A novel approach to secure communication in mega events through Arabic text steganography utilizing invisible Unicode characters

Introduction

The internet has completely changed the technological means via which data is shared in many social interactions. As globalization has progressed, there has been a higher dependence on social media to share data by both individuals and organizations especially during mega events (Al-Khaldy et al., 2022). When a text message is sent via short message service (SMS), email, or social media, the information included in the message is transmitted as plain text, which implies that such information is vulnerable to malicious attacks and unauthorized access. In some cases, this information may be sensitive or confidential, such as passwords, banking credentials, or even some secrets that an individual seeks to send to a friend. As a result, there is an increasing need for intelligence and multimedia security studies that incorporate covert communication, whose primary goal is data concealment (Ahvanooey et al., 2019).

Text hiding, also known as information or data hiding, has gathered a lot of interest recently because of its widespread use and its applications in the network communication and cybersecurity sectors. It is the process of embedding secret data such that it is invisible to adversary or casual readers using a cover text. In the context of mega events, data hiding is important for several reasons. Mega events attract massive crowds, making it easier for eavesdroppers to blend in and covertly listen to conversations. The sheer volume of people creates a challenging environment for security personnel to monitor every individual effectively. Mega events often involve the exchange of sensitive information among organizers, security teams, and other stakeholders. Eavesdropping on these communications can lead to the compromise of security protocols, emergency plans, or confidential details. Mega data breaches, which usually involve millions of compromised data records, can incur significant costs. Figure 1 shows the cost of mega data breaches in 2023 (Invisible Characters, 2022). As depicted in the figure, the greater the amount of data compromised, the higher the cost. As the number of cyber attacks are increasing dramatically in recent mega events, as depicted in Fig. 2 (Poitevin, 2023; SOCRadar, 2022), the cost of data breaches due to these attacks could be substantial unless data is protected in a secure manner. Data hiding helps protect this information from unauthorized access or interception by potential threats. Furthermore, mega events typically use tickets or access credentials for attendees. Data hiding can be employed to embed additional security features, making it more difficult for counterfeiters to replicate tickets or access passes. This enhances the overall security of the event by reducing the risk of unauthorized entry. In addition, security teams and organizers need to communicate discreetly to ensure the effectiveness of security measures. Data hiding allows for covert communication, reducing the risk of intercepted messages and maintaining the element of surprise in security operations. Also, attendees at mega events may have concerns about privacy, especially in the age of pervasive surveillance. Data hiding techniques can be used to secure personal information, making it more challenging for malicious actors to exploit or misuse attendees’ data (Ahvanooey et al., 2019).

Information hiding is classified into two categories in practice: watermarking and steganography. Watermarking safeguards cover media against damaging attacks such as modifications, forgeries, and plagiarism by demonstrating ownership. In contrast, steganography is concerned with the invisible transmission of confidential information in such a way that no one can detect it. In other words, the goal of steganography is concealing the fact that a medium contains secret data (Ahvanooey et al., 2019).

Steganography is the art and science that involves hiding messages within other messages to conceal information. The phrase “secret message” refers to the concealed message, while “cover message” or “cover media” refers to the message used to cover the secret message. The resulting message, which combines the cover message and the secret message, is known as the “stego message” (Souvik, Banerjee & Sanyal, 2011; Gutub, Al-Alwani & Mahfoodh, 2010). The goal of steganography is to conceal the secret message by making use of any redundancy in the cover message while maintaining the integrity of the cover media (Gutub & Fattani, 2007).

Basically, any digital file format can be used as a cover media. Based on the cover media, steganography techniques can be classified into five main categories, image, audio, video, text, and network steganography (Jebur et al., 2023). Information is concealed using the noise present in the cover media in image, audio, and video steganography. On the other hand, in text steganography, redundancy to conceal secret information can take on different forms, including altering a text’s formatting, changing some words, and using random character sequences (Bennett, 2004). In network steganography, data is concealed by making use of the header and payload fields of network protocols. This is achieved by creating hidden channels between a secret sender and a secret recipient (Jebur et al., 2023).

Text-based steganography poses unique challenges due to the scarcity of redundant information in text files compared to other carrier files like images or audio, making it a particularly challenging form of steganography to implement effectively. However, text steganography offers several practical advantages. It requires minimal memory for storage, facilitating efficient memory usage. Its lightweight nature enables seamless transmission over networks, ensuring swift data exchange. Additionally, it reduces printing costs by embedding data within textual content, making it cost-effective for printing. Text steganography is also widely used across social media platforms for covert communication due to its simplicity and long-standing history of use. These practical benefits of text steganography emphasize its efficiency, versatility, and adaptability across various communication channels and mediums, making it the most suitable tool for secure communication in mega events. Therefore, this article focuses on text steganography as the preferred method for ensuring discreet and confidential communication amidst the unique challenges of mega events.

