A secure fingerprint hiding technique based on DNA sequence and mathematical function

Problem statement

The secure transmission of sensitive data is a critical concern in today’s digital age. While DNA steganography offers a potential solution by utilizing the vast capacity of DNA sequences for data hiding, there is a need for an effective and secure technique that specifically addresses the protection of fingerprint images. Biometric systems, including fingerprinting, are widely used for individual recognition; however, if biometric information is compromised, it cannot be changed, necessitating robust security measures. The challenge lies in developing a technique that combines steganography and cryptography to securely transmit fingerprint images while maintaining confidentiality.

Motivation

The motivation behind this research is to address the need for a secure fingerprint hiding technique based on DNA sequences and mathematical functions. Current approaches often focus on general data hiding methods, but the specific requirements of fingerprint images necessitate a tailored solution. By leveraging the capabilities of DNA steganography and cryptography, this research aims to provide an efficient and secure method for embedding fingerprint images into DNA sequences. The proposed technique offers a low probability of being cracked, a high number of hiding positions, and efficient execution times. The utilization of the Feistel network encryption process, along with a mathematical function and an insertion technique, enhances the security and confidentiality of the embedded data. The evaluation of the technique using the NIST statistical test suite and comparison with other research papers further validates its effectiveness. By introducing random ambiguous bits and increasing noise, the technique adds an additional layer of security to protect against various attacks. However, it is important to acknowledge the limitations of the proposed technique, as it can only handle small images within DNA sequences and is not suitable for larger-scale data such as video or audio.

Organization of the article

The article is structured into several sections for better organization and coherence ‘Introduction’ encompasses the introduction, main objective, and motivation. Moving forward, ‘Cryptographic Background’ delves into the cryptographic background, providing essential context for the study. In ‘Related Works’, the article presents a comprehensive review of related works. Section ‘Design and Methodology’ outlines the research methodology, while also detailing the design and methodology. The findings and outcomes of the study are presented in ‘Experimental Results’, titled ‘Results’. Furthermore, ‘Discussion’ provides an in-depth exploration of the discussion surrounding the results. Finally, the article concludes in ‘Conclusion’.

Biological background

DNA is the molecule that carries genetic information in human beings (Giuliani, Sarti & Di Virgilio, 2019). It is present in almost every cell of the human body and determines characteristics such as eye color, hair color, skin color, and sex (Campbell, 2017; Rosenberg, 2017). Human DNA consists of approximately 3 billion bases and is organized into 23 pairs of chromosomes (46 in total) found in the nucleus of each cell (Campbell, 2017; Jindal et al., 2017). DNA is composed of nucleotides, which are small subunits with four types of bases: adenine (A), thymine (T), guanine (G), and cytosine (C) (Abbasy et al., 2012; Appuswamy et al., 2019; Kiss, 2018; Shen, 2019). These nucleotides are connected through complementary base pairings, with A bonding to T through two hydrogen bonds and C bonding to G through three hydrogen bonds (Khalifa, 2013; Kiss, 2018; Management Association, 2018; Shen, 2019).

A biometric system is a technology that processes information about an individual to identify and distinguish them. Fingerprint identification is a widely used biometric method due to the unique characteristics of each person’s fingerprints (Daluz, 2018). Fingerprints are an inherent part of the biometric science, which utilizes physical characteristics for identification purposes (Douglas et al., 2018). Each person’s fingerprints are unique, with no two fingerprints found to be the same among different individuals (Daluz, 2018; O’Hagan & Calder, 2020). This uniqueness extends to each finger having a distinct fingerprint, and fingerprints do not change with age (Alsmirat et al., 2019). As a result, the probability of two individuals having the same fingerprint is extremely low, estimated to be one in 64 billion (O’Hagan & Calder, 2020). However, if a fingerprint image is stolen, it cannot be modified or changed. Therefore, it is crucial to ensure the security and protection of sensitive fingerprint images. This research aims to develop an efficient and secure technique for hiding sensitive images, including fingerprint images. The technique involves converting a fingerprint image into a binary representation and then embedding it into a DNA sequence after applying appropriate encryption methods.

