Convolutional neural networks (CNNs) can automatically learn data patterns to express face images for facial expression recognition (FER). However, they may ignore effect of facial segmentation of FER. In this paper, we propose a perception CNN for FER as well as PCNN. Firstly, PCNN can use five parallel networks to simultaneously learn local facial features based on eyes, cheeks and mouth to realize the sensitive capture of the subtle changes in FER. Secondly, we utilize a multi-domain interaction mechanism to register and fuse between local sense organ features and global facial structural features to better express face images for FER. Finally, we design a two-phase loss function to restrict accuracy of obtained sense information and reconstructed face images to guarantee performance of obtained PCNN in FER. Experimental results show that our PCNN achieves superior results on several lab and real-world FER benchmarks: CK+, JAFFE, FER2013, FERPlus, RAF-DB and Occlusion and Pose Variant Dataset. Its code is available at https://github.com/hellloxiaotian/PCNN.

In this paper, we develop an online change-point detection procedure in the covariance structure of high-dimensional data. A new stopping rule is proposed to terminate the process as early as possible when a change in covariance structure occurs. The stopping rule allows spatial and temporal dependence and can be applied to non-Gaussian data. An explicit expression for the average run length is derived, so that the level of threshold in the stopping rule can be easily obtained with no need to run time-consuming Monte Carlo simulations. We also establish an upper bound for the expected detection delay, the expression of which demonstrates the impact of data dependence and magnitude of change in the covariance structure. Simulation studies are provided to confirm accuracy of the theoretical results. The practical usefulness of the proposed procedure is illustrated by detecting the change of brain’s covariance network in a resting-state fMRI data set. The implementation of the methodology is provided in the R package OnlineCOV.

Few-shot aerial image object detection aims to rapidly detect object instances of novel category in aerial images by using few labeled samples. However, due to the complex background of aerial images, few labeled samples of novel categories, and the model trained with the few-shot learning paradigm is biased towards the base categories, it greatly increases the difficulty of identifying foreground objects of novel categories. In addition to this, tiny object detection is always a hot potato in aerial image object detection, and it is even more difficult for few-shot object detection. To this end, we propose a Few-shot aerial image object detection with Confidence-Iou collaborative proposal filtration and Tiny object constraint loss (FsCIT). Specifically, we first introduce a new confidence-iou collaborative proposal filtration scheme to RPN, which combines the unbiased IoU scores between the two bounding boxes with foreground-background confidence scores to filter redundant region proposals and rescue more foreground proposals for the novel categories in RPN. Then, we design a new tiny object loss constraint term to attempt at overcoming the challenge of tiny object detection in few-shot aerial image object detection. This term considers the central point distance, the size of ground-truth bounding boxes, and the distances between the four edges of the ground-truth bounding box and the predicted bounding box. Experiments on DIOR, AI-TOD and HRRSD datasets show that FsCIT is effective and can improve the performance of few-shot aerial image object detection.

Automatic image captioning is to conduct the cross-modal conversion from image visual content to natural language text. Involving computer vision (CV) and natural language processing (NLP), it has become one of the most sophisticated research issues in the artificial-intelligence area. Based on the deep neural network, the neural image caption (NIC) model has achieved remarkable performance in image captioning, yet there still remain some essential challenges, such as the deviation between descriptive sentences generated by the model and the intrinsic content expressed by the image, the low accuracy of the image scene description, and the monotony of generated sentences. In addition, most of the current datasets and methods for image captioning are in English. However, considering the distinction between Chinese and English in syntax and semantics, it is necessary to develop specialized Chinese image caption generation methods to accommodate the difference. To solve the aforementioned problems, we design the NICVATP2L model via visual attention and topic modeling, in which the visual attention mechanism reduces the deviation and the topic model improves the accuracy and diversity of generated sentences. Specifically, in the encoding phase, convolutional neural network (CNN) and topic model are used to extract visual and topic features of the input images, respectively. In the decoding phase, an attention mechanism is applied to processing image visual features for obtaining image visual region features. Finally, the topic features and the visual region features are combined to guide the two-layer long short-term memory (LSTM) network for generating Chinese image captions. To justify our model, we have conducted experiments over the Chinese AIC-ICC image dataset. The experimental results show that our model can automatically generate more informative and descriptive captions in Chinese in a more natural way, and it outperforms the existing image captioning NIC model.