More specifically, this work focuses on the Arabic language, which holds considerable importance as a widely spoken language across numerous countries. It holds profound cultural and religious significance as the language of the Quran, the central religious text in Islam. With approximately 1.6 billion Muslims worldwide, Arabic serves as the language of a significant portion of the global population (Al-Nofaie & Gutub, 2020). Moreover, given that one of the largest and most significant mega events in the world, Hajj, occurs within an Arabic-speaking context, targeting the Arabic language in this article is both logical and pertinent.

The proposed method in this article is based on hiding secret bits within Arabic letters through the utilization of invisible Unicode characters, where each letter can hide two secret bits. With this capacity of each letter, this technique offers a notably greater capacity than previous approaches.

The article is organized as follows. ‘Background Material’ provides a background material about the features of Arabic language and Unicode encoding utilized in this study. In ‘Related Work’, a review of the related work is covered. The methodology of the proposed work is explained in ‘Methodology’. In ‘Experimental Results’ the experimental results of the proposed work are discussed and a comparison with similar previous methods are discussed in ‘Discussion and Comparison’. Finally, the article is concluded in ‘Conclusion’ highlighting some potential directions for future research.

Background Material

This work utilizes some characteristics of Arabic letters and the Unicode encoding system. In this section, these features and characteristics are highlighted. First, the characteristics of Arabic letters is discussed in ‘Arabic letter characteristics’. Then, the features of the Unicode encoding is covered in ‘Unicode characteristics’

Arabic letter characteristics

The Arabic language holds a significant position among human languages, and it is one of the sixth official languages of the United Nations. Around 400 million individuals worldwide use Arabic as their primary language. Its alphabet comprises 28 characters. Notably, Arabic letters are typically connected to each other in written texts, unlike languages like English, where characters are usually written individually (Thabit et al., 2021). Arabic language and its characters have some characteristics that might be useful when thinking about data hiding (Al-Nofaie, Fattani & Gutub, 2016; Alanazi, Khan & Gutub, 2022).

The unique right-to-left writing system of Arabic, contrasting with the left-to-right orientation of languages like English or French, is a significant characteristic. This unidirectional feature not only affects letter arrangement but also shapes the structure of written Arabic, including numerical representation. This consistency aids Arabic readers in processing text visually, ensuring a cohesive reading experience. Understanding this aspect is crucial in applications like data hiding and steganography, emphasizing the need to preserve readability while concealing information within Arabic text.

Another characteristic is connectivity of letters to neighbouring letters within words. Letters can occur at the start, end, middle of a word, or isolated. Some Arabic letters connect with both preceding and following letters, while others only join with preceding ones. Additionally, certain letters cannot connect to either preceding or following ones. Table 1 shows the connected and un-connected Arabic letters.

Furthermore, in Arabic script, some letters are characterized by the presence of dots, while others are free of dots, as shown in Table 2. The dots may be one, two or three, and they may be above or beneath the letter.

Table 1:

Connected and un-connected Arabic letters.

Arabic letters that may be connected to previous or following letters
Arabic letters that may be connected only to previous letters
Arabic letter that may not be connected to previous nor following letters

DOI: 10.7717/peerjcs.2236/table-1

Table 2:

Dotted and un-dotted Arabic letters.

Arabic letters that have dots
Arabic letters that are free of dots

DOI: 10.7717/peerjcs.2236/table-2

These characteristics are not unique for Arabic languages. Other languages, such as Persian and Urdu, have the same characteristics. Therefore, the proposed methodology could be adapted to be utilized in such languages.

Related Work

Various approaches have been proposed in Arabic text steganography, utilizing different techniques to conceal information within Arabic text. The main categorizes of these techniques are based on Arabic letters characteristics, diacritical marks, kashida character, Unicode encoding, and Arabic poetry system (Thabit et al., 2021).

One of the approaches that are based on Arabic letters characteristics involves altering the placement or arrangement of dots within Arabic letters to encode hidden information. By strategically manipulating dot patterns, steganographers can embed covert messages while maintaining the appearance of the text. Early research has explored the utilization of points within Arabic and Persian letters to conceal sensitive information. For instance, in Shirali-Shahreza & Shirali-Shahreza (2006) single secret bit (0 or 1) was hidden within Arabic letters by adjusting the positions of the dots. A similar study proposed in Odeh et al. (2012) addressed Arabic letters with multiple points, assigning two bits to each multipoint letter to double the hidden bits. However, a challenge arose with this method: the retyping process, which could destroy concealed bits. The authors proposed a solution by consolidating all data into a single file to limit future font changes. While this approach enhances capacity and decreases suspicion in the covert text, it is characterized by longer processing times, fixed output format, and vulnerability to retyping or scanning.