The proposed algorithm

The proposed algorithm comprises three algorithms for the encryption and hiding process, as well as three algorithms for decryption and extracting the fingerprint image. Let us delve into each of these algorithms:

1. Encryption and hiding algorithms:

   1. Algorithm 1: Fingerprint Encryption—This algorithm takes the fingerprint image and encryption keys as inputs. It converts the image into a sequence of bits and performs encryption using a specific encryption algorithm. The output is the encrypted sequence of bits representing the fingerprint.

   2. Algorithm 2: DNA Sequence Generation—This algorithm generates a DNA sequence based on the encrypted fingerprint and a set of predefined rules, see Table 2. It maps the encrypted bits to corresponding DNA bases, creating a DNA sequence that carries the encrypted fingerprint information.

   3. Algorithm 3: DNA Hiding—This algorithm hides the DNA sequence within a larger DNA sequence or within a specific DNA region. It employs techniques like insertion or substitution to embed the DNA sequence, making it less apparent and increasing the difficulty of detection.

2. Decryption and extraction algorithms:

   1. Algorithm 4: DNA Extraction—This algorithm extracts the hidden DNA sequence from the larger DNA sequence or specific DNA region. It identifies and isolates the portion of DNA that contains the hidden information.

   2. Algorithm 5: DNA Decoding— This algorithm decodes the extracted DNA sequence back into the encrypted fingerprint sequence of bits, reversing the mapping process performed during encryption.

   3. Algorithm 6: Fingerprint Decryption—This algorithm decrypts the sequence of bits obtained from the DNA decoding process, using the decryption keys and the appropriate decryption algorithm. The output is the original fingerprint image.

Table 3 provides a summary of the notations used in the implementation of these algorithms, along with their descriptions. These notations assist in understanding the variables, parameters, and operations involved in the algorithms. By implementing these algorithms, the proposed technique aims to securely encrypt and hide a fingerprint image within a DNA sequence. The decryption and extraction algorithms facilitate the retrieval of the original fingerprint image from the hidden DNA sequence. This technique can be useful for protecting sensitive fingerprint data and ensuring its confidentiality.

Encoding and hiding process

The encoding and hiding process of the proposed technique can be described using three algorithms: the Feistel network algorithm, the converting binary image to DNA base algorithm, and the hiding process algorithm. The flowchart for these procedures is depicted in Fig. 1. An overview of each algorithm’s role:

Decoding and extracting a fingerprint process

The decoding and extracting process in the proposed technique involves three algorithms: the extracting image algorithm, the converting DNA base to binary image algorithm, and the reversing Feistel network algorithm. These algorithms are responsible for retrieving the hidden fingerprint image from the encrypted and hidden DNA sequence. The flowchart for the decryption and extraction procedures can be seen in Fig. 2.

Extracting image algorithm

The extracting image algorithm utilizes three keys to separate the fingerprint base sequence from the real DNA base sequence. These keys are essential for correctly extracting the hidden information and ensuring the integrity of the retrieved fingerprint image.

Converting DNA base to binary image algorithm

The converting DNA base to binary image algorithm performs the reverse process of the “Converting Binary Image to DNA Base” algorithm used during encoding. It converts the fingerprint base sequence, obtained from the extraction process, back into a binary sequence. This step is necessary to restore the original format of the fingerprint image.

Reversing feistel network algorithm

The reversing Feistel network algorithm acts as the counterpart to the encryption Feistel network algorithm. It decodes the binary sequence representing the fingerprint image, reversing the encryption process performed during encoding. By applying the reverse steps of the Feistel network, including permutation and decryption rounds, the algorithm produces the decrypted fingerprint image.

Figure 2 illustrates the flowchart of the entire decoding and extracting process, providing a visual representation of how the three algorithms work together to recover the fingerprint image from the hidden DNA sequence.

