Text Watermarking Research Articles

Scalable watermarking for identifying large language model outputs

Large language models (LLMs) have enabled the generation of high-quality synthetic text, often indistinguishable from human-written content, at a scale that can markedly affect the nature of the information ecosystem1–3. Watermarking can help identify synthetic text and limit accidental or deliberate misuse4, but has not been adopted in production systems owing to stringent quality, detectability and computational efficiency requirements. Here we describe SynthID-Text, a production-ready text watermarking scheme that preserves text quality and enables high detection accuracy, with minimal latency overhead. SynthID-Text does not affect LLM training and modifies only the sampling procedure; watermark detection is computationally efficient, without using the underlying LLM. To enable watermarking at scale, we develop an algorithm integrating watermarking with speculative sampling, an efficiency technique frequently used in production systems5. Evaluations across multiple LLMs empirically show that SynthID-Text provides improved detectability over comparable methods, and standard benchmarks and human side-by-side ratings indicate no change in LLM capabilities. To demonstrate the feasibility of watermarking in large-scale-production systems, we conducted a live experiment that assessed feedback from nearly 20 million Gemini6 responses, again confirming the preservation of text quality. We hope that the availability of SynthID-Text7 will facilitate further development of watermarking and responsible use of LLM systems.

A Survey of Text Watermarking in the Era of Large Language Models

Text watermarking algorithms are crucial for protecting the copyright of textual content. Historically, their capabilities and application scenarios were limited. However, recent advancements in large language models (LLMs) have revolutionized these techniques. LLMs not only enhance text watermarking algorithms with their advanced abilities but also create a need for employing these algorithms to protect their own copyrights or prevent potential misuse. This paper conducts a comprehensive survey of the current state of text watermarking technology, covering four main aspects: (1) an overview and comparison of different text watermarking techniques; (2) evaluation methods for text watermarking algorithms, including their detectability, impact on text or LLM quality, robustness under target or untargeted attacks; (3) potential application scenarios for text watermarking technology; (4) current challenges and future directions for text watermarking. This survey aims to provide researchers with a thorough understanding of text watermarking technology in the era of LLM, thereby promoting its further advancement.

Emerging Arabic Text Watermarking Utilizing Combinations of Different Diacritics

Emerging Arabic Text Watermarking Utilizing Combinations of Different Diacritics

Fortifying Textual Integrity: Evolutionary Optimization-powered Watermarking for Tampering Attack Detection in Digital Documents

Digital document helps as the lifeblood of present communication, yet their vulnerability to tampering poses major safety anxieties. Digital text watermarking is an effective mechanism to protect the reliability of text-based data in the digital. Introducing a hidden layer of accountability and safety, allows individuals and organizations to trust the written word and make sure the truth behind all the files. Watermarking model identifies the tampering attack by inspecting the embedded signature for distortions or alterations. Watermarks can able to mechanically classify and repair themselves once tampered with, improving document resilience. Watermarking acts as a powerful tool to detect tampering attacks in digital document. By embedding strong and imperceptible watermarks in document distribution or creation, alterations are recognized by specialized procedure. This study introduces an Evolutionary Optimizer-powered Watermarking for Tampering Attack Detection in Digital Document (EO-WTAD3) model. The main intention of EO-WTAD3 approach is to support textual integrity using the applications of metaheuristic optimizer algorithm based watermarking technique for detecting tampering attacks in digital document. In the EO-WTAD3 method, a digital watermarking method has been proposed for the ownership verification and document copyright protection using data mining concept. Moreover, the EO-WTAD3 technique utilizes the concepts of data mining to define appropriate characteristics from the document for embedding watermarks. Moreover, fractional gorilla troops optimization (FGTO) algorithm can be applied for the assortment of optimal situation of watermarks in the content, ensuring both imperceptibility and strong to tamper. The performance validation of the EO-WTAD3 methodology takes place employing multiple datasets. The extensive result analysis portrayed that the EO-WTAD3 system accomplishes improve solution with other existing approaches with respect distinct aspects.