Another feature of Arabic letters, which is the sharp edges and geometric shapes, was also utilized to hide secret information. In Roslan, Mahmod & Udzir (2011), this technique was employed by using two keys to determine whether dotted or undotted letters are used for concealing secret bits and specifying the precise positions within the letters where the information is hidden. Each edge of a letter is assigned a unique code, facilitating the concealment of bits within those edges. Similarly, in Mersal et al. (2014), the concept of sharp edges was utilized, with a 24-bit random key concealed within the initial sharp edges of the letters in the cover text. The positioning of the letter within the text determines the sequence of random numbers used to extract the binary representation of the secret message. This binary number forms the code sequence for the message, with each code number representing a binary bit concealed within a specific character of the text. Furthermore, Roslan et al. (2014) introduced the primitive structural method, which incorporates sharp edges, dots, and typo proportion from calligraphy writing. Each Arabic character offers multiple hiding spots for secret bits, determined by its structural features and proportion calculation, enhancing the capacity for concealing information within the text.

Diacritical marks, such as vowel signs and other phonetic symbols, have been also utilized to encode hidden information within the text, exploiting the subtle variations in their placement and appearance. In Aabed et al. (2007), it was noted that the fatha diacritic is almost as prevalent as other diacritics in Arabic, leading to the suggestion of using it to encode “1”, with other diacritics representing “0”. However, this approach may attract undue attention from readers. An improved method proposed in Gutub et al. (2010b) introduced two algorithms based on the number of diacritics needed to conceal secret bits, with one utilizing fixed-size blocks of secret bits and the other mapping consecutive bits of the same value to their respective run lengths. Additionally, Gutub et al. (2008) explored steganography using multiple diacritics, employing both textual and image-based approaches. In Bensaad & Yagoubi (2013), three steganographic methods employing diacritics were discussed: one involving the inclusion or removal of diacritics based on secret bit values, another employing a switch technique, and a third introducing parity bits for each letter. Moreover, Memon & Shah (2011) employed reversed fatha in Arabic and Urdu text to convey concealed messages, a method that matches hidden messages character by character to cover articles. Ahmadoh & Gutub (2015) presented an embedding algorithm utilizing kasrah and fatha diacritics to hide messages, fragmenting messages into binary value arrays and concealing them within respective diacritics based on their parity. In another work, Malalla & Shareef (2016a) introduced a modified fatha for Arabic text steganography, which encrypts secret messages using AES before concealing them within text using a slightly altered fatha to avoid detection. Recently, Gutub (2024) proposed a hiding technique based on multiple diacritics. The author presented two models for the proposed technique. In each model, two diacritics are selected to hide secret bits. If the secret bit is 1, the selected diacritic is kept and the neighbouring diacritics other than the two selected diacritics are deleted. On the other hand, if the secret bit is 0, the selected diacritic is deleted and the neighbouring diacritic is preserved.

In another approach, kashida (also called extension or tatweel character) is utilized for steganography. Kashida, which is a typographical feature used in Arabic script to elongate certain letters for aesthetic or spacing purposes without affecting the content of the written message, can be manipulated to encode hidden data by varying the length or position of the elongated strokes within the text. In Gutub, Al-Alwani & Mahfoodh (2010), kashida was strategically employed with one kashida concealing “0” and two consecutive kashidas hiding “1”, optimizing the process by encoding all possible forms of Arabic letters using 6 bits and introducing a “finishing character” to denote completion. The MSCUKAT algorithm proposed in Al-Nazer & Gutub (2009) scans cover objects to identify suitable letter locations for kashida insertion based on secret bit values. Another algorithm proposed by the authors of Al-Haidari et al. (2009) encodes secret messages as numbers by strategically inserting kashidas within extendable letters. This concept was further developed by the authors of Gutub & Al-Nazer (2010) with the implementation of MSCUKAT. In Gutub et al. (2007), kashida was utilized with both pointed and un-pointed letters to conceal secret bits, while Gutub et al. (2010a) and Al-Haidari et al. (2009) enhanced security by selectively utilizing potential kashida positions and employing secret keys for bit-carrying positions. The kashida variation algorithm (KVA) in Odeh, Elleithy & Faezipour (2013) aimed to enhance robustness by randomly concealing bits in text blocks. Similarly, Alhusban & Alnihoud (2017) proposed four embedding strategies to conceal two secret bits using kashida after specific letters. Techniques such as compressing secret messages using Gzip and encrypting with AES, as shown in Malalla & Shareef (2016b), have been integrated with kashida embedding methods, along with other approaches like hiding voice files within text files using kashida and specific words (Al-Oun & Alnihoud, 2017). Despite the effectiveness of kashida for steganography, challenges such as susceptibility to suspicion and large output file sizes persist. Moreover, efforts to streamline algorithm complexity for improved extraction of hidden information remain limited.

