Yuling Liu

• Hunan University

Sparse R-CNN is a new paradigm of object detection, which predicts objects in a sparse way. However, there are some limitations in Sparse R-CNN. One is the presence of weak prior information caused by fixed learnable proposal boxes and features across different images, necessitating excessive iterations for the model to refine its predictions; the...

AI-generated content has accelerated the topic of media synthesis, particularly Deepfake, which can manipulate our portraits for positive or malicious purposes. Before releasing these threatening face images, one promising forensics solution is the injection of robust watermarks to track their own provenance. However, we argue that current watermar...

Since different kinds of face forgeries leave similar forgery traces in videos, learning the common features from different kinds of forged faces would achieve promising generalization ability of forgery detection. Therefore, to accurately detect known forgeries while ensuring high generalization ability of detecting unknown forgeries, we propose a...

Generative text steganography has received considerable attention in the covert communication community for the benefit of sending secret messages without the need to modify carriers. Existing methods typically choose the next word when generating a stego-text based on conditional probability encoding of candidates, which may lead to generating ina...

We explore the reasons for the poorer feature extraction ability of vanilla convolution and discover that there mainly exist three key factors that restrict its representation capability, i.e. regular sampling, static aggregation and limited receptive field. With the cost of extra parameters and computations, existing approaches merely alleviate pa...

Detecting if two functions in different compiled forms are similar has a wide range of applications in software security. We present a method that leverages both semantic and structural features of functions, learned by a neural-net model on the underlying control-flow graphs (CFGs). In particular, we devise a neural function-similarity regressor (...

Classical machine learning works ineffectively when the data set is small. Recently, few-shot learning is proposed to solve this problem. Few-shot learning models a few samples through the prior knowledge. We could divide few-shot learning into various categories depending on where the prior knowledge is extracted from. There are mainly three class...

Due to the imbalance between a large number of litigation cases and the number of judicial personnel, many legal documents to be processed greatly increase the burden of legal practitioners. So the intelligent processing of legal documents is especially important. At present, machine learning and deep learning have made great achievements in the in...

Maximal biclique enumeration is a fundamental problem in analysing bipartite graphs, which has a wide range of real applications, such as web mining, recommendation systems, and social network analysis. As real-world bipartite graphs constantly evolve over time, it is useful and necessary to incrementally maintain maximal bicliques in dynamic bipar...

In large dynamic graphs, it is often impractical to store and process the entire graph. To contend with such graphs, lossless graph summarization is a compression technique that can provide a succinct graph representation without losing information. However, given a fully dynamic graph stream, the problem of lossless summarization is computationall...

In recent years, fatigue driving has been a serious threat to the traffic safety, which makes the research of fatigue detection a hotspot field. Research on fatigue recognition has a great significance to improve the traffic safety. However, the existing fatigue detection methods still have room for improvement in detection accuracy and efficiency....

Stocks price prediction is a current hot spot with great promise and challenges. Recently, there have been many stock price prediction methods. However, the prediction accuracy of these methods is still far from satisfactory. In this paper, we propose a stock price prediction method that incorporates multiple data sources and the investor sentiment...

Steganography based on generative adversarial networks (GANs) has become a hot topic among researchers. Due to GANs being unsuitable for text fields with discrete characteristics, researchers have proposed GAN-based steganography methods that are less dependent on text. In this paper, we propose a new method of generative lyrics steganography based...

With the large-scale application of cloud storage, how to ensure cloud data integrity has become an important issue. Although many methods have been proposed, they still have their limitations. This paper improves some defects of the previous methods and proposes an efficient cloud data integrity verification scheme based on blockchain. In this pap...

In rapid and massive graph streams, it is often impractical to store and process the entire graph. Lossless graph summarization as a compression technique can provide a succinct graph representation without losing information. However, the problem of lossless streaming graph summarization is computationally and technically challenging. Although the...

Facial recognition technology has always been a challenging field in computer vision and pattern recognition. At present, facial recognition technology has been widely used in daily life. Much research has been done on facial recognition systems. However, we still need to continuously enhance and improve the facial recognition system in practical a...

Due to the great success of convolutional neural networks (CNNs) in the area of computer vision, the existing methods tend to match the global or local CNN features between images for near-duplicate image detection. However, global CNN features are not robust enough to combat background clutter and partial occlusion, while local CNN features lead t...

Text classification has always been an increasingly crucial topic in natural language processing. Traditional text classification methods based on machine learning have many disadvantages such as dimension explosion, data sparsity, limited generalization ability and so on. Based on deep learning text classification, this paper presents an extensive...

As an important way of communication, the wireless network has brought great convenience to people’s life because of its rapid development. It is a hot topic to protect the data of the wireless network stored in the cloud. In order to solve the problem, this paper presents a privacy protection scheme based on confusing keyword. Based on user histor...

