Hanzhou Wu

Shanghai University | SHU

Linguistic steganography (LS) aims to embed secret information into a highly encoded text for covert communication. It can be roughly divided to two main categories, i.e., modification based LS (MLS) and generation based LS (GLS). MLS embeds secret data by slightly modifying a given text without impairing the meaning of the text, whereas GLS uses a...

Watermarking by Zernike moments has been proven to be effective in providing high rotational resistance. However, due to the high computational complexity, the conventional video watermarking methods using Zernike moments are developed for videos with low resolution. Moreover, according to the properties of Zernike moments, only the matrices of equ...

Recent studies show that scaling pre-trained language models can lead to a significantly improved model capacity on downstream tasks, resulting in a new research direction called large language models (LLMs). A remarkable application of LLMs is ChatGPT, which is a powerful large language model capable of generating human-like text based on context...

Recently, more and more attention has been focused on the intellectual property protection of deep neural networks (DNNs), promoting DNN watermarking to become a hot research topic. Compared with embedding watermarks directly into DNN parameters, inserting trigger-set watermarks enables us to verify the ownership without knowing the internal detail...

Recent years have witnessed the prosperous development of Graph Self-supervised Learning (GSSL), which enables to pre-train transferable foundation graph encoders. However, the easy-to-plug-in nature of such encoders makes them vulnerable to copyright infringement. To address this issue, we develop a novel watermarking framework to protect graph en...

It has been established that convolutional neural networks are susceptible to elaborate tiny universal adversarial perturbations (UAPs) in natural image classification tasks. However, UAP attacks against face recognition systems have not been fully explored. This paper proposes a spatial perturbation method that generates UAPs with local stealthine...

Copy-move forgery is a common audio tampering technique in which users copy the contents of one speech and paste them into another region of the same speech signal, thus achieving the effect of tampering with the semantics. To verify the authenticity of the audio, this paper proposes a method to detect and localize audio copy-move forgery based on...

Mainstream transferable adversarial attacks tend to introduce noticeable artifacts into the generated adversarial examples, which will impair the invisibility of adversarial perturbation and make these attacks less practical in real-world scenarios. To deal with this problem, in this paper, we propose a novel black-box adversarial attack method tha...

The objective of linguistic steganography is to embed additional data in text carriers for covert communication, whereas linguistic steganalysis, as a counter technology to linguistic steganography, aims at revealing the existence of additional data within unknown texts. Early linguistic steganography algorithms alter a text carrier to embed additi...

With the extensive adoption of generative models across various domains, the protection of copyright for these models has become increasingly vital. Some researchers suggest embedding watermarks in the images generated by these models as a means of preserving IP rights. In this paper, we find that existing generative model watermarking introduces h...

As generative models find broader applications in Internet of Things (IoT) image processing tasks, safeguarding the copyright of these models assumes increasing significance. Embedding watermarks on the output images generated by such models has been proposed by some researchers as a means of protecting intellectual property. However, prevailing me...

The susceptibility of Graph Neural Networks (GNNs) to backdoor attacks poses a significant potential threat to GNN-based Internet of Things (IoT) systems. In such attacks, GNNs are manipulated to behave abnormally when presented with maliciously crafted samples known as trigger samples, which can be exploited by attackers for their malicious intent...

Multi-view subspace clustering aims to cluster the data lying in a union of subspaces with low dimensions. The commonly used spectral clustering performs the final clustering based on an n×n affinity graph, which suffers from relative high time and space complexity. Some existing works have chosen key anchors with uniform sampling strategy or K -...

Multi-view clustering has gained great progress recently, which employs the representations from different views for improving the final performance. In this paper, we focus on the problem of multi-view clustering based on the Markov chain by considering low-rank constraints. Since most existing methods fail to simultaneously characterize the relat...

Multi-view clustering has become an important research topic in machine learning and computer vision communities, which aims at achieving a consensus partition of data points across different views. However, the existing multi-view clustering methods fail to simultaneously consider the pairwise and high-order correlations among different views in t...

Protecting deep neural networks (DNNs) against intellectual property (IP) infringement has attracted an increasing attention in recent years. Recent advances focus on IP protection of generative models, which embed the watermark information into the image generated by the model to be protected. Although the generated marked image has good visual qu...

The low-hit-zone (LHZ) frequency hopping sequence (FHS) sets are widely applicable in quasi-synchronous frequency hopping multiple-access (QS-FHMA) systems. In order to reduce mutual interference (MI) in the zone around the signal origin between different users, we recommend the LHZ FHS set instead of the conventional FHS set. In this letter, we pr...