In literature, Unicode characters have also been utilized for concealing information within text. One of the oldest proposals for using white spaces to hide secret bits was presented in Bender et al. (1996), where three scenarios were proposed to hide bits within text, the first at the end of sentences, the second at the end of lines, and the third between words. In Shirali-Shahreza & Shirali-Shahreza (2010), the resemblance between certain Arabic and Persian letters, such as “ ” and “ ” has been exploited, where Persian letters are used to represent “0” and Arabic letters for “1”. Additionally, in Shirali-Shahreza & Shirali-Shahreza (2008b), the isolated and connected forms of Arabic and Persian letters are leveraged to encode binary data, with the ZWJ character aiding in rendering the isolated forms. Furthermore, zero-space characters like ZWJ and ZWNJ were used in Shirali-Shahreza & Shirali-Shahreza (2008d) to hide secret bits by controlling letter connections. The special “La” character in Arabic is employed in Shirali-Shahreza (2007) and Shirali-Shahreza & Shirali-Shahreza (2008a) for concealing secret bits, with different representations used to encode “0” and “1”. Moreover, each Arabic letter’s multiple Unicode codes were utilized in Shirali-Shahreza & Shirali-Shahreza (2008c) to hide secret bits, distinguishing between representative and positional codes for encoding binary data. The concealment is performed word by word, as combining representative and contextual codes within a single word isn’t feasible due to text viewer limitations. The approach introduced in Alanazi, Khan & Gutub (2022) resolved this issue by employing pseudo-spaces and extension characters, thereby enhancing the method’s capacity for concealing information. In Obeidat (2017), three scenarios for concealing information in Arabic text were explored, involving character alterations, preserving cover letters, and adjusting Unicode based on letter type. Two Arabic text steganography methods were proposed in Al-Nofaie, Gutub & Al-Ghamdi (2021). The first, called Kashida-PS, builds upon previous work in Gutub, Al-Alwani & Mahfoodh (2010) and Al-Nofaie, Fattani & Gutub (2016) by incorporating the Kashida feature with PS. The second method, called PS-betWords, utilizes PS between words. Furthermore, Alotaibi & Elrefaei (2018) presented two Arabic text watermarking methods leveraging word spaces, with the first method enhancing a previous approach proposed in Alotaibi & Elrefaei (2016) by incorporating dotting features, and the second method selectively inserting different spaces based on watermark bits.

The Arabic poetry system, as described in Khan (2014), is capable of hiding secret bits, as each Arabic poem contains a representation of binary units. The approach involves assuming that the positions of embedded binary bits within poems contain the secret bits. The actual secret bit corresponds either to the binary position or its reverse. To enhance the capacity of this embedding technique, diacritics and kashida approaches are employed.

The combination of multiple steganographic techniques can enhance the concealment and robustness of hidden information. For example, combining Unicode with kashida was employed in several methods. In Al-Nofaie, Fattani & Gutub (2016), kashida is inserted between Arabic letters to represent a secret bit of one, while white spaces are used to represent zero, with two consecutive white spaces indicating the presence of a zero. A similar approach was adopted in Taha, Hammad & Selim (2020), where kashida and Unicode methods utilize small spaces for concealing bits, with specific patterns indicating the presence of a one. Moreover, Malalla & Shareef (2017) employed techniques like Gzip compression and AES encryption alongside kashida and Unicode methods to embed secret messages. In Alanazi, Khan & Gutub (2022), Medium Mathematical Spaces (MSPs), ZWJ, ZWJN, and kashida were combined to conceal one secret bit per alteration of format, whitespace, or kashida insertion. Kadhem & Ali (2017) proposed a method that merges Unicode with diacritics, utilizing RNA encoding for secret messages and non-printed characters to obscure codes, with compression using modified run length encoding (RLE). Lastly, Alshahrani & Weir (2017) explored the integration of kashida with diacritics, where fatah and consecutive kashidas are used to conceal different bits, demonstrating the versatility of combining different encoding techniques for steganographic purposes.

Table 3 summarizes the main pros and cons of previous work, classified by the techniques used. From this table, it is evident that methods utilizing kashida or diacritics raise suspicion, as kashida are not commonly used in most Arabic texts, and diacritics, when used, should appear on all letters, not just some. Other techniques, such as those using the points of letters or similar letters, have low capacity because not all letters can be utilized to hide secret data. Although the technique of using sharp edges does not add visible characters and can hide more secret bits, it is not robust enough due to the need for a reference table and code sequence that must be securely transmitted to the receiver and correctly retrieved. In general, methods utilizing Unicode have better imperceptibility, as they minimize the addition of visible characters. However, most of these methods cannot hide more than one bit per character and some suffer from incorrect connectivity between letters due to the use of different Unicode forms. Table 4 provides a more detailed analysis of methods based on Unicode techniques.