Few-shot learning (FSL), which aims to rapidly recognize unseen categories with limited samples, has attracted wide attention in aerial image scene classification. However, the existing methods generally train and evaluate the model within a dataset, and changing the dataset requires retraining and evaluation, which only realizes the generalization of intra-dataset. Considering meta-learning, this brings in a natural assumption: FSL should learn meta-knowledge from cross-domain heterogeneous tasks and then can generalize to new data distributions (e.g., datasets) with few samples. To this end, we propose a new benchmark, dubbed aerial image few-shot dataset (AIFS-DATASET), which is composed of diverse datasets and can provide more realistic heterogeneous task distributions. On AIFS-DATASET, we use many heterogeneous tasks, across multi-domains without any aerial image category, to train the model, achieving “see more.” Then we transfer the learned knowledge to new tasks in aerial images to evaluate the generalization performance of the model, thus acquiring a “well-informed” few-shot aerial image scene classification model. Moreover, the challenges of inter-class similarity and intra-class discrepancy in aerial images still exist. We also develop a dual constrained distance metric learning (DC-DML) framework to deal with the variable learning tasks adaptively and to achieve compact data distribution within a class and clear distribution gaps between classes from the perspective of metric learning. DC-DML mainly uses a task-adapted feature extractor while devising a novel distance metric with a cross-class bias penalty. By conducting experiments on AIFS-DATASET, we observed that DC-DML outperforms the current prevailing FSL approaches by a large margin.

In this article, we attempt to achieve one-shot object detection by mimicking the human ability to learn new concepts under limited reference, which aims at detecting all object instances of an unseen class in a target image when given a query image of the same unseen class. However, this one-shot learning ability of human benefits from the fact that human brain can quickly extract and process the associated information between the query–target images, which is an issue for the one-shot object detection framework to overcome. Moreover, the feature extraction of the query class in target images is intractable due to the complex and diversified background of remote sensing images. To solve these issues, we propose a solo-to-collaborative dual-attention network (SCoDANet) to hierarchically (image itself/pairs) enhance image feature representations. It consists of three components: 1) solo-attention head that strengthens the compactness of intraclass feature representations of an image and avoids background interference by selectively aggregating the similar features from the spatial and channel dimensions, respectively; 2) dual coattention module that guides RPN to generate an expected set of region proposals related to the query class by mining the coinformation of each query–target feature pair; and 3) nonlinear matching that provides a measure of similarity between the query feature and proposals of the target image to further learn a more robust detector. Our extensive experiments over two benchmarks demonstrate the effectiveness of our method under the one-shot scenario of detecting seen and unseen object categories.

Few-shot scene classification aims to recognize unseen scene concepts from few labeled samples. However, most existing works are generally inclined to learn metalearners or transfer knowledge while ignoring the importance to learn discriminative representations and a proper metric for remote sensing images. To address these challenges, in this article, we propose an end-to-end network for boosting a few-shot remote sensing image scene classification, called discriminative learning of adaptive match network (DLA-MatchNet). Specifically, we first adopt the attention technique to delve into the interchannel and interspatial relationships to automatically discover discriminative regions. Then, the channel attention and spatial attention modules can be incorporated with the feature network by using different feature fusion schemes, achieving “discriminative learning.” Afterward, considering the issues of the large intraclass variances and interclass similarity of remote sensing images, instead of simply computing the distances between the support samples and query samples, we concatenate the support and query discriminative features in depth and utilize a matcher to “adaptively” select the semantically relevant sample pairs to assign similarity scores. Our method leverages an episode-based strategy to train the model. Once trained, our model can predict the category of query image without further fine-tuning. Experimental results on three public remote sensing image data sets demonstrate the effectiveness of our model in the few-shot scene classification task.