Intellectual property protection of deep neural networks is receiving attention from more and more researchers, and the latest research applies model watermarking to generative models for image processing. However, the existing watermarking methods designed for generative models do not take into account the effects of different channels of sample i...

As a self-supervised learning paradigm, contrastive learning has been widely used to pre-train a powerful encoder as an effective feature extractor for various downstream tasks. This process requires numerous unlabeled training data and computational resources, which makes the pre-trained encoder become the valuable intellectual property of the own...

The purpose of image steganalysis is to determine whether the carrier image contains hidden information or not. Since JEPG is the most commonly used image format over social networks, steganalysis in JPEG images is also the most urgently needed to be explored. However, in order to detect whether secret information is hidden within JEPG images, the...

In frequency hopping communication, time delay and Doppler shift incur interference. With the escalating upgrading of complicated interference, in this paper, the time-frequency two-dimensional (TFTD) partial Hamming correlation (PHC) properties of wide-gap frequency-hopping sequences (WGFHSs) with frequency shift are discussed. A bound on the maxi...

In quasi-synchronous FH multiple-access (QS-FHMA) systems, no-hit-zone frequency-hopping sequences (NHZ-FHSs) can offer interference-free FHMA performance. But, outside the no-hit-zone (NHZ), the Hamming correlation of traditional NHZ-FHZs maybe so large that the performance becomes not good. And in high-speed mobile environment, Doppler shift phen...

Intellectual property protection of deep neural networks is receiving attention from more and more researchers, and the latest research applies model watermarking to generative models for image processing. However, the existing watermarking methods designed for generative models do not take into account the effects of different channels of sample i...

Reversible Data Hiding in Encrypted Domain (RDHED) is an innovative method that can keep cover information secret and allows the data hider to insert additional information into it. This article presents a novel data hiding technique in an encrypted text called Reversible Data Hiding in Encrypted Text (RDHET). Initially, the original text is conver...

Deep neural networks (DNNs) have already achieved great success in a lot of application areas and brought profound changes to our society. However, it also raises new security problems, among which how to protect the intellectual property (IP) of DNNs against infringement is one of the most important yet very challenging topics. To deal with this p...

As a self-supervised learning paradigm, contrastive learning has been widely used to pre-train a powerful encoder as an effective feature extractor for various downstream tasks. This process requires numerous unlabeled training data and computational resources, which makes the pre-trained encoder become valuable intellectual property of the owner....

Recently, more and more attention has been focused on the intellectual property protection of deep neural networks (DNNs), promoting DNN watermarking to become a hot research topic. Compared with embedding watermarks directly into DNN parameters, inserting trigger-set watermarks enables us to verify the ownership without knowing the internal detail...

To achieve a good trade-off between the data-embedding payload and the data-embedding distortion, mainstream reversible data hiding (RDH) algorithms perform data embedding on a well-built prediction error histogram. This requires us to design a good predictor to determine the prediction errors of cover elements and find a good strategy to construct...

Anomaly detection of surveillance video has become a critical concern in computer vision. It can be used for real-time monitoring and the timely generation of alarms and is widely applied in transportation systems and security systems. An unsupervised anomaly detection method for surveillance video based on frame prediction is implemented in this p...

Digital video watermarking has become a hot research topic in recent years due to the increasing demand of protecting the intellectual property of video data. Even though many conventional video watermarking methods have been reported in past years, few of them are resistant to high-intensity geometric attacks, which motivates the authors in this p...

With the fast development of natural language processing, recent advances in information hiding focus on covertly embedding secret information into texts. These algorithms either modify a given cover text or directly generate a text containing secret information, which, however, are not reversible, meaning that the original text not carrying secret...

Linguistic steganography (LS) conceals the presence of communication by embedding secret information into a text. How to generate a high-quality text carrying secret information is a key problem. With the widespread application of deep learning in natural language processing, recent algorithms use a language model (LM) to generate the steganographi...

With the widespread use of deep neural networks (DNNs) in many areas, more and more studies focus on protecting DNN models from intellectual property (IP) infringement. Many existing methods apply digital watermarking to protect the DNN models. The majority of them either embed a watermark directly into the internal network structure/parameters or...

This letter is concerned with tackling both of the multi-access interference (MAI) and the follower jamming (FJ) in frequency-hopping multiple-access (FHMA) systems via novel FH sequence (FHS) design. We consider low-hit-zone (LHZ) FHS which provides lower MAIs in quasi-synchronous FHMA systems but suffers from FJ attack. We introduce an improved L...

Subspace clustering aims to fit each category of data points by learning an underlying subspace and then conduct clustering according to the learned subspace. Ideally, the learned subspace is expected to be block diagonal such that the similarities between clusters are zeros. In this paper, we provide the explicit theoretical connection between spe...