Ambiguity bits

Additionally, the presence of ambiguity bits is mentioned, which are added to enhance the security of the encryption process. These bits introduce small terms into the ciphertext during encryption, increasing the noise level in the transmitted data. The receiver, equipped with techniques to decode the ambiguity bits, can recover the original information. The inclusion of ambiguity bits makes data identification and decryption more challenging for unauthorized users, thereby enhancing security and protecting against unauthorized access. It’s worth noting that the use of ambiguity bits also reduces data efficiency to some extent, as additional bits need to be transmitted. However, this trade-off is deemed necessary for stronger encryption and increased security (Hamad, 2014).

Cracking probability

Cracking probability refers to the likelihood of successfully breaking the proposed technique and uncovering the hidden secret message, in this case, the fingerprint image. It is an important measure of the security level provided by the hiding mechanism. The cracking probability is influenced by various factors and is calculated by considering specific variables used in the technique to hide the sensitive data.

In the context of the proposed algorithm, the cracking probability was assessed to determine the probability of an attacker successfully revealing the hidden fingerprint image using cryptographic or steganographic methods. The calculation of cracking probability takes into account six factors, which are crucial in evaluating the security of the technique.

Reference of DNA

The reference sequences that were used in the proposed technique were real sequences. There are several public databases that work as banks for real DNA. Together, they provide approximately 163 million DNA sequences available publicly. Therefore, this is the first piece of information the intruders need to break the fingerprint image in the DNA sequence. Thus, in the worst case, the probability of predicting a reference DNA sequence from a set of 163 million sequences is: (1)${\displaystyle DNASeq=\frac{1}{1.63\ast 10^8}}$

Binary coding rule

The binary coding of DNA bases A, C, G, and T gives different combinations of two bits, which is equal to 4! = 24. The probability of an intruder making a successful guess at this stage is (2)${\displaystyle BCR=\frac{1}{24}}$

Size of the message

This factor refers to the probability of an attacker’s successfully revealing the size of a fingerprint image and prefix DNA sequence. The fake DNA is available to the attacker to crack the hidden fingerprint image, and the probability of an intruder’s successfully guessing is: (3)${\displaystyle MsgSize=\frac{1}{n-1}}$

The Random segments of the secret bits

This factor refers to the fingerprint image bits that are segmented to hide inside the DNA sequence. It is difficult for intruders to know how many segments are divided. Thus, they would need to try two segments, three segments, four segments, and so on. Therefore, the probability of guessing the segmentation of an image is: (4)${\displaystyle Rssb=\frac{1}{2s-1}}$

The random segments of the DNA sequence

This factor refers to the DNA sequence that is segmented randomly by generated key values and thus the probability of an intruder’s making a successful guess for: (5)${\displaystyle Rsds=\frac{1}{2s-1}}$

Feistel encryption process

The encryption process that was chosen for the proposed technique was a Feistel network. A Feistel structure consists of a single round of encryption which itself consists of a substitution step followed by a permutation step. The probability of an intruder making a successful guess is: (6)${\displaystyle Fes=\frac{1-1}{2n}}$

According to the previous analysis, the cracking probability of an intruder making a successful guess at the proposed algorithm is given by: (7)${\displaystyle TheProposedTechnique=\frac{1}{1.63\ast 108}\times \frac{1}{24}\times \frac{1}{n-1}\frac{1\times }{2m-1}\times \frac{1-}{2s-1}\frac{1}{2n}}$

Execution time

Execution time is an important factor to consider when evaluating the efficiency of a technique. In Fig. 4, we present a comparison of the execution time of the proposed technique with two related techniques, referred to as Srilatha & Murali (2016) and Biswas et al. (2019). The comparison was conducted on different file sizes, ranging from 1 KB to 15 KB. As expected, larger file sizes generally required more time to execute the technique.