The plenoptic 2.0 video records time-varying light fields varying with much higher spatial resolution for each subaperture image. However, the huge data volume becomes a significant obstacle of the plenoptic 2.0 video application. Therefore, this paper proposes an efficient compression method designed for plenoptic 2.0 video coding. First, the matching blocks relative to the current block are determined by imaging principles. The matching blocks and current block are imaged by the same object under neighboring microlens, which are highly correlated with each other. Then, weighted prediction by using the matching blocks as reference blocks is adopted for the intra prediction of current block. Compared with HEVC, experimental results show that the proposed method can save 9.63% and 4.73% bitrate under the All Intra and Random Access configuration, respectively.

High mobility and high spectrum efficiency are two main challenges for 5G communication. Many new technologies are proposed to meet these requirements. Orthogonal time frequency space (OTFS) is a novel scheme to combat the high Doppler effect under time-varying channel. In this paper, an index modulation based orthogonal time frequency space (OTFS-IM) system is proposed for improving BER performance under both high mobility and high spectrum efficiency. In the scheme, the information is conveyed by both constellation symbols and indices bits on Delay-Doppler domain. A low complexity detection strategy based on integrated MMSE-ML detection is proposed, which is very flexible and suitable to high spectrum efficiency signal. By simulation, compared with the conventional OTFS, the proposed OTFS-IM system can achieve better BER performance, maintaining good tolerability against Doppler effect, especially for higher spectral efficiency.

Recently, light field images have received extensive attention due to their potential applications. Since they take up a huge memory because of its super-high resolution, efficient compression methods are fundamentally required. In this paper, we propose a novel intra prediction mode by using depth-adaptive convolutional neuro network (DCNN). Light field projection finds the imaging response distribution for each object point using the depth estimated from each macropixel in the light field image. The highly correlated imaging responses are used to select the neural network structure. The network structure also adapts to the to-be-encoded block size. Adding the proposed DCNN-based prediction mode into the rate-distortion optimization loop with other 35 intra prediction modes of HEVC, the proposed encoding scheme achieves a significant bit-rate saving compared to representative compression approaches with limited computational complexity increment. Statistical data are also provided and analyzed to demonstrate the efficiency of the proposed method.

The plenoptic video recording light fields varying with time has the large-range motion and complex macropixel structure, which brings a great challenge to efficient compression. Based on the analysis of the relationship between temporal motion and macropixel arrangement, an efficient motion estimation algorithm is proposed for plenoptic video compression. It finetunes motion vector predictor (MVP) by the macropixel-constrained collocated position of the current prediction unit in the nearest macropixel, and uses all the macropixel-constrained collocated positions, derived from the finetuned MVP, as motion search candidates for better complexity and compression-efficiency trade-off. The experimental results demonstrate that the proposed algorithm outperforms multi-view based compression and the pseudovideo based compression by an average of 56.40% and 78.72% bitrate reduction, respectively. Compared with HEVC, the proposed method can also save 7.65% bitrate.

Orthogonal time frequency space (OTFS) modulation scheme is based on a novel 2-dimensional modulation technique designed to transform the time-varying multi-path channel into a time invariant Delay- Doppler channel. In general, the Message Passing (MP) algorithm works well for data detection in OTFS, but its complexity is quite large. In this paper, a low complexity MP algorithm for data detection in OTFS is presented based on a Doppler compensation precoder and a modified stopping criterion. Besides, an improved Approximate Message Passing (AMP) algorithm based on covariance processing, having almost the same complexity as that of conventional AMP algorithm, but exhibiting better bit error rate(BER) performance is presented, as demonstrated by simulation results.