The proposed method in this study aims to leverage the benefits of Unicode techniques while addressing the limitations of previous work.

Methodology

The methodology of the proposed technique is based on hiding secret bits on the letters of a cover text utilizing invisible Unicode characters. The utilization of invisible Unicode characters to hide secret data has been proposed in several previous works, as discussed in ‘Related Work’. However, the novelty of the proposed methodology in this work is based on combining the utilization of invisible Unicode characters with two characteristics of the Arabic letters explained in ‘Arabic Letter Characteristics’. The first is whether the letter is dotted or not, which has been utilized for hiding secret data in several previous works such as (Roslan, Mahmod & Udzir, 2011; Alhusban & Alnihoud, 2017) and (Alotaibi & Elrefaei, 2018). The second is whether the letter is connected or not, which has not been utilized before –up to our knowledge –as proposed in this work.

Based on the two characteristics of each Arabic letter of a word, a two-bit value DC is assigned to that letter, as shown in Table 5. The two-bit DC of each character of the word is XORed with two secret bits. The result of the XOR is a two-bit value EN. Based on EN, an invisible white space is appended at the end of the word. Using this methodology, each letter hides two secret bits and will correspond to one white space. Therefore, four invisible characters are required.

The selection of invisible Unicode characters to be utilized in the proposed method is based on the following criteria:

1. Any invisible Unicode character inserted should not be visible in any way. In other words, the cover Arabic text should not be visually changed after inserting the invisible Unicode character. This increases the imperceptibility of the proposed method.

2. The inserted invisible Unicode characters should not be omitted when the stego text is sent via communication media or copied and pasted through different software. This increases the robustness of the proposed method.

Among the available invisible Unicode characters, we selected the four invisible characters shown in Table 6, which all satisfy the above mentioned criteria (Invisible Characters, 2022; Unicode Explorer, 2024).

In addition to the above four invisible white spaces used to hide secret bits, two other invisible characters are used at the end of the hiding process, as will be explained below. These two characters are \ufe00 and \ufe01.

UTF-16 is used in this work because it has fixed size of two bytes, while UTF-8 has different sizes. Specifically, some of the Arabic letters and the selected invisible characters are represented in more bytes using UTF-8 than UTF-16.

The advantages of the proposed method compared to previous work can be summarized in the following:

1. The proposed method can hide two secret bits for each character, while all previous methods based on Unicode could hide one secret bit per character at most. This advantage increases the capacity of the proposed method in comparison to other works.

2. Using the four selected white spaces to hide secret bits after words of the cover text increases the imperceptibility of the stego text. In other words, no one can notice the insertion of these white spaces in the stego text since there is no visible difference between the cover text and the stego text.

3. Adding one white space for each character helps in detecting alterations of the stego text, since the number of white spaces after each word should be equal to the number of characters in that word.

In addition, it is worth noting that secret bits could be hidden directly using the selected four invisible Unicode characters instead of XORing them with the characteristics of the corresponding Arabic letter. However, using the proposed XORing step has several advantages:

1. The secret bits cannot be extracted directly from the Unicode characters without knowing the characteristics of the corresponding letter in the text. This is important if someone has noticed the insertion of invisible characters and tried to extract secret bits from them.

2. The XORing with the characteristics of the Arabic letters will scramble the pattern of secret bits. The advantage of this is specifically important for structured secret bits, such as all ones, all zeros, or alternating patterns.

The proposed method consists of two main processes: one for embedding the secret message in the cover text, and the other for extracting it from the stego text. In the following, these two processes are explained in detail.

Embedding algorithm

The embedding process, which is shown in detail in Algorithm 1 , consists of the following steps:

1. Preprocessing: In this step, the number of secret bits is checked. If it is odd, a redundant “0” is appended to the end of the secret bits. This is because each character will hide two secret bits, so the number of secret bits should be even. In addition, the number of characters in the cover text should be more than or equal to twice the number of secret bits.

2. Hiding secret bits: For each word of the cover text, the following steps are performed:

   1. Find the number of characters in each word, which is checked by reading characters until a normal space (\u0020) is found.

   2. For each character of the word, a two-bit value DC is given based on its characteristics, as shown in Table 5.

   3. The two-bit DC of each character of the word is XORed with two secret bits S ₁S ₀. The result of the XOR is a two-bit value EN; as illustrated in the following equation: (1)${\displaystyle EN=DC\oplus S_1{S}_0}$

   4. Based on EN, one of the invisible white spaces is appended at the end of the word, as shown in Table 7.

   5. The above steps are repeated until all secret bits are hidden.

3. Post-processing: After all secret bits are encoded, an invisible Unicode character is appended to indicate the number of secret bits, whether odd or even. For odd number of secret bits, \ufe00 is inserted. Otherwise, \ufe01 is appended.

