3

Key generated from test sample before encryption

Full size image
Fig. 4
figure 4

Key generated from test sample after encryption

Full size image

3.3 Extraction and secret message reconstruction process

The algorithm below shows the proposed method for reconstruct the secret message from the image and generated key received:

Algorithm 2: Extraction for pixel locations and reconstruction for the secret message

Input: Image and generated key.

Output: Secret message.

Step 1: Decrypt the generated key with AES method.

Step 2: Decode the generated key to get location of the pixel and character position in secret message.

Step 3: Read the color value from pixel location and color channel.

Step 4: Convert the color value into ASCII and reconstruct the secret message.

End

The proposed system block diagram is shown in Fig. 5. Both the sender and receiver will have the same system. From the sender, the system will require the cover image and the secret message as input for the user. The system will convert the secret message into ASCII value and find the matches with the cover image. After the matching process is completed, the AES method will be used to encrypt the key generated before sending itto the receiver along with the image to maintain the integrity of the message. The receiver will receive the key and the image, after which the system will use the AES method to decrypt the key and reconstruct the secret message from pixel coordinates and character location.

Fig. 5
figure 5

Proposed system block diagram

Full size image

4 Experiment results and analysis

The implementation of the proposed model has been carried out by using NetBeans 8.2 with 6 sample images. The experiment has been carried out on a computer that is running with Windows 10 Operating System, Intel i7 2.5 GHz processor and 4 GB RAM. Fig. 6 shows the dataset that will be used for the experiment. The dataset used for the experiment is fromAlsarayreh [5]. The performance of the proposed system has been measured based on percentage of matching value and levels of security between the current method and the proposed method. The images used for the experiment are airplane, Lena, Elaine, mandrill, pepper, and splash.

Fig. 6
figure 6

dataset used for the experiment (Alsarayreh et al., 2017)

Full size image

The image quality and imperceptibility are strongly related. If there are any changes in the image quality, then the imperceptibility will be reduced. The image quality will be calculated by using PSNR as defined in Eq. 1.

4.1 Equation 1 PSNR calculation

$$ PSNR=10x{\log}_{10}\frac{255^2}{MSE}(dB) $$
(1)

Where MSE (Mean Square Error) will be calculated as defined in Eq. 2. Where W is image width and H is the height, with Iijand \( {I}_{ij}^{\prime } \) are the pixel values of the cover image and stego image, respectively.

4.2 Equation 2 MSE calculation

$$ MSE=\frac{1}{WxH}{\sum}_{i=1}^W{\sum}_{j=1}^H{\left({I}_{ij}-{I}_{ij}^{\prime}\right)}^2 $$
(2)

Table 3 shows the result of possible matches found from each dataset. It is shown that the proposed method can find more matches than the current method. The increase in the matches is occurring when the entire color channels are used for finding the matches instead of only using one color channel. It also shows that the proposed method has more possible matches value than the current method.

Table 3 Results of number of possible matches found
Full size table

From the Table 4, the results of the experiment are shown, based on search time in millisecond (ms). It shows that the proposed system higher speed in finding the matches compared to the current system. The proposed system does not require high computational power and can find the matches quickly. Figure 7 shows that there is a significant gap between the proposed method and the current system in terms of search time. The reason that the proposed system having faster speed is that in the current method, the process scans the pixels for the entire image for every character in the secret image and saves the pixel location in the array. Then, it will randomly choose one pixel location for key construction. By doing this, it consumes a lot of time if the secret message is long. Therefore, the proposed method will only use the first matching pixel location for key generation and in the matching process; after the match has been found, it will skip to the next character of the secret message instead. Thus, it reduces computational complexity and saves time.

Table 4 Results of time consumption
Full size table
Fig. 7
figure 7

Number of characters and searching time

Full size image

From Table 5 and Fig. 8, it’s shown that the proposed system produces a larger key size than the current system. It happens because in the proposed system the key will be encrypted with AES encryption, but it is a good trade off because it provides more security and also the proposed system has a large payload capacity.

Table 5 Results of generated key size
Full size table
Fig. 8
figure 8

Number of character and key size

Full size image

Table 6 shows the comparison between the proposed method and other Steganography systems. It shows that the proposed method maintains the image quality and also has large payload capacity. The result is considered very promising, due to having a great payload capacity, high image quality, and low time consumption, and it also secures the key with encryption before sending it to the receiver. Furthermore the Table 7 shows the Comparative results of Current Best Solution and Proposed Method.

Table 6 Comparison between proposed method and Existing Steganography Method
Full size table
Table 7 Comparative results of Current Best Solution and Proposed Method
Full size table

5 Conclusion

The result above shows the difference in the number of possible matching values according to the ASCII number between the current method and proposed method. The result shows that the proposed method improves the possible matching numbers as well as the security of the generated key.

In conclusion, by using the entire color channels to find the matching value, it improves the possibility of the match. The AES also helps secure the generated key to secure the integrity of the exchanged information from third-parties. The limitation of the proposed method is that the matching value is still limited to the color distribution of the cover image. If the matches cannot be found, then the system will not be able to produce the key.

Considering the limitations, future research should investigate the possibility to produce the key from the cover image when the matching value cannot be found. Future research can also increase the security, availability, and also robustness to improve the method.

The proposed method and the current method have been examined by using NetBeans 8.2 and it has been demonstrated that the proposed method is capable of more matching values than the current method. The key produced by the proposed method is also more secure from third parties than the current method.