Although the development of the Internet has enabled more and more texts to be transmitted in electronic forms, printed documents are still an irreplaceable form of transmission. Whether it is the copyright protection of book authors or the prevention of the leakage of confidential documents, there is an urgent need for text watermarking resistant...

We present a Recurrent Neural Network (RNN) Encoder-Decoder model to generate Chinese pop music lyrics to hide secret information. In particular, on a given initial line of a lyric, we use the LSTM model to generate the next Chinese character or word to form a new line. In so doing, we generate the entire lyric from what has been generated so far....

Coverless information hiding has become a hot topic because it can hide secret information (SI) into carriers without any modification. Aiming at the problems of the low hiding capacity (HC) and mismatch in text big data, a novel method of coverless information hiding by retrieving the massive amount of web text on the Internet. Firstly, the propos...

Coverless information hiding does not make any modifications to the carrier, so it can effectively resist the various steganalysis and detection algorithms. However, the existing methods still have some problems that include low hiding capacity and unsatisfactory hiding success rate. To address these problems, this paper proposes a method of text c...

In recent years, the wide applications of the wireless sensor networks have achieved great success. However, the security is a critical issue in many scenarios ranging from covert military operations to the organization of the social unrest. Because the traditional encrypting methods are easy to arouse suspicion, an adaptive audio steganography met...

Data outsourcing has become an important application of cloud computing. Driven by the growing security demands of data outsourcing applications, sensitive data have to be encrypted before outsourcing. Therefore, how to properly encrypt data in a way that the encrypted and remotely stored data can still be queried has become a challenging issue. Se...

Sentiment analysis is an increasingly important area in NLP to extract opinions and sentiment expressed by humans. Traditional methods are often difficult to tackle the problems of different sample distribution and domain dependence, which seriously limits the development of sentiment classification. In this paper, a novel sentiment analysis method...

Rationale and objectives: The purpose of this study was to measure and analyze interobserver disagreement in rating diagnostic characteristics of pulmonary nodules on computed tomography scans using the Lung Imaging Database Consortium and Image Database Resource Initiative (LIDC/IDRI) database, and then to provide investigators with understanding...

A robust image hashing method based on radon transform and invariant features is proposed for image authentication, image retrieval, and image detection. Specifically, an input image is firstly converted into a counterpart with a normalized size. Then the invariant centroid algorithm is applied to obtain the invariant feature point and the surround...

A novel ROI-based reversible data hiding scheme in encrypted medical images is proposed. Firstly, a content owner partitions an original medical image into the region of interest (ROI) and the region of noninterest (RONI), and then encrypts the image using an encryption key. A data-hider concatenates the least significant bits (LSB) of the encrypte...

This paper proposes an improved method for reversible data hiding in encrypted image based on Zhang’s algorithm when the embedding ratio is considered. A new fluctuation function is presented to evaluate the smoothness of image block. The fluctuation values of blocks are calculated, and then are sorted in ascending order. The blocks that the fluctu...

Communications via instant message tools have become increasingly popular in people's daily lives. However, one of the main issues in communication is the transmission of secret information. There are many methods for covert communications. In this paper, a novel text steganography method in chat is proposed which utilizes emoticons and interjectio...

In recent years, due to the appealing features of cloud computing, more and more sensitive or private information has been outsourced onto the cloud. Although cloud computing provides convenience, privacy and security of data becomes a big concern. For protecting data privacy, it is desirable for the data owner to outsource sensitive data in encryp...

A robust zero-watermarking algorithm is proposed based on merging features of sentences for Chinese text document authentication. In the scheme, a text is first segmented into sets of sentences, where a semantic code for every word can be obtained. Then the sentence entropy is calculated by the frequency of semantic codes, and the sentence relevanc...

A novel ROI-based reversible data hiding scheme is proposed for medical images, which is able to hide electronic patient record (EPR) and protect the region of interest (ROI) with tamper localization and recovery. The proposed scheme combines prediction error expansion with the sorting technique for embedding EPR into ROI, and the recovery informat...

Content authentication of text document has become a major concern in the current digital era. In this paper, a zero-watermark algorithm is proposed for Chinese text documents content authentication. Firstly, the frequencies of different part-of-speech (POS) tags are obtained through natural language processing technology. And they are used to calc...

This paper proposes a robust image hashing method which is robust against common image processing attacks and geometric distortion attacks. In order to resist against geometric attacks, the log-polar mapping (LPM) and contourlet transform are employed to obtain the low frequency sub-band image. Then the sub-band image is divided into some non-overl...

With the development of Internet, a huge volume of digital contents floated over the Internet daily. Digital watermarking is an effective way to realize authentication and copyright protection for digital contents over. Text is a dominant part of the internet contents, but there are few techniques available for text protection. This paper proposes...