A flowchart of the embedding algorithm is shown in Fig. 3, which highlights the main steps of the algorithm. Table 8 shows an example of the embedding process, where the inserted invisible white spaces are bolded.

Table 7:

The corresponding invisible white space for the two-bit value EN.

EN	The corresponding invisible white space
0 0	\u061c
0 1	\u200f
1 0	\u200c
1 1	\u034f

DOI: 10.7717/peerjcs.2236/table-7

Algorithm 1: Embedding Process
Input: Secret bits, Cover text
Output: Stego text
1.	i= 1
2.	CharCount= 0
3.	SecretCount= number of secret bits
4.	ifSecretCount mod 2 = 1
5.	Append 0 to secret bits \\add a redundant bit if number of secret bits is odd
6.	end if
7.	while not End of Secret Bits
8.	Read next two secret bits S 1S 0
9.	Read ith character Text[i ] of the cover text
10.	T [i ] = Unicode(Text[i ])
11.	whileT [i ] ! = space (\u0020)
12.	CharCount++ \\count number of characters in each word
13.	ifT [i ] is dotted
14.	D [i ] = 1
15.	else
16.	D [i ] = 0
17.	end if
18.	ifT [i ] is not connected
19.	C [i ] = 0
20.	else
21.	C [i ] = 1
22.	end if
23.	DC [i ] =D [i ] & C [i ] \\& is append operation
24.	EN=DC [i ] ⊕S 1S 0
25.	caseEN:
26.	00: W [CharCount ] = \u061c
27.	01: W [CharCount ] = \u200f
28.	10: W [CharCount ] = \u200c
29.	11: W [CharCount ] = \u034f
30.	end case
31.	i ++
32.	Read Text[i]
33.	T[i] = Unicode(Text[i])
34.	end while
35.	j=i –1
36.	for (count= 1 to CharCount; count ++)
37.	insert W [count ] at T [j ]
38.	end for
39.	CharCount= 0
40.	end while
41.	ifSecretCount mod 2 = 0
42.	T [i +1] = \ufe01 \\indicates end of even number of secret bits
43.	else
44.	T [i +1] = \ufe00 \\indicates end of odd number of secret bits
45.	end if
46.	for (k= 1 to i +1, k ++)
47.	OutputFile[k ] = Character(T [k ]) \\convert unicode into characters
48.	end for
49.	while not End of cover text \\copy the remaining characters from cover text to output file
50.	Read ith character Text[i] of the cover text
51.	OutputFile[k ] = Text[i ]
52.	k++; i++
53.	end while

Extracting algorithm

Algorithm 2 shows the details of the extracting process, which is depicted in Fig. 4. The algorithm consists of the following steps:

Table 8:

An example of the embedding process.

The bolding indicatines where invisible white spaces are inserted.

Cover text
Unicode of the cover text	\u0644\u0627\u0020 \u0625\u0644\u0647\u0020 \u0625\u0644\u0627\u0020 \u0627\u0644\u0644\u0647\u0020 \u0645\u062d\u0645\u062f\u0020 \u0631\u0633\u0648\u0644\u0020 \u0627\u0644\u0644\u0647
Secret bits	0100011111100011111011
Unicode of the Stego text	\u0644\u0627\u061c\u061c\u0020 \u0625\u0644\u0647\u200f\u200c\u200c\u0020 \u0625\u0644\u0627\u200c\u200f\u034f\u0020 \u0627\u0644\u0644\u0647\u034f\u034f\u200c\ufe01\u0020 \u0645\u062d\u0645\u062f\u0020 \u0631\u0633\u0648\u0644\u0020 \u0627\u0644\u0644\u0647

DOI: 10.7717/peerjcs.2236/table-8

1. Read a character from the stego text until one of the two invisible white spaces indicating the end of secret bits (\ufe00 or \ufe01) is found.

2. Until the read character is a space (\u0020), do the following:

   1. Count the number of Arabic letters and the number of invisible white spaces.

   2. For each Arabic character, find the corresponding two-bit value DC, as explained in Table 5.

3. For each word, the value of each Arabic Letter DC is XORed with the value EN of the corresponding invisible white space, as shown in Table 7, which results in two of the secret bits S ₁S ₀. This is illustrated using the following equation: (2)${\displaystyle S_1{S}_0=DC\oplus EN}$

4. If the last read character is the invisible white spaces (\ufe00), which indicates an odd number of secret bits, ignore the last extracted secret bit.