Linguistic steganography (LS) conceals the presence of communication by embedding secret information into a text. How to generate a high-quality text carrying secret information is a key problem. With the widespread application of deep learning in natural language processing, recent algorithms use a language model (LM) to generate the steganographi...

Benefiting from the rapid development of computer hardware and big data, deep neural networks (DNNs) have been widely applied in commercial speaker recognition systems, achieving a kind of symmetry between “machine-learning-as-a-service” providers and consumers. However, this symmetry is threatened by attackers whose goal is to illegally steal and...

Linguistic steganography (LS) aims to embed secret information into a highly encoded text for covert communication. It can be roughly divided to two main categories, i.e., modification based LS (MLS) and generation based LS (GLS). Unlike MLS that hides secret data by slightly modifying a given text without impairing the meaning of the text, GLS use...

Embedding multiple watermarks into a digital object enables multiple purposes to be realized. In this paper, we present a multi-party watermark embedding framework based on frequency-hopping sequences (FHSs). In the proposed work, a certain number of FHSs are generated in advance and then randomly assigned to multiple users. Each user uses an assig...

Frequency-hopping sequence (FHS) sets with low-hit-zone (LZH) can be well applied in quasi-synchronous (QS) frequency-hopping multiple-access (FHMA) systems to reduce the mutual interference among different users. On the other hand, LHZ-FHS sets with wide-gap (WG) property can effectively resist the broadband blocking interference, the single frequ...

In this letter, we propose a novel linguistic steganographic method that directly conceals a token-level secret message in a seemingly-natural steganographic text generated by the off-the-shelf BERT model equipped with Gibbs sampling. Compared with all modification based linguistic steganographic methods, the proposed method does not modify a given...

Wireless sensor network (WSN), as one of the core technology in the Industrial Internet of Things (IIoT) system, plays a critical role in collecting data for the monitoring areas. Security and privacy are essential to ensure the trustworthy completeness of data access and transmission for the WSN-based IIoT system. Recently, a series of excellent p...

Multi-view clustering aims at simultaneously obtaining a consensus underlying subspace across multiple views and conducting clustering on the learned consensus subspace, which has gained a variety of interest in image processing. In this paper, we propose the Semi-supervised Structured Subspace Learning algorithm for clustering data points from Mul...

Recently, neural networks have become a promising architecture for some intelligent tasks. In addition to conventional tasks such as classification, neural networks can be used for data hiding. This paper proposes a data hiding scheme to transmit different data to multiple receivers via the same neural network simultaneously. Additional data are em...

Adversarial example generation has been a hot spot in recent years because it can cause deep neural networks (DNNs) to misclassify the generated adversarial examples, which reveals the vulnerability of DNNs, motivating us to find good solutions to improve the robustness of DNN models. Due to the extensiveness and high liquidity of natural language...

In this paper, we introduce a graph representation learning architecture for spatial image steganalysis, which is motivated by the assumption that steganographic modifications unavoidably distort the statistical characteristics of the hidden graph features derived from cover images. In the detailed architecture, we translate each image to a graph,...

Abstract Data hiding aims to embed a secret message into a digital object such as image by slightly modifying the object content without arousing noticeable artefacts. The resultant object containing hidden information will be sent to a desired receiver via some insecure channels, e.g. images transmitted through noisy channel, social networks are v...

Recent advances in linguistic steganalysis have successively applied CNNs, RNNs, GNNs and other deep learning models for detecting secret information in generative texts. These methods tend to seek stronger feature extractors to achieve higher steganalysis effects. However, we have found through experiments that there actually exists significant di...

In order to protect the intellectual property (IP) of deep neural networks (DNNs), many existing DNN watermarking techniques either embed watermarks directly into the DNN parameters or insert backdoor watermarks by fine-tuning the DNN parameters, which, however, cannot resist against various attack methods that remove watermarks by altering DNN par...

Reversible data hiding (RDH) has become a hot spot in recent years as it allows both the secret data and the raw host to be perfectly reconstructed, which is quite desirable in sensitive applications requiring no degradation of the host. A lot of RDH algorithms have been designed by a sophisticated empirical way. It is not easy to extend them to a...

In this paper, we present a novel method to covert communication based on a host deep neural network (DNN) itself, which is totally different from many traditional works that embed secret data into a digital image since: 1) there has no direct image transmission between the data hider and the data receiver, 2) there has no modification to the image...

Information hiding allows us to hide secret information into digital objects such as images without significantly distorting the objects. The object containing hidden information will be transmitted to a data receiver via a probably insecure channel. To securely transmit the object carrying hidden information, the distortion caused by data embeddin...