UniHaCh: Unicode and Hash Function Supported with Counting and Frequency Recurrence of Arabic Characters for Quranic Text Watermarking

UniHaCh: Unicode and Hash Function Supported with Counting and Frequency Recurrence of Arabic Characters for Quranic Text Watermarking

Improving ANiTW Performance Using Bigrams Character Encoding and Identity-Based Signature

Several threats, such as misinformation and hoax, exist in social media as a frequently used technology to exchange information. These threats can be prevented if the reader can verify the message’s content integrity and source authenticity. One recent solution to overcome these threats was proposed by Ahvanooey et al. called ANiTW. ANiTW uses an invisible text watermarking scheme to verify social media message integrity and identify the message source. However, ANiTW is still vulnerable to undetected malicious text modification. Moreover, ANiTW has no tools to authenticate the extracted message source. The proposed method aims to overcome those vulnerabilities and improve ANiTW performance as a tool to verify message content integrity and message source authenticity in social media. To overcome vulnerabilities in ANITW, a part of the message hash and IBS-based signature is introduced as the watermark. The message integrity is verifiable by using part of the message hash as a watermark to detect any malicious. Moreover, implementing an IBS-based signature enables any reader to verify the message source’s authenticity easily. In addition, to improve ANITW performance in embedding capacity and invisibility, a novel bigram-based character encoding and embedding rule are also introduced. Based on the experiment and analysis result, it is proven that the proposed method can perform better than ANiTW in the security, embedding capacity, and invisibility aspects.

Arabic text watermarking tuned for medical e-record semi-authentication

This study focuses on Arabic text watermarking semi-verification utilizing recent counting-based secret-sharing to partially validate the ownership and correctness for all sensitive medical e-records. The benefit of this approach is its semi-trust of e-records, as needed services provided, even if complete text medical report is a bit delayed for full verification. The Proposed work is presents two approaches of text-watermarking which are Seed Shares approach and WM Shares approach. The capacity results of our approaches is achieved without effect of secret bits hidden. The security is achieved via using XOR operation and CBSS technique. The comparisons result shows that the Accuracy percentage of the seed shares approach is much better than the WM in terms of tampering attack. The exploration tested its proposed strategy on standard Arabic benchmark of 42 Nawawi Hadiths which provided promising remarks, which can also be useful for Urdu and Farsi texting.

A three-dimensional text image watermarking model based on multilayer overlapping of extracted two-dimensional information

Aiming at the problem that watermark information is easy to be lost in two-dimensional text image watermarking, a three-dimensional text image watermarking model is proposed based on multilayer overlapping of extracted two-dimensional information, and a specific method is accordingly realized by means of embedding, extracting and overlapping of multiple watermarks in sequence. We discuss the influence of the sequence order, positional distancing and color difference of the extracted multiple two-dimensional information on displaying a three-dimensional text watermark image. In addition, considering the crucial evaluation indices of the invisibility and robustness for the watermarking algorithm, the selection method for superior embedding regions of multiple watermarks is also constructed to enhance the robust performance of watermarks under the premise of effective invisibility. Moreover, we embed the multiple two-dimensional information into the host image by using the undecimated dual-tree complex wavelet transforms - bidiagonal singular value decomposition algorithm. In this way, the extracted multiple two-dimensional information is automatically overlapped to generate the three-dimensional text watermarking according to the optimal matching parameters. We use standard image processing data set to carry out experiments, the results show that the peak signal to noise ratio before and after embedded watermarks is higher than 39 dB, and the peak signal to noise ratio between the original watermarks and the extracted watermarks is more than 37 dB, which is superior to the current reported watermark model. Therefore, it is suggested that the proposed model shows good invisibility and robustness and can reduce the loss of watermark during transmission and attack as much as possible.