Algorithm 2: Extracting Process
Input: Stego text
Output: Extracted secret bits
1.	i= 1
2.	CharCount= 0
3.	SpaceCount= 0
4.	Read ith character Text[i ] of the Stego text
5.	T [i ] = Unicode(Text[i ])
6.	whileT [i ] ! = (\ufe00 or \ufe01) \\not end of secret bits
7.	whileT [i ] ! = (\u0020) \\not end of a word
8.	caseT [i ]:
9.	\u061c:
10.	SpaceCount ++
11.	S [SpaceCount ] = 00
12.	\u200f:
13.	SpaceCount ++
14.	S [SpaceCount ] = 01
15.	\u200c:
16.	SpaceCount ++
17.	S [SpaceCount ] = 10
18.	\u034f:
19.	SpaceCount ++
20.	S [SpaceCount ] = 11
21.	others:
22.	CharCount ++
23.	ifT [i ] is dotted
24.	D [CharCount ] = 1
25.	else
26.	D [CharCount ] = 0
27.	end if
28.	ifT [i ] is not connected
29.	C [CharCount ] = 0
30.	else
31.	C [CharCount ] = 1
32.	end if
33.	DC [CharCount ] =D [CharCount ] & C [CharCount ]
34.	end case
35.	i ++
36.	Read Text[i ]
37.	T [i ] = Unicode(Text[i ])
38.	end while
39.	ifCharCount ! =SpaceCount
	print(“Text may be altered.”)
	else
	forj= 1 to SpaceCount; j ++
40.	S 1S 0=DC [j ] ⊕S [j ]
41.	Insert S 1S0to OutputFile
42.	end for
	CharCount= 0
43.	SpaceCount= 0
44.	i ++
45.	Read Text[i ]
46.	T[i ] = Unicode(Text[i ])
	end if
47.	end while
48.	ifT [i ] = \ufe00     \\odd number of secret bits
49.	Delete last bit from OutputFile
50.	end if

Experimental Results

The proposed method has been implemented using the Python 3 programming language. In order to evaluate our proposed steganography method, performance measurements of steganography techniques should be applied. Steganography techniques are assessed based on four key criteria: capacity, imperceptibility, robustness, and security. In the following, the experimental results of the proposed method in terms of these four measurements are discussed.

Security and imperceptibility

Imperceptibility measures how well the technique preserves the quality and appearance of the cover object, ensuring that any changes are indiscernible to human observers, while security assesses the difficulty of unauthorized parties detecting or extracting the hidden message without knowledge of the embedding method or key, ensuring the confidentiality and integrity of the hidden data (Alotaibi & Elrefaei, 2018). Security and imperceptibility are closely related as the aim of both is to make it difficult to detect the presence of hidden data within the cover text.

For imperceptibility, the proposed method does not add any extra visible characters beyond those originally present in the cover text, thus achieving a very high imperceptibility with 100% imperceptibility ratio.

For security evaluation, the security ratio is computed using the following equation (Al-Nofaie, Gutub & Al-Ghamdi, 2021): (5)${\displaystyle \text{Security ratio}=\frac{\text{Amount after discount}}{\text{Original character}}\times 100}$ where the amount after discount is calculated using the following equations: (6)${\displaystyle \text{Discount value}=\frac{\text{Original characters}\times \text{Excess characters}}{100}}$ (7)${\displaystyle \text{Amount after discount}=\text{Original characters-Discount value}}$

To evaluate the performance of the proposed method, we conducted a number of experiments using the last third Surah of the Holy Quran, Surah ‘Al-Ikhlas’, as the cover text. The results are summarized in Table 10 and illustrated in Fig. 5. In these experiments, different sizes of secret messages were hidden in the cover text, both with and without diacritics. As depicted in the figure, the security of the proposed method increases as the number of secret bits decreases, which shows that there is a tradeoff between capacity and security, as proved in Zhang & Li (2004).

Table 10:

Security ratio of the proposed method.

The bolded values are the capacity and security ratios, which are the important results shown in the tables.

Results without diacritics					Results with diacritics
# of secret bits	Excess characters	Discount value	Amount after discount	Security ratio	# of secret bits	Excess characters	Discount value	Amount after discount	Security ratio
94	50	30.5	30.5	50%	178	92	94.76	8.24	8%
47	25	15.25	45.8	75%	89	46	47.38	55.62	54%
24	13	7.93	53.1	87%	44	23	23.69	79.31	77%
12	7	4.27	56.7	93%	2	12	12.36	90.64	88%

DOI: 10.7717/peerjcs.2236/table-10

Discussion and Comparison

In this section, comparisons of the performance of the proposed method with related work are discussed. The related methods shown in Table 4 are selected, which all use Unicode in their methodologies. The comparison is discussed in the main measurements of steganography: capacity, security, and robustness.