The results showed that the proposed technique achieved lower execution time values compared to Srilatha & Murali (2016) for the different file sizes. This indicates that the proposed technique is more efficient in terms of execution time when compared to Srilatha & Murali (2016). However, it is worth noting that the proposed technique had a higher execution time compared to Biswas et al. (2019). This suggests that Biswas et al. (2019) may have better performance in terms of execution time than the proposed technique. Overall, the comparison of execution time provides insights into the relative efficiency of the proposed technique compared to other related techniques. It helps to assess the computational cost associated with implementing the proposed technique and allows for a better understanding of its performance in practical applications.

Decoding and extracting time

Decoding and extracting time is another important aspect to consider when evaluating the performance of a data hiding technique. In Fig. 5, we present a comparison of the decoding and extracting time between the proposed technique and two related techniques, referred to as Srilatha & Murali (2016) and Biswas et al. (2019).

The comparison was conducted on various file sizes, similar to the execution time comparison. However, it is important to note that the proposed technique generally took more time for decoding and extracting procedures compared to its execution procedures.

The results showed that the proposed technique had higher decoding and extracting time values compared to Srilatha & Murali (2016) and Biswas et al. (2019) for several file sizes. This indicates that the proposed technique requires more time to decode and extract the hidden data compared to the other techniques.

One of the main factors contributing to the increased decoding and extracting time in the proposed technique is the process of converting a fingerprint image into binary and vice versa. This conversion process takes a significant amount of time. It is important to note that the other techniques do not involve this conversion process since their final form of data is text, while the proposed technique is specifically designed for hiding an image.

Overall, the higher decoding and extracting time of the proposed technique compared to other techniques can be attributed to the additional steps involved in handling image data. Although the decoding and extracting time is relatively longer, it is necessary to consider the specific requirements and objectives of the technique, particularly when it comes to hiding image data.

Cracking probabilities

The cracking probabilities of the proposed technique and several other techniques discussed in the literature review, including Malathi et al. (2017), Sajisha & Mathew (2017), Agrawal, Srivastava & Sharma (2014), Marwan, Shawish & Nagaty (2015), Abdullah, Eesa & Abdo (2019), are presented in Table 4. The proposed technique demonstrated the lowest cracking probability among these techniques, indicating a higher level of security.

BPN

In terms of bits per nucleotide (BPN), which refers to the number of secret bits that can be embedded in each nucleotide of the DNA sequence, the proposed technique achieves an average of 1.53 BPN. This value is approaching the upper bound of 2 BPN, indicating efficient utilization of the DNA sequence for data hiding. Table 5 illustrates the performance of the proposed technique in hiding secret bits in eight DNA sequences, comparing it with three existing techniques: insertion, substitution, and complementary pair techniques. The proposed technique outperforms the existing techniques, providing higher BPN values, as shown in Fig. 6.

Furthermore, the BPN results of the proposed technique were compared with the results of Malathi et al. (2017), Agrawal, Srivastava & Sharma (2014), Marwan, Shawish & Nagaty (2015) techniques, as shown in Tables 5 and 6. The comparison results indicate that the proposed technique achieves competitive BPN values. A visual representation of the BPN results can be seen in Fig. 7.

Overall, the cracking probabilities demonstrate the enhanced security of the proposed technique, while the BPN analysis highlights its efficient utilization of the DNA sequence for data hiding compared to existing techniques.

The NIST result

The NIST statistical test suite results of the proposed technique were compared with the results from Biswas, Gupta & Haque (2019). Table 7 presents the comparison results, with the proportion shown in the fourth and fifth columns. It can be observed that the proposed technique’s runs test and universal test results are clearly outside the expected interval, while Biswas, Gupta & Haque (2019) only has the longest run test outside the interval. This indicates that the sequence generated by the proposed technique, as well as the longest run test in Biswas, Gupta & Haque (2019), cannot be considered completely random. The non-random results are marked with an asterisk in the table. The level of significance α for all tests is considered to be 0.01, and a P-value below α is required to prove randomness. The P-value results, shown in the second and third columns of Table 7, reveal that the proposed technique has a P-value of zero for the universal and approximate entropy tests, indicating non-randomness. Similarly, Biswas, Gupta & Haque (2019) also has three test results below the standard P-value, namely rand, universal, and approximate entropy tests. Therefore, both techniques exhibit non-randomness in the universal and approximate entropy tests with zero values. However, the proposed technique has higher P-values than Biswas, Gupta & Haque (2019) in several tests, such as block frequency, cumulative sums, rank, FFT, non-overlapping template, and linear complexity.