Watermarking Using Energy-LSB Embedded Method

Digital watermarking technology is increasingly used to protect copyright and demonstrate ownership of digital multimedia (such as text, music, photos, and videos). In order to safeguard intellectual property rights and rules of ownership for multimedia, this project suggests a text watermark algorithm. The process of hiding little text or grayscale images is the main focus, though. The masking of a watermark text in a high-color or high-density area of the block picture has been proposed using an embedding technique based on an energy function and the Least Significant Bits (LSB) method. Even with various format types and picture sizes chosen to conceal and cover a changing message size, the quality results demonstrate that the watermark image suffers from less distortion than the cover image, and the suggested algorithm is powerful to conceal a random watermark text even with smaller block sizes. An additional optional choice to encrypt the text watermark before embedding is also recommended because doing so would make it harder for hackers to read the text. This text can be encrypted using the Caesar cipher method before embedding is implemented in images.The experimental results of using the suggested algorithm for embedding and extracting watermark text for various sizes in a large number of images were satisfactory, giving a level of peak signal-to-noise ratio (PSNR) and Signal to Noise Ratio (SNR) with low mean square error (MSE) values. However, PSNR degrades more quickly than LBS as the watermark text size increases, so it was determined that it is more suitable for applying a watermark rather than a stego It is employed in order to share information securely

Language universal font watermarking with multiple cross-media robustness

The rapid development of digital devices and the increasing demand for copyright protection and leakage tracing of documents result in new demands for robust text watermarking. It becomes more important for a utility text watermarking scheme to have robustness in cross-media channels since the camera shooting operation to papers or screens has become very convenient. To achieve the corresponding cross-media robustness, this paper proposes a font-based text watermarking scheme, which is applicable to most commonly used languages while the previous works were usually designed for specific languages. To generate fonts with high efficiency, we propose a novel glyph centroid modification-based font generation algorithm, which can automatically create target-similar fonts. And we design an effective watermarking scheme that utilizes the relative centroid position (RCP) of glyphs to represent watermark signals. Compared with existing font-based text watermarking schemes that artificially design fonts, the proposed scheme ensures higher efficiency by generating the font codebook automatically. In addition, the proposed RCP-based watermarking scheme can achieve the robustness against lossy compression and several kinds of cross-media distortions. The watermark message can be extracted without the need to recognize the semantic information of glyphs by optical character recognition (OCR).

Tracing Text Provenance via Context-Aware Lexical Substitution

Text content created by humans or language models is often stolen or misused by adversaries. Tracing text provenance can help claim the ownership of text content or identify the malicious users who distribute misleading content like machine-generated fake news. There have been some attempts to achieve this, mainly based on watermarking techniques. Specifically, traditional text watermarking methods embed watermarks by slightly altering text format like line spacing and font, which, however, are fragile to cross-media transmissions like OCR. Considering this, natural language watermarking methods represent watermarks by replacing words in original sentences with synonyms from handcrafted lexical resources (e.g., WordNet), but they do not consider the substitution’s impact on the overall sentence's meaning. Recently, a transformer-based network was proposed to embed watermarks by modifying the unobtrusive words (e.g., function words), which also impair the sentence's logical and semantic coherence. Besides, one well-trained network fails on other different types of text content. To address the limitations mentioned above, we propose a natural language watermarking scheme based on context-aware lexical substitution (LS). Specifically, we employ BERT to suggest LS candidates by inferring the semantic relatedness between the candidates and the original sentence. Based on this, a selection strategy in terms of synchronicity and substitutability is further designed to test whether a word is exactly suitable for carrying the watermark signal. Extensive experiments demonstrate that, under both objective and subjective metrics, our watermarking scheme can well preserve the semantic integrity of original sentences and has a better transferability than existing methods. Besides, the proposed LS approach outperforms the state-of-the-art approach on the Stanford Word Substitution Benchmark.

Fusion of Watermarking and Steganography for Protecting Image Ownership

The use of digital images has become very common because of the rapid increase of the internet over time. Moving digital images over the internet is easy, but keeping ownership is complex, and serious issues have emerged. Forgery, fraud, and pirating of this content are rising. Different techniques used to protect images, like watermarking and steganography, but these methods are not enough toprotect. So, providing new techniques is essential for protecting image ownership. We have proposed a fusion method of steganography and watermarking in this work. First, the secret message is encoded within the original image using the LSB technique to obtain the stego image. Secondly, the watermarking process is applied on the stego image using text watermarking or image watermarking to provide stego-watermarked-image. The proposed fusion watermarking and steganography method is very useful for protecting image ownership over insecure communication channels. An attacker cannot get the desired watermarked image from the stego-watermarked-image without knowing the secret message hiding inside it using the LSB technique. The proposed method is efficient, simple and secure; it provides significant protection for image ownership.