Security and imperceptibility

As shown in Table 4, one of the main drawbacks of the methods based on the use of different Unicode forms is that they suffer from incorrect connectivity between some letters, even though they tried to avoid that. For instance, the example given in Alanazi, Khan & Gutub (2022) shows that the sentence “ ” has changed to “ ” after hiding the secret message. It is clear that the letter “ ” does not connect correctly to the previous letter. Another example is given in Obeidat (2017), where also the letter “ ” does not connect correctly to the previous letter. This degrades the imperceptibility of these methods. Therefore, in this section, these methods, which are M10 to M15, are excluded from the comparison. In addition, the approaches that have very low capacity have also been excluded from the comparison, because the secret message will be very small if these approaches are included. These approaches are the ones that hide secret messages word by word, which are M1 to M3.

To evaluate the performance of the proposed method against the other related approaches in terms of security and imperceptibility, we conducted five experiments using five different secret messages: all ones (Exp1), all zeros (Exp2), alternating ones and zeros (Exp3), random ones and zeros with more ones than zeros (Exp4), and random ones and zeros with more zeros than ones (Exp5). The cover text selected is the last third Surah of the Holy Quran, Surah ‘Al-Ikhlas’. The number of bits in the secret messages is the maximum that could be hidden among all methods, which is 56 bits. The results of these experiments are shown in Table 14 and depicted in Fig. 8.

Table 14:

Security and imperceptibility comparison.

The bolded texts show the average of security and imperceptibility.

	method	M4	M5	M6	M7	M8	M9	Proposed
Exp1	imperceptibility ratio	54%	8%	2%	100%	100%	8%	100%
	Security ratio	44%	44%	-12%	-68%	44%	44%	72%
Exp2	imperceptibility ratio	100%	100%	51%	100%	100%	100%	100%
	Security ratio	100%	100%	44%	100%	100%	100%	72%
Exp3	imperceptibility ratio	77%	54%	28%	100%	100%	54%	100%
	Security ratio	72%	72%	16%	52%	72%	72%	72%
Exp4	imperceptibility ratio	74%	48%	25%	100%	100%	48%	100%
	Security ratio	68%	68%	12%	34%	68%	68%	72%
Exp5	imperceptibility ratio	82%	62%	31%	100%	100%	62%	100%
	Security ratio	77%	77%	21%	44%	77%	77%	72%
average imperceptibility ratio		77%	54%	27%	100%	100%	54%	100%
average security ratio		72%	72%	16%	32%	72%	72%	72%

DOI: 10.7717/peerjcs.2236/table-14

As illustrated in the figure, the proposed method, along with M7 and M8, achieves the highest imperceptibility ratio of 100%.

Regarding security, as discussed in ‘Security and imperceptibility’, the proposed method maintains the same security ratio regardless of the secret bits. In contrast, the security of other approaches varies based on the structure of the secret bits. As illustrated in the figure, the proposed method exhibits a commendable level of security, with a 72% security ratio on average, outperforming M6 and M7, and achieving equal average security with the other approaches, namely, M4, M5, M8, and M9.

Conclusion

This article introduces a novel approach to secure communication during mega events using Arabic text steganography with invisible Unicode characters. In the proposed method each Arabic letter can hide two secret bits based on the two characteristics of that letter: whether it is dotted or not, and whether it is connected or not.

The method offers several advantages: it enables the transmission of small secret messages within Arabic text, making it more compact compared to audio or video covers. Moreover, it’s adaptable to social media platforms, which are prevalent communication channels during mega events. Additionally, the stego text can be sent via email or WhatsApp while preserving the embedded messages’ integrity. Given the widespread use of Arabic in communication, particularly during mega events like Hajj and Umrah, this method holds significant relevance.

The proposed method demonstrates superior performance across key steganography metrics. It surpasses previous methods with an average capacity ratio of 178% because it hides two secret bits in all characters of the cover texts, except for normal spaces, whereas the other approaches hide one bit per character at most, and some of them hide less than that. In addition, the proposed method achieves a perfect imperceptibility ratio of 100% because it does not add any visible characters. Furthermore, it exhibits commendable levels of robustness, as it can detect 70% of possible attacks.

Although the proposed method achieved a good level of security of an average of 72% security ratio, with a comparable performance to other related approaches, this aspect is one of its main limitations because the size of the stego text is larger than the cover text. However, since there is a tradeoff between security and capacity, this limitation is acceptable, especially given the perfect imperceptibility of the proposed method.

As a future direction, steganographic techniques with even higher degrees of security and resilience may be developed by utilizing the remarkable imperceptibility and high capacity of the proposed method.

Supplemental Information