On the other hand, Biswas, Gupta & Haque (2019) surpasses the proposed technique and has higher P-values in frequency, runs, longest run, overlapping template, and serial tests. In terms of key space analysis, the proposed technique utilizes four different random keys: the encryption key (EK), DNA segment key, fingerprint segment key, and first position key. The EK is used in the Feistel network technique employed in the encryption scheme. It has a size of 32 bits, as it operates on the right section of a 64-bit block. The DNA segment key, fingerprint segment key, and first position key are used in the hiding process. The key space of the proposed technique comprises all possible permutations of these keys, which should be large enough to provide protection against various attacks.as u seen in the following discussion:

Key space

In the proposed technique, the key space refers to the set of all possible permutations of the keys used in the encryption algorithm. A larger key space provides greater security against various attacks. In this technique, four different random keys are employed: the EK, DNA segment key, fingerprint segment key, and first position key. Each key serves a specific purpose in the encryption process.

Encryption key

The EK in the proposed technique is used in the Feistel network encryption scheme. This scheme operates by dividing the secret fingerprint image into blocks, with each block consisting of 64 bits. Each block is further divided into two sections: the left section and the right section. During the encryption process, the right section of the block remains unchanged. However, the left section undergoes an encryption operation that takes two inputs: the encryption key and the right section. The encryption process applied to the left section depends on the specific algorithm used in the Feistel network. In the proposed technique, the size of the EK is determined by the size of the right section, which is 32 bits. Therefore, ${\displaystyle \text{the encryption key size is}=2^{32}}$ resulting in a key space of 4,294,967,296 possible permutations for the encryption key. This large key space contributes to the security of the encryption process and makes it more challenging for unauthorized users to decrypt the encrypted data.

DNA segment key

The DNA segment key is one of the keys used in the hiding process. It is employed to select a specific segment of the DNA sequence that will be utilized to hide the fingerprint image.

In the proposed technique, there are three additional random keys used in the hiding process: the DNA segment key (DSK), the fingerprint segment key (FSK), and the first position key (FPK). The DNA segment key (DSK) determines how the DNA sequence will be divided into segments to accommodate the hiding of secret fingerprint bits.

This key is chosen randomly and has a key space of 2⁸, which means there are 256 possible combinations.

Fingerprint segment key

The fingerprint segment key (FSK) is used to divide the fingerprint DNA sequence into segments that will be hidden within the DNA sequence. Like the DSK, this key is chosen randomly and also has a key space of 2⁸.

First position key

The first position key (FPK) is used to determine where the hiding process will start within the DNA sequence after the encryption process. This key is chosen randomly and has a key space of 2³².

Therefore, the proposed technique has a key space of ²80, which represents a large number of possible combinations for the keys and enhances the security of the hiding mechanism. To calculate the overall key space (KS) of the proposed technique, we multiply the key spaces of all four keys together: (8)${\displaystyle KS=\left(2^{32}\right)\ast \left(2^8\right)\ast \left(2^8\right)\ast \left(2^{32}\right)=2^{80}.}$

By combining these four different random keys in different permutations, the key space of the proposed technique is formed. A larger key space increases the complexity of the encryption process and makes it more challenging for attackers to decipher the hidden information.