Securing Arabic Contents Algorithm for Smart Detecting of Illegal Tampering Attacks

The most common digital media exchanged via the Internet is in text form. The Arabic language is considered one of the most sensitive languages of content modification due to the presence of diacritics that can cause a change in the meaning. In this paper, an intelligent scheme is proposed for improving the reliability and security of the text exchanged via the Internet. The core mechanism of the proposed scheme depends on integrating the hidden Markov model and zero text watermarking techniques. The watermark key will be generated by utilizing the extracted features of the text analysis process using the third order and word level of the Markov model. The Embedding and detection processes of the proposed scheme will be performed logically without the effect of the original text. The proposed scheme is implemented using PHP with VS code IDE. The simulation results, using varying sizes of standard datasets, show that the proposed scheme can obtain high reliability and provide better accuracy of the common illegal tampering attacks. Comparison results with other baseline techniques show the added value of the proposed scheme.

An Optimal Text Watermarking Method for Sensitive Detecting of Illegal Tampering Attacks

Due to the rapid increase in the exchange of text information via internet networks, the security and authenticity of digital content have become a major research issue. The main challenges faced by researchers are how to hide the information within the text to use it later for authentication and attacks tampering detection without effects on the meaning and size of the given digital text. In this paper, an efficient text-based watermarking method has been proposed for detecting the illegal tampering attacks on the Arabic text transmitted online via an Internet network. Towards this purpose, the accuracy of tampering detection and watermark robustness has been improved of the proposed method as compared with the existing approaches. In the proposed method, both embedding and extracting of the watermark are logically implemented, which causes no change in the digital text. This is achieved by using the third level and alphanumeric strategy of the Markov model as a text analysis technique for analyzing the Arabic contents to obtain its features which are considered as the digital watermark. This digital watermark will be used later to detecting any tampering of illegal attack on the received Arabic text. An extensive set of experiments using four data sets of varying lengths proves the effectiveness of our approach in terms of detection accuracy, robustness, and effectiveness under multiple random locations of the common tampering attacks.

POTENTIAL APPLICATIONS OF LINGUISTIC STEGANOGRAPHIC TRIGGER-CONTAINERS FOR TEXT WATERMARKING PURPOSES

The paper gives analysis of feasibility of using linguistic steganographic trigger-containers as means of linguistic-based text watermarking. The proposed approah is based on the previous experimental research in the context of Russian native speaker text juncture perception. It was posited that specific minimal text modifications addressed in the paper may be used as means of text watermarking with the aim of tracking the leak of information for the purposes of taking legal actions, enforcing non-disclosure agreements, and testing for internal vulnerabilities. There is analyzed the viability of altering paragraph juncture points in Russian texts and, through the use of the corresponding trigger-containers, usage of this alteration as a means of linguistic watermarking.

Text Steganography in Image depending on Radon _ Barker code Transforms

- With the widespread usage of the internet around the world, both public and private data are now widely disseminated on the internet. The security of these online records is a pressing issue that must be addressed immediately. Text watermarking is a technique for adding watermarks to text documents to help preserve their authenticity and credibility. For many years, text watermarking has been a hot topic of research. This paper describes a new method for embedding text in a picture that uses a radon transform-based Barker code to make the device more reliable and ensure that the text is returned without error. The main advantages of this research are the high security and the big capacity of data to be hiding since the use of radon transform is very helpful in extracting the data. All the image noise does not effect on the stego-image except the paper and salt noise effect on Root Mean Square (RMS) since its value is 0.274 and contract enhancement effect on RMS and Mean Square Error (MSE) by the values 0.586 and 0.343 respectively. Also the SSIM value is very acceptable since there is an obvious similarity between the original image and the stego-image (after stegonagraphy process).