Symmetric NMF (SNMF) is able to determine the inherent cluster structure with the constructed graph. However, the mapping between the empirically constructed similarity representation and the desired one may contain complex structural and hierarchical information, which is not easy to capture with single learning approaches. Then, we propose a nove...

Recent linguistic steganalysis methods model texts as sequences and use deep learning models to extract discriminative features for detecting the presence of secret information in texts. However, natural language has a complex syntactic structure and sequences have limited representation ability for text modeling. Moreover, previous methods tend to...

Data hiding is a technique to embed secret data into cover multimedia for covert communication. In this letter, we propose a method to disguise the data hiding tools, including a data embedding tool and a data extraction tool, as a deep neural network (DNN) with ordinary task. After training the DNN for both style transfer and data hiding tasks, wh...

Deep Convolutional Neural Networks (DCNNs) are capable of obtaining powerful image representations, which have attracted great attentions in image recognition. However, they are limited in modeling orientation transformation by the internal mechanism. In this paper, we develop Orientation Convolution Networks (OCNs) for image recognition based on t...

Social media plays an increasingly important role in providing information and social support to users. Due to the easy dissemination of content, as well as difficulty to track on the social network, we are motivated to study the way of concealing sensitive messages in this channel with high confidentiality. In this paper, we design a steganographi...

Data hiding aims to embed a secret payload into a cover object without introducing significant degradation of the cover. The resulting object containing hidden information, typically also called stego, will not arouse obvious suspicion from the monitor. A number of data hiding systems have been designed for digital images and only a few focus on vi...

Many learning tasks require us to deal with graph data which contains rich relational information among elements, leading increasing graph neural network (GNN) models to be deployed in industrial products for improving the quality of service. However, they also raise challenges to model authentication. It is necessary to protect the ownership of th...

Watermarking neural networks is a quite important means to protect the intellectual property (IP) of neural networks. In this paper, we introduce a novel digital watermarking framework suitable for deep neural networks that output images as the results, in which any image outputted from a watermarked neural network must contain a certain watermark....

Embedding data into texture or high-frequency region will not arouse noticeable artifacts, meaning that, a low distortion can be achieved when suited region for data embedding is selected. It motivates us to extract texture features and spatial-correlation features to evaluate carrier characteristics for quantizing robustness. Meantime, since large...

The modern social networks are huge and complex, with lots of users and connections, and are well suited for steganography. Recently, Wu et al. introduce a novel steganographic approach through graph structure, which is represented by a series of sequential interactions over online social networking service (SNS). However, since the SNS is public t...

As an effective means to content authentication and privacy protection, reversible data hiding (RDH) permits us to hide a payload such as authentication data in a media file. The resulting marked content will not introduce noticeable artifacts. In order to achieve superior payload-distortion performance, the conventional RDH algorithms often exploi...

With the development of smartphones and mobile wireless network, social discovery applications are becoming more and more popular now. These applications help users to chat with nearby people and make friends. However, the location of a user could be exploited by attackers in some situations. This paper concentrates on the protection of location pr...

Separable reversible watermarking enables two encoders to separately embed a payload in a cover, and the original cover can be reconstructed by cooperation. It is required to limit the embedding distortion for the two encoders so that the marked content will not be seriously degraded, whereas both encoders expect to embed a sufficient payload. It m...

Social bots are computer software designed for content production and interaction with humans. With the popularity of images in social networks, social bots need to have visual awareness of image content while only understanding texts is far from enough to be active in social networks. We introduce a novel task, Visual Social Comment (VSC), in whic...

Developing the technology of reversible data hiding based on video compression standard, such as H.264/advanced video coding, has attracted increasing attention from researchers. Because it can be applied in some applications, such as error concealment and privacy protection. This has motivated us to propose a novel two-dimensional reversible data...

The performance of the steganography detector built on deep learning has been superior to the traditional feature-based methods, and more adaptive methods for steganalysis are beginning to emerge. However a single model may encounter a bottleneck in classification accuracy due to the absent diversity of training data and parameter configuration, it...

Conventional steganography embeds secret data into an innocent cover object such as image and video. The resulting stego object will be sent to the desired receiver via an insecure channel. Though the channel monitor cannot distinguish between normal objects and objects containing hidden information, he has the ability to intercept and alter the ob...

The widespread use of text over online social networks makes it quite suitable for steganography. Conventional text steganography usually transmits the secret data by either slightly modifying the given text or hiding the secret data through synonym replacement. The rapid development of deep neural networks (DNNs) has led automatically generating t...