In the present era, the secret to the success of an organization is all about the information it can gather. Also, of concern is how effectively it can stop others from accessing the information generated through its operations and processes. The popularity and availability of the Internet, easy access to digital storage devices, have made the creation, replication, and distribution of digital media hassle-free. This has led to the strong need of developing methods for preventing copyright breaches [1,2].
The technique of digital watermarking is being applied widely to the situations where an organization wants to deter the data from spilling into the public domain. It is extremely crucial in cases where the company is in direct fiduciary relationship with its customers and must protect their information as well [1].
WCNC-2021: Workshop on Computer Networks & Communications, May 01, 2021, Chennai, India. EMAIL: shweta.du@gmail.com (Shweta Wadhera) ORCID: 0000-0002-3123-9976 (Shweta Wadhera) Digital Watermarking is a method to provide protection from any tampering or alteration [3,4,5]. It provides security and authentication to digital content. The digital watermarking process involves the insertion of signal, information into the original media content. The inserted information is then uncovered and extracted to report the actual owner of the digital media.
Watermarking involves embedding data called watermark or digital signature or label into the digital media. This watermark can be extracted for revealing the authenticity of the media object [6]. As an example of a watermark, we can imagine a visible "seal" over an image. A digital watermarking algorithm has three basic parts: -1. Watermark 2. The Encoding algorithm 3. The decoding algorithm [1,7,8]. This technique may be used for copyright safeguarding and embedding fingerprints, placing authentication for data integrity checks, and confidential communication. Another important application of watermarking techniques includes tracing illegal users with an objective that the owner can approach the regulatory authorities. It can be useful for ensuring that data pertaining to people who buy and sell digital media, kept in record for each transaction. Further track of this data can be kept for controlling copyright breaches. In fact, strict measures must be implemented for this unlawful distribution of digital content.
The paper is consolidated as follows: section 2 presents the classification of digital watermarking based on various characteristics. Section 3 describes the prominent features of digital watermarking. Section 4 covers the recent areas of application for watermarking. Section 5 presents the various types of attacks. In section 6 we analyze the highlights and results of related work by various authors and section 7 presents the conclusions and future scope in the direction of image watermarking.
The section describes the classification of digital watermarking based on various criteria such as robustness, perceptibility, domain, and detection process, multimedia. Further we also briefly present the different popular watermarking techniques.
Robust: Robust watermarking is preferred when copyright information is required to be inserted. Robustness is indicated if the embedded watermark even after some attack is not damaged. It can withstand several attacks. It is seen that for copyright protection, a robust watermark is advantageous.
Fragile: One can easily find out from the status of the watermark if the data has been altered. For integrity protection, this watermark is preferred.
Semi-fragile: Some extent of change is tolerated by a semi-fragile watermark [2,3,6].
Perceptible: A watermark that is visible is known as perceptible.
Imperceptible: Incase the watermark is invisible then it is known as imperceptible watermark. In this, information which is embedded into the image is not visible. In such cases one can prove the ownership of your image with the help of imperceptible watermark [2,26].
Frequency domain: First, the transforming of the image to the frequency domain is carried out. In this type of watermarking different transformation techniques are applied such as DCT, DFT, and DWT [27].
Spatial Domain: Watermarking in this domain moderately changes the value of pixels, of arbitrarily selected portions of images; the watermark is inserted in the host image. No conversion or transformation is applied in the spatial domain. LSB, Patchwork method, SSM Modulation are some of the popular spatial domain-based techniques [2,3].
Generally, the watermarking carried out in the frequency domain is more robust as compared to the one carried out in the spatial domain.
In this method, the watermark is inserted in the least significant bit (LSB of image pixels. Ideally, either of two ways is used for embedding. In one approach, the LSB of an image is substituted with a pseudo-noise (PN) sequence, while another approach adds this PN sequence to the LSB. LSB technique provides ease of use but compromises on the robustness parameter against attacks.
Patchwork method randomly picks n pairs of image points (x, y). The data in the x region is lightened and data in the y region is darkened. However, this technique can withstand a series of attacks, but it lacks in terms of capacity. This method employs a technique of subdividing an image signal into non-overlapping blocks of 8 × 8 size. Further, block-wise DCT is performed, and thereafter the choice of coefficients to be watermarked is carried out. Finally, an inverse DCT is applied on each 8 × 8 block to obtain the signed image.
In this method, the image is subjected to a sequence of low-pass and high-pass filters. An image is decomposed into four equal sub-bands where each sub-band comprises low frequency (LL), horizontal features (LH), vertical features (HL), and diagonal features (HH). It is a preferred algorithm as it provides a robust and secure watermarking method [2,3,27].
Blind: Those watermarking processes fall in this category wherein the removal of the embedded information requires only the watermarked image. Its applications can be copyright protection, evoting, etc. [9].
Non-Blind: In this type of watermarking, the process copies along with the text data, the host image, and the inserted information for the retrieval of the watermark. Its application is seen in copyright protection [9].
Text Watermarking: It consists of components like words, punctuations, sentences, etc.
Transformation is done on one of these components and is embedded as a watermark [3].
Image Watermarking: Large-size images are there which has to be watermarked. In the case of images, we require robust watermarks, which should be imperceptible [3].
Video watermarking: In this case, it is difficult to get imperceptible watermark.