A Hybrid Intelligent Text Watermarking and Natural Language Processing Approach for Transferring and Receiving an Authentic English Text Via Internet

Abstract Due to the rapid increase in the exchange of text information via internet networks, the security and the reliability of digital content have become a major research issue. The main challenges faced by researchers are authentication, integrity verification, and tampering detection of the digital contents. In this paper, a Robust English Text Watermarking and Natural Language Processing Approach (RETWNLPA) is proposed based on word mechanism and first level order of Markov model to improve the accuracy of tampering detection of sensitive English text. The RETWNLPA approach embeds and detects the watermark logically without altering the original text document. Based on the hidden Markov model (HMM), the first-level order of word mechanism is used to analyze the interrelationship between English text. The extracted features are used as watermark information and integrated with text zero-watermarking techniques. To detect eventual tampering, RETWNLPA has been implemented and validated with attacked English text. Experiments were performed on four datasets of varying sizes under random locations of common tampering attacks. The simulation results prove the tampering detection accuracy of our method against all kinds of tampering attacks. Comparison results show that RETWNLPA outperforms baseline approaches HNLPZWA (an intelligent hybrid of natural language processing and zero-watermarking approach) and ZWAFWMMM (Zero-Watermarking Approach based on Fourth level order of Word Mechanism of Markov Model) in terms of tampering detection accuracy.

Data Aggregation and Privacy Preserving Using Computational Intelligence

In today's smart world, the privacy protection of data is an important issue. Data is distributed, reproduced, and disclosed with extensive use of communication technologies. Many non-traditional challenges arise with the rapid increase of IoT devices for system design and implementation. However, security and privacy are the main issues in IoT. With advanced technologies, an illegal copy of the content can easily be generated and shared. Therefore, it is crucial for users to protect and secure their data. In the said perspective, an efficient third-generation watermarking technique is proposed, which works on the computational intelligence model to insert a large amount of robust watermark and make an extra effort to hide more information than first and second-generation techniques. The Advanced Encryption Standard (AES) encryption algorithm is employed to guarantee secure communication, which has a significantly less computational cost. The proposed technique evaluated parameters including security, robustness, imperceptibility, and capacity. The results of the proposed technique are compared to existing text watermarking methods, which illustrates it is secure, robust, imperceptible, and inserts a large amount of watermark information through computational intelligence.

Advanced security and privacy technique for digital text in smart grid communications

In the current era of technology, digital contents copyright protection is a significant issue in the Internet of Things and smart grid. Digital contents are distributed, reproduced, and disclosed with extensive use of communication technologies such as the cloud. It has become quite easy to produce an illegal copy of digital content. Limited techniques are available for text documents ownership verification and copyright protection in the Internet of Things. However, most of the existing techniques produce distortion during watermark insertion, which directly impacts imperceptibility. We propose a robust and secure digital text watermarking technique for Microsoft Word documents in the said perspective. The watermark is extracted with high probability, which illustrates that the proposed technique is robust, imperceptible, and improves the capacity. The results of the proposed technique are compared with the previous techniques of text watermarking. The proposed technique is imperceptible with the Peak signal-to-noise ratio is 33.65, and the similarity percentage is 99.42%. The length of the secret message is improved from 0.2 to 1.24 kilobytes.

Secure and robust color image dual watermarking based on LWT-DCT-SVD

Now days advancements in multimedia technology and Information and Communication technologies (ICTs) has raised various security related concerns. Digital image watermarking is one of the new and popular techniques for the protection of multimedia content. This paper proposes an approach for digital image watermarking in which Lifting wavelet transform (LWT), Discrete Cosine Transform (DCT) and Singular Value Decomposition (SVD) have been used. The host color image is decomposed using LWT and LH3, LL3 sub bands of the third level LWT are transformed by DCT and then SVD. Security of logo (image watermark) is enhanced by using message-digest (MD5) hash algorithm and then DCT-SVD are applied to it before embedding it into LH3 sub band. The robustness of Info (Text watermark) is increased by encoding it using Hamming error correcting code. Error correcting code increases the length of watermark therefore arithmetic coding is applied that provides lossless compression. Experimental results are computed for different color image models (RGB, YIQ, YCbCr) at different gain factors, size of text watermark, different types of cover image and for attacks. The proposed method has good performance for the imperceptibility of watermarked image and robustness of watermarks. The performance of this algorithm can be improved by using various optimization techniques.