Due to the environment and equipment are not controllable, the process of face image acquisition is inevitable to be interfered by external factors, and there are usually only a small number of available face images. Insufficient samples are not conducive to face recognition. Therefore, it is a popular scheme to produce virtual samples based on the available training samples. In this study, the authors first take the symmetry of human face into account, and propose a novel method to generate virtual samples. Then a representation‐based classification method and the score fusion strategy are applied to both original face images and virtual images to perform face recognition. Several sparse representation‐based classification algorithms are compared on ORL, FERET and GT databases. Experimental results show that the authors’ method is effective for improving the face recognition.

In online surveys, many people are reluctant to provide true answers due to privacy concerns. Thus, anonymity is important for online message collection. Existing solutions let each member blindly shuffle the submitted messages by using an IND-CCA2 secure cryptosystem. In the end, the message sender's identities are protected since no one knows the message submission order. These approaches cannot efficiently handle groups of large size.In this paper, we propose an efficient anonymous message submission protocol aimed at a practical group size. Our protocol is based on a secret sharing scheme and a symmetric key cryptosystem. We propose a novel method to aggregate members' messages into a message vector such that a group member knows only his own position in the submission sequence. The protocol is accountable for capturing malicious members breaking the protocol execution. We provide a theoretical proof showing that our protocol is anonymous under malicious attacks. We also discuss our simulation results to demonstrate the efficiency of our protocol.

Air pollution is becoming a serious issue that threatens everyone's daily life. Accordingly how to measure the local air quality easily and quickly becomes an urgent problem. Sensor-based methods are relatively expensive, and image based air quality measurement is a promising direction as it is often less cost. Meanwhile web service of precise air quality measurement is of great importance as it allows to timely monitor the air pollution and can provide recommendations for decision makers. This paper devises an effective web service to address this challenging problem. Specifically we offer a service letting mobile device users to upload photos taken outdoor with meta information. Once the background web service computation is finished, we return the air quality level at their location to the users. In our service system, it includes three basic modules: 1) using GPS location information to get the basic air quality value from the nearest official air quality station, 2) using photo by our air quality assessment based on dictionary learning for image representation to compute air quality. In this method we add ℓ21norm to the target function to promote the nonzero coefficients of the words aggregate within similar quality level, 3) using photo uploaded by user by our CNN based photo air pollution estimation method to obtain the air quality estimation. Finally we combine the above three estimations to reach the final estimation to the end users. Empirical experiments are conducted on the real-world dataset that collaborates the efficacy of our method.

Among all image representation and classification methods, sparse representation has proven to be an extremely powerful tool. However, a limited number of training samples are an unavoidable problem for sparse representation methods. Many efforts have been devoted to improve the performance of sparse representation methods. In this study, the authors proposed a novel framework to improve the classification accuracy of sparse representation methods. They first introduced the concept of the approximations of all training samples (i.e., virtual training samples). The advantage of this is that the application of virtual training samples can allow noise in original training samples to be partially reduced. Then they proposed an efficient and competent objective function to disclose more discriminant information between different classes, which is very significant for obtaining a better classification result. The devised sparse representation method employs both the original and virtual training samples to improve the classification accuracy since the two kinds of training samples makes sample information to be fully exploited in a good way, also satisfactory robustness to be obtained. The experimental results on the JAFFE, ORL, Columbia Object Image Library (COIL‐100) AR and CMU PIE databases show that the proposed method outperforms the state‐of‐art image classification methods.

Authenticating whether two smartphones are in close proximity is important in smartphone security. For example, the authentication result can be used to pair two devices and construct a secure communication channel between them. Many existing proximity authentication systems rely on short range networks-the communication is usually restricted in short range networks. However, this approach is inadequate when we want to verify whether the communication distance is within a few centimeters, i.e. near field. To address this challenge, many other techniques construct systems based on the near field communication (NFC) system. Unfortunately, only a small portion of smart devices in the current market are equipped with NFC chips. The purpose of this paper is to provide a close proximity authentication system which does not depend on NFC chips. We devise a system to achieve close proximity authentication by using correlated finger movements on the two smartphones. Human input usually contains errors and is of low entropy, which affects the usability and security of our system. We solve these issues in an efficient way, considering the limited computational resources on smart devices. Our system does not need any prior secret information shared between the two devices, and generates the same high-entropy cryptographic key for both devices in a successful authentication. The efficiency of the system is validated by evaluations on Motorola Droid smartphones.

Asserting whether two devices are in close proximity is very important to many smartphone assisted security systems. For example, the smartphone based two-factor authentication usually requires the smartphone to stay in close proximity to the other device during authentication. However, relay attacks pose a serious threat to existing approaches for proximity assertions. In this paper, we present a novel near field assertion system that restricts the distance between the two devices to the scale of several centimeters. Our system explores acoustic communications and can prevent relay attacks. The generated assertion is a confidential binary sequence known only to the two devices. Our system is fully automated and light-weight, as demonstrated by extensive evaluations on a prototype.

One of the latest trends of realizing innovative Command and Control (C&C) channels involves leveraging Online Social Networks (OSNs) as a C&C channel. The number of botnets targeting the smartphones and the sophistication of those botnets have progressively increased. Due to their mobility, smartphones connect to a variety of networks which makes it harder for network centric detection of botnets in smartphones. This paper approaches the problem of detecting bot traffic from a host based detection perspective. In this paper, we first propose an innovative C&C that leverages "public information" in OSNs combined with a Username Generation Algorithm. We then propose a new system to detect the bots that leverage the above mentioned type of C&C channel. Our insight is that the user generated web traffic on the smartphones will be significantly different from the requests made by the bots that leverage OSNs as C&C channel. Our approach involves building a profile of the smartphone user based on his web usage and then comparing that profile to subsequent usage to detect anomalous behavior. The Preprocessing phase clusters the web usage based on domains and extracts relevant features. In the next step, we use classification algorithm to build the user profile and assign a score of mismatch to the domains compared to the user behavior. If the score crosses a threshold, then the traffic to that domain is perceived to be different from normal user traffic to that domain and the user will be notified. Based on his response, the model will be updated to incorporate the change into it. We implemented a prototype bot and detection system and evaluated it on real-world user traffic. Our system reports an accuracy of 76%, with false positive rate of less than 1%.

Location based service is an indispensable part of today's mobile era. While it brings a lot of benefits to people, the breach to individual location privacy is always a concern and impedes the smooth development of location based service. A user can be easily tracked once she subscribes or uses the service from an untrusted location based service server. In this paper, we try to address this problem by proposing a secure and efficient location based service system. In our system, a user does not leak any of her location information while she can still obtain the desired information associated with the location. We propose a novel method to map a user's current location to the index of the information stored in the location based service server. We demonstrated the efficiency of our system through simulations.

Reminder applications are essential applications in mobile devices. Since most smart devices are equipped with accurate localization capabilities, location based reminders emerge in recent smart devices. A user could add a location based reminder which reminds the user to do something once she enters or leaves a location. Convenient as these applications are, a user can be easily tracked once she installs these applications. We propose a secure and efficient location based reminder system. In our system, the reminder location and reminder message are stored in the form of ciphertext on the cloud server. The cloud server is able to preform a private location match without knowing anything about the user's location information. We propose a novel method to represent the user's reminder area in order to save the storage space and computation time of both users and the cloud server. We demonstrate the efficiency of our system in our simulations.