Graphic Watermarking: In 2D or 3D digital graphics, a watermark is embedded. It provides copyright protection.
Mentioned here are various features of Digital Watermarking.
Robustness: The robustness feature indicates that the digital watermark can resist various processing operations and attacks. Then it is considered to be robust [14].
Imperceptibility: The imperceptibility feature indicates that a digital watermark should not be seen by the human eye. One should not be able to see the embedded watermark. It can only be identified by specialized procedures. The watermark should be such that the viewer should not be able to see it and the process of embedding a watermark should be such that the quality of the content is maintained [14].
Security: The security feature indicates that irrespective of targeted attacks, the inserted digital watermark cannot be removed. Watermark security describes that altering or removing a watermark without any deterioration to the host signal should be arduous. Watermarking security can be explained as a way to provide secrecy, ownership, and protection of data [2].
Capacity: The amount of information embedded in a watermarked image also known as data payload [3].
Copyright protection: As we are aware that images can be easily circulated and are freely available over the internet. These images can be used commercially. So copyright protection of data is needed and for this Digital Watermarking is very useful. The inserted digital watermark will be used to identify the copyright owner [2,9].
Fingerprint: Fingerprinting in digital watermarking can be put up as a method for embedding some distinctiveness. The fingerprint should be difficult to alter. The information inserted is related to the customer. It is through this fingerprinting that it is revealed about those authorized customers who are involved in the circulation of copyright data by breaching the agreement [2,27]. Broadcast Monitoring: Over the years it has been seen that availability and accessibility to media content have increased exponentially. Also, the content is available through the internet. In such times, it has become important for content owners and copyright owners to know about the real distributor of content. Digital Watermarking has an important role here [3,26].
Medical Application: Embedding of patient name in the MRI-scan, CT-scan or X-ray reports can be done using visible watermarking. The treatment of the patient depends on these medical reports. So to avoid mixing of reports the technique of visible watermarking can be used [9].
Electronic Voting System: The Internet has spread all over the country from big towns to small villages. Electronic voting helps to carry out elections, keeping the security aspect into consideration [9].
Remote Education: Lack of teachers poses a big problem in small villages. Smart Technology needs to be adopted for distance learning. In this case watermarking plays its role in the authentic transmission of study material over the internet [9,27].
A digital Watermarking scheme is always assessed by the fact that how robust it is over attacks. Attack on any watermark is used to harm the inserted watermark or enfeeble the watermark's discovery. Hampering the protection, which a watermark provides to digital content, is the objective of any attack. Watermarking attacks can be classified as follows -Geometry attack, Protocol attack, Cryptographic attack, and removal attack [10,11,12].
Geometry Attack: Such processing is done over the watermark image which alters the geometry of the image like rotation, cropping, etc. These can be further classified into -scaling, cropping, rotation, and translation [6]. Removal Attack: This attack aims at removing the inserted data from the digital image. If it is not able to, yet they try to destroy the embedded information [29]. Protocol Attack: The attacks which come under this category, do not damage the embedded data. Two types of Protocol attacks are there: Invertible, Copy attack. So a watermark should be noninvertible and should not be copied. A watermark is invertible when the attacker removes his own watermark from the host data. The attacker then pretends to be the owner of the data. This shows that for copyright protection, watermarks should be non-invertible [13,29].
Copy Attack: It is also a form of Protocol attack. In this also the watermark is not destroyed. Instead, the attacker estimates the watermark from host data. It is then copied to other data [13].
Cryptographic Attack: These types of attacks include those which break the security in watermarking techniques. With this, they can extract the inserted watermark data or can insert some delusive watermark. The Brute-force and Oracle attacks fall under this category [29].
In order to present an exhaustive survey on image watermarking, we created a repository of more than 50 papers on Mendeley and 35 papers were used to develop the background of digital watermarking and 19 papers were found relevant in the area of image watermarking.
Table 1 summarizes the watermarking schemes proposed by various research groups in the past few years through the comparative analysis. 1, a study and evaluation of work done in digital watermarking techniques in past are enumerated. The spatial domain and frequency domain techniques are some popular techniques examined in the past. Also, it has been observed that the spatial domain digital watermarking technique is less robust and hence less preferred. The performance of the watermarked image is evaluated through robustness, imperceptibility, security, and capacity. Among these the most preferred criteria, were the visual imperceptibility of the watermarked image and the robustness of the watermarking. In fact, the future work holds scope by combining techniques and using them in hybrid form to not only enhances the robustness of the watermarked image, but it may also reduce the drawbacks of each method considered separately.
This paper gives an overview of techniques of digital image watermarking along with the detailed classification and characteristics. The various application areas such as medical, remote education, military, electronic voting systems have been presented. It has been seen that data security has become a top priority due to the extensive transmission of digital data. So, for providing authorized data or safeguarding important data, digital watermarking is used. The performance of the watermarked images is evaluated through robustness, imperceptibility, security, and capacity. These are analyzed using peak signal to noise ratio and bit-error ratio. It was observed that robustness was a preferred criterion. Invisible watermarking is carried out for content authentication and proof of ownership. Research groups have preferred frequency domain techniques and have tried to work on balancing between robustness and visual imperceptibility. Through the paper, we analyze various watermarking methods in digital images used in the recent past.