Key sensitivity

In the proposed technique, the sensitivity of the keys was tested to evaluate the robustness of the encryption and hiding process. The sensitivity test involved modifying a single bit in any key value of the decryption process while keeping the other key values unchanged. The purpose of this test was to determine if even a minor alteration in any key would affect the decryption process and prevent the correct recovery of the fingerprint image. The results of the sensitivity test showed that when any bit of the keys was modified, the decryption process failed, and a new encrypted image was generated instead of recovering the original fingerprint image. This indicates that the proposed technique is highly sensitive to key alterations. Even a small change in any key value can significantly impact the decryption process and prevent the accurate retrieval of the hidden fingerprint. As shown by Table 8 key sets used in the proposed technique.

The high sensitivity of the proposed technique to key modifications enhances its security. It ensures that any unauthorized alteration in the keys would render the decryption process ineffective and pre vent unauthorized access to the hidden fingerprint image.

Resistance to several attacks

The proposed technique has been evaluated for its resistance against various types of attacks. One such attack is the known-plaintext attack (KPA), where the attacker possesses some known plaintexts and their corresponding ciphertexts. The objective of the attacker in a KPA is to analyze this information and deduce the encryption key used in the system.

Known-plaintext attack

To ensure the secrecy of the proposed technique against KPA, it incorporates multiple mechanisms in addition to relying on keys. These additional mechanisms enhance the security of the technique and make it highly resistant to such attacks. As a result, the cracking probability of the proposed technique is very low, indicating that it is highly unlikely for a known-plaintext attack to successfully break the security of the technique. By employing multiple layers of protection and not solely relying on keys, the proposed technique provides a robust defense against known-plaintext attacks. Its low cracking probability demonstrates its effectiveness in preserving the confidentiality and integrity of the encrypted data, making it a secure choice for protecting sensitive information.

Chosen-plaintext attacks

In chosen-plaintext attacks (CPA), the attacker has the ability to select specific plaintexts and obtain their corresponding ciphertexts. It is important to note that there is no deterministic encryption technique that can provide complete security against chosen-plaintext attacks. Instead, any encryption technique that aims to be secure against CPA must be probabilistic in nature. To demonstrate the resistance of the proposed technique against CPA, the XOR test operation is used. The XOR operation is performed on two pairs of plaintexts and their respective ciphertexts: (9)${\displaystyle A1\oplus A1\prime =B1\oplus B1.}$

If the above equation holds true, it indicates that the proposed technique can resist chosen-plaintext attacks. The XOR test helps verify that the encryption process is probabilistic and not vulnerable to attacks where the attacker can selectively choose plaintexts and observe their corresponding ciphertexts. By exhibiting the desired properties in the XOR test, the proposed technique demonstrates its ability to withstand chosen-plaintext attacks and maintain the security of the encrypted data.

Spoof attacks

Spoof attacks are a type of active attack that exploit weaknesses in biometric systems, including fingerprint systems, to deceive the system and gain unauthorized access. These attacks involve using artificial or fabricated fingerprints to bypass the biometric sensor. Spoof attacks have been found to have a high success rate, with over 70% effectiveness in breaking fingerprint systems. To address this vulnerability, the proposed technique converts the fingerprint image into a long sequence of DNA bases. This conversion process makes it extremely difficult for an attacker to steal a fingerprint image stored as DNA within the system. The attacker would need to extract fingerprints from the DNA, remove ambiguity bits, and decrypt the data using the Feistel network technique to recover the original image, making it virtually impossible to steal the fingerprint image.

Ciphertext-only attack

A ciphertext-only attack (COA) is a type of attack where the attacker only has access to the encrypted ciphertext and attempts to determine the original plaintext. One example of a COA is a brute force attack.

Brute force attack

A brute force attack involves systematically trying every possible key or password until the correct one is discovered. In the context of the proposed technique, if an attacker attempts a brute force attack, they would need to try every possible combination of keys: the encryption key, DNA segment key, fingerprint segment key, and first position key. The use of four different keys in the proposed technique results in a large key space, making it computationally infeasible for an attacker to exhaustively try all possible keys within a reasonable time frame. This key space sensitivity strengthens the resistance of the proposed technique against brute force attacks.