In this paper, we propose a secure cloud-assisted location based reminder system. The proposed system is secure and responsive. Our system outsources the location testing task --- testing whether the current location is near a reminder location --- to the cloud server such that the device synchronization is not necessary in the system. This feature makes the proposed system more responsive, especially when the reminder message is of large size, e.g., audio, images. Above all, the proposed system protects a user's location privacy and the confidentiality of the reminder message. The system is designed in a way that the cloud server can perform location testing for a user but cannot learn about her current location, reminder locations, and reminder messages. We prove the security of the proposed system and demonstrate its efficiency using simulations on a Motorola Droid smartphone.

Many smartphone applications use near field communication (NFC) systems to guarantee that two smartphones appear in near field when interactions take place. “Google Wallet” is one such example. This kind of guarantee is called near field authentication, which is to authenticate whether two smartphones stay closely to each other. Using NFC systems is a natural option for near field authentication. However, NFC systems rely on NFC chips, which are not available on many smartphones, especially the low-end smartphones. This obstructs the popularization of NFC applications. In this paper, we propose a simple and lightweight system to perform near field authentication without using NFC chips. The idea is to put two smartphones side by side and let the user slide his finger across the two smartphone screens. The two smartphones then extract correlated feature values from the movement on each screen to authenticate each other. When the two smartphones are in near field, our system generates the same cryptographic key for both. The key can be used by another upper system to carry out confidential communications. Our system is proved to be secure in the random oracle model. We demonstrate its efficiency on Motorola Droid smartphones in our evaluations.

Near field communication (NFC) systems provide a good location-limited channel so that many security systems can use it to force the participants to stay close to each other. Unfortunately, only a small number of smart devices in the market are equipped with NFC chips that are essential for NFC systems. The purpose of this paper is to provide the same feature, called near field authentication (NFA), without using NFC chips. We propose an easy-to-use system to achieve NFA by using human finger movement on the touch screens of two nearby smart devices. Our system does not need any prior secret information shared between two devices and generates the same high-entropy cryptographic key for both devices in a successful authentication. The efficiency of the system is demonstrated by our evaluation on a Motorola Droid smartphone.

In current location based social networks (LBSNs), users expose their location when they check in at a venue or search a place. The release of location privacy could lead to a severe breach of other privacy, such as identity or health condition. In this paper, we propose a framework to safeguard users' location information as well as the check-in records. Considering the special demands in LBSNs, we design a novel index structure to provide a fast search for users when they check in at the same venue frequently. At the same time, our framework outsources the heavy cryptographic computations to the server to reduce the computational overhead for mobile clients. Due to the dynamic feature of LBSNs, our framework uses a lightweight approach to handle a user's revoked friends and new friends. We prove the security of our framework in the random oracle model and demonstrate its efficiency on a Motorola Droid phone.

Recent success of online social networks (OSNs) motivates the study of security issues in OSNs. A fundamental but challenging security issue in OSNs is to authenticate a friend's real identity. A solution to this issue will benefit a number of OSN security protocols. Existing solutions require users securely obtain some secret information from their friends before authentication takes place, which is not always possible in OSNs. In this paper, we propose a new authenticated key exchange protocol based on the exclusive secrets shared between friends. It provides identity authentication and key exchange in a plain setting, i.e., users do not need to securely exchange or distribute any information beforehand. The protocol is designed to work with low-entropy input information, because human beings are not good at dealing with a large amount of information. Another advantage of our protocol is its tolerance of input errors considering human error is always a possibility. We prove the security of the protocol in the universal composability (UC) framework and demonstrate its efficiency.

In this paper, we construct a system which can hide users' identity when they visit untrusted third party storage sites. We also define a fine-grained access control policy for the data owner to freely define who can access the record. That is to say, the data owner divide his friends into several groups and issues them corresponding credentials for accessing his data. However, he can adds a friend at any time in a revocation list (RL) so that that friend could not access the data owner's data any more even if he has credentials. We theoretically prove the security of our protocols, and evaluate the performance of our protocols through simulations.