A new steganographic method for data hiding in Microsoft Word 2007–2010 (Microsoft Corp., Redmond, WA, USA) files that use Office Open XML (OOXML) format is proposed. Secret information can be imperceptibly embedded into OOXML documents by splitting up the printable text, which is defined by the main document body of the OOXML format document. The...

In P2P network, the existing researches focus on protecting the security of information transmission, or ensuring users' privacy. Security certification is a basic demand of P2P network, including the general authenticity, credibility, integrity and so on. Privacy protection for transactions is the high-level requirement of security, including conf...

If a document copy detection method is effective, it need to be able to detect plagiarized contents with high accuracy. Most of the document copy detection methods proposed so far can do well in identifying an exact copy of document, paragraph, or sentence. But when words and sentences structure change, false negative rate of them is high. We propo...

MS Office documents could be illegally copied by offenders, and forensic investigators still face great difficulty in investigating and tracking the source of these illegal copies. This paper mainly proposes a forensic method based on the unique value of the revision identifier (RI) to determine the source of suspicious electronic documents. This m...

A crucial issue of authorship identification is to find a set of features which can represent the writing style of a particular author. Due to Chinese special characteristics, the features used for identifying Chinese authorship are relatively few. In this paper, we propose a method of Chinese authorship identification based on Dependency Grammar....

In order to improve the accuracy of Chinese license plate location, a method based on the number of binary image jump and mathematical morphology is proposed in the paper. After the pretreatment of vehicle plate image and threshold values segmentation, the top and the bottom location of license plate can be determined by obtaining the number of bin...

In this study, a novel secure steganographic method was proposed by taking advantage of Multiple-Choice Questions (MCQs). Selecting a series of MCQs to automatically generate a stego text could conceal secret information. The proposed method could achieve a considerable embedding bit rate while being able to hide more information by reordering the...

N-gram frequency is always an important indicator in corpus-based natural language processing. As the corpus being larger, time and space consumption is always a serious issue. In this paper, we proposed a two-stage inverted index based on bigrams. The proposed index is constructed by two levels of variable length vectors and the two characters of...

The digital information revolution and increasing of outsourced data in many organizations caused significant changes in the global society where digital data are available everywhere with free of cost. Proving ownership rights of outsourced relational databases is a crucial issue in today internet-based application environments and in many content...

In some critical application areas, it is not allowed to modify the contents of the text, like military, legal and literature fields. Therefore, restoring the original contents of the text becomes a practical and important issue for text watermarking. This study aims to deal with this problem. We firstly present the concept of reversible text water...

A novel method of natural language information hiding is presented based on Chinese mathematical expression. The method can convert Chinese text into bit string by using the thought of Chinese mathematical expression, and then embed the secret message into the text by utilizing the replacement of synonyms, variant forms of the same word, the adding...

As text steganography becomes a new research hotspot of security communication recently, text steganalysis, whose purpose is to detect the presence of secret message in text, has attracted more and more attention. At present, all existing methods concerning text steganography can be roughly divided into three categories: those based on invisible ch...

Microsoft PowerPoint files (PPT files for short) are widely used today but steganographic schemes for them have rarely been reported. Through detailed analysis, we discover that the PPT file update mechanism, which is employed by Microsoft PowerPoint Application to update the corresponding PPT file when saving a modified presentation, can produce r...

In this study, a novel natural language watermarking scheme of Chinese texts based on syntactic transformations is proposed. In the proposed scheme, the so-called sentence weight defined in the study is used to carry the watermark bit. During the embedding process, the transformable sentences are first grouped and then their sequence is permuted to...

Communications via Microsoft Power Point (PPT for short) documents are commonplace, so it is crucial to take advantage of PPT documents for information security and digital forensics. In this study, we propose a new method of text steganography, called MIMIC-PPT, which combines text mimicking technique with characteristics of PPT documents. Firstly...

Because of the existence of a large number of non-totally interchangeable synonymy sets, the present natural language steganographic approaches based on Chinese synonymy substitution may destroy the semantic coherence after synonymy substitutions. To deal with this disadvantage, this paper proposes an improved steganographic algorithm based on syno...

Linguistic steganography, as a method of text steganography, is becoming a hot spot. To investigate the linguistic steganography for Chinese text, a Chinese linguistic steganography algorithm is presented by utilizing the existing Chinese information processing techniques. The algorithm is based on the substitution of synonyms and variant forms of...

After analyzing some popular algorithms for text steganography, and according to the peculiarity of text steganography, this paper proposed an approach to steganalyse text based on its noise, and then designed and implemented a system for text Steganalysis. The system can judge whether or not a text has a secret data by detecting the noise values o...

A novel text watermarking algorithm is presented. It combines natural language watermarking and Chinese syntax based on BP neural networks. Since the watermarking signals are embedded into some Chinese syntactic structure rather than the appearance of text elements, the algorithm is totally based on the content that can prove to be very resilient....