Search

Cross-view geo-localization is to spot images of the same geographic target from different platforms, e.g., drone-view cameras and satellites. It is challenging in the large visual appearance changes caused by extreme viewpoint variations. Existing methods usually concentrate on mining the fine-grained feature of the geographic target in the image center, but underestimate the contextual information in neighbor areas. In this work, we argue that neighbor areas can be leveraged as auxiliary information, enriching discriminative clues for geolocalization. Specifically, we introduce a simple and effective deep neural network, called Local Pattern Network (LPN), to take advantage of contextual information in an end-to-end manner. Without using extra part estimators, LPN adopts a square-ring feature partition strategy, which provides the attention according to the distance to the image center. It eases the part matching and enables the part-wise representation learning. Owing to the square-ring partition design, the proposed LPN has good scalability to rotation variations and achieves competitive results on three prevailing benchmarks, i.e., University-1652, CVUSA and CVACT. Besides, we also show the proposed LPN can be easily embedded into other frameworks to further boost performance., accepted by TCSVT

With the rapid development of the Internet of things (IoT) technology, how to effectively predict the network security situation of the IoT has become particularly important. It is difficult to quantify the IoT network situation due to a large number of historical data dimensions, and there are also has the problem of low accuracy for IoT network security situation prediction with multi-peak changes. To solve the above problems, this paper proposed an adaptive IoT network security situation prediction model, which makes the IoT network security situation prediction accuracy higher. Firstly, the paper used the entropy correlation method to calculate the network security situation value sequence in each quantization period according to Alarm Frequency (AF), Alarm Criticality (AC), and Alarm Severity (AS). Then, the security situation values arranged in time series are fragmented through the sliding window mechanism, and then the adaptive cubic exponential smoothing method is used to initially generate the IoT network security situation prediction results. Finally, the paper built the time-varying weighted Markov chain to predict the error value and modify the initial predicted value based on the error state. The experimental results show that the model has a better fitting effect and higher prediction accuracy than other models, and this model’s determination coefficient is 0.811. Compared with the other two models, the sum of squared errors in this model is reduced by 78 %-82 %. The model can better reflect the changes in the IoT network security situation over a while.

Point-of-Interest (POI) recommendation, i.e., suggesting POIs that a user is likely to visit, is a key task to improve user experience in location based social networks (LBSNs). Existing models either focus on geographical influence without considering other factors such as social influence and temporal influence or rely on linear methods to combine different modeling factors, lacking a sophisticated and systematical way to learn representations for users and POIs for recommendation. To remedy these issues, in this work we propose GNN-POI, a generic POI recommendation framework that leverages Graph Neural Networks (GNNs), which demonstrate powerful modeling capacity to learn node representations from node information and topological structure to improve POI recommendation. Specifically, we construct a LBSN graph comprising of two types of nodes, i.e., user node and POI node. For a target user, her preference representation is learned by combining (1) representations of her social connection nodes and (2) representations of the visited POI nodes. For social connection nodes integration, in order to model the complicated and multifaceted social influence, an attention mechanism is applied to learn strengths of heterogeneous social relations; for location nodes integration, we utilize Bi-directional Long Short-Term Memory (Bi-LSTM) to model users’ sequential check-in behavior, taking into account geographical and temporal features. Extensive experiments conducted over three real LBSN datasets show that the proposed GNN based framework significantly outperforms the state-of-the-art POI recommendation models in terms of precision, recall and Normalized Discounted Cumulative Gain (NDCG).

This work presents a security analysis of the QUIC handshake protocol based on symbolic model checking. As a newly proposed secure transport protocol, the purpose of QUIC is to improve the transport performance of HTTPS traffic and enable rapid deployment and evolution of transport mechanisms. QUIC is currently in the IETF standardization process and will potentially carry a significant portion of Internet traffic in the emerging future. For a better understanding of the essential security properties, we have developed a formal model of the QUIC handshake protocol and perform a comprehensive formal security analysis by using two state-of-the-art model checking tools for cryptographic protocols, i.e., ProVeirf and Verifpal. Our analysis shows that ProVerif is generally more powerful than Verifpal in terms of verifying authentication properties. Moreover, both tools produce a counterexample to some security properties, which reveal a design flaw in the current protocol specification. Last but not least, we analyze the root causes of this counterexample and suggest a possible fix.

Incremental Learning is a particular form of machine learning that enables a model to be modified incrementally, when new data becomes available. In this way, the model can adapt to the new data without the lengthy and time-consuming process required for complete model re-training. However, existing incremental learning methods face two significant problems: 1) noise in the classification sample data, 2) poor accuracy of modern classification algorithms when applied to modern classification problems. In order to deal with these issues, this paper proposes an integrated classification model, known as a Pre-trained Truncated Gradient Confidence-weighted (Pt-TGCW) model. Since the pre-trained model can extract and transform image information into a feature vector, the integrated model also shows its advantages in the field of image classification. Experimental results on ten datasets demonstrate that the proposed method outperform the original counterparts.

Instance-correspondence (IC) data are potent resources for heterogeneous transfer learning (HeTL) due to the capability of bridging the source and the target domains at the instance-level. To this end, people tend to use machine-generated IC data, because manually establishing IC data is expensive and primitive. However, existing IC data machine generators are not perfect and always produce the data that are not of high quality, thus hampering the performance of domain adaption. In this paper, instead of improving the IC data generator, which might not be an optimal way, we accept the fact that data quality variation does exist but find a better way to use the data. Specifically, we propose a novel heterogeneous transfer learning method named Transfer Learning with Weighted Correspondence (TLWC), which utilizes IC data to adapt the source domain to the target domain. Rather than treating IC data equally, TLWC can assign solid weights to each IC data pair depending on the quality of the data. We conduct extensive experiments on HeTL datasets and the state-of-the-art results verify the effectiveness of TLWC.

The dense connection is a powerful technique to build wider and deeper convolution neural networks (CNNs) for handling several computer vision tasks. Despite the excellent performance, it consumes numerous parameters and produces a large weight model file. We studied the distribution of convolution layers and proposed a hard parameter sharing approach known as convolution pool (CP) for compressing dense-connection-based image restoration CNN models. CP is used to reallocate the parameters to specific convolution layers to ensure that some can be shared in different layers. We design a set of dense-connection-based baselines for three typical image restoration tasks, including image denoising, super-resolution, and JPEG deblocking