Authenticating users in online social networks (OSNs) is different from traditional authentication, because the participants in the authentication may not share any prior secret information. In this paper, we propose a decentralized authentication framework to help users authenticate unknown users in an OSN. In our framework, a user requests certificates from other trusted users to prove his identity in authentication. However, collecting certificates is constrained by the fact that trust is usually attritted with the length of a trust chain. Considering this constraint, our framework utilizes a decentralized online learning approach to help users collect more certificates. We further prove that the number of the collected certificates by each user using our protocols is close to optimum.

Group ranking is a necessary process used to find the best participant from a group. Group ranking has many applications, including online marketing, personal interests matching and proposal ranking. In an online virtual environment, participants want to do group ranking without leaking any of their private information. In this work, we generalize this scenario as a privacy preserving group ranking problem and formulate the privacy requirements of this problem. We propose a fully distributed privacy preserving group ranking framework and prove its security in the honest but curious model. The core of our framework is a novel multiparty sorting protocol, which guarantees that an adversary cannot link the private information to its owner's identity as long as the owner's final ranking is hidden from the adversary. Our protocol is efficient in computational overhead and communication rounds compared to existing works, as demonstrated by our analysis and simulation.

In online surveys, many people are not willing to provide true answers due to privacy concerns. Thus, anonymity is important for online message collection. Existing solutions let each member blindly shuffle the submitted messages by using the IND-CCA2 secure cryptosystem. In the end, all messages are randomly shuffled and no one knows the message order. However, the heavy computational overhead and linear communication rounds make it only useful for small groups. In this paper, we propose an efficient anonymous message submission protocol aimed at a practical group size. Our protocol is based on a simplified secret sharing scheme and a symmetric key cryptosystem. We propose a novel method to let all members secretly aggregate their messages into a message vector such that a member knows nothing about other members' message positions.We provide a theoretical proof showing that our protocol is anonymous under malicious attacks. We then conduct a thorough analysis of our protocol, showing that our protocol is computationally more efficient than existing solutions and results in a constant communication rounds with a high probability.

Making new connections is a crucial service in current online social networks. However, such a service can also raise a big security concern. For example, two strangers become friends in the online social network and they want to communicate securely. The messages transmitted between them could be encrypted using their public keys and decrypted using their private keys. However, one can not determine whether this public key indeed belongs to the claimed user. In this paper, we design a system to authenticate two strangers in fast mixing online social networks. We make a thorough analysis on our system and show that there is a high probability of finding witness users in fast mixing social networks.

In this paper, a novel statistical algorithm for linguistic steganography detection, which takes advantage of distribution of words in the text segment detected, is presented. Linguistic steganography is the art of using written natural language to hide the very presence of secret messages. Using the text data, which is the foundational media in Internet communications, as its carrier, linguistic steganography plays an important part in Information Hiding (IH) area. The previous work was mainly focused on linguistic steganography and there were few researches on linguistic steganalisys. We attempt to do something to help to fix this gap. In our experiment of detecting the three different linguistic steganography methods: NICETEXT, TEXTO and Markov-chain-Based, the total accuracies on discovering stego-text segments and normal text segments are found to be 87.39% 95.51%, 98.50%, 99.15% and 99.57% respectively when the segment size is 5 kB, WkB, 20 kB, 30 kB and 40 kB. Our research shows that the linguistic steganalysis based on distribution of words is promising.

Word shift is a fundamental format based text steganography. It embeds secret information in text by shifting words slightly. Compared with study on steganography, research on its steganalysis is still in its infancy. In this paper, we present a blind steganalysis method to detect word shift in PDF document. Our method is to find features sensitive to word shift and use classifier to learn and remember feature differences between natural document and stego one. In order to design sensitive features, we propose two concepts "neighbor difference" and "environment equal", which reveal the spaces' statistical property. Then, we divided the PDF document into two types to make our method work efficiently. At last, we design a series of experiments to demonstrate performance of our method. The detection accuracy of our method can be up to 93.3%. Our initial results shows that our proposed concepts are very useful in text steganalysis and offer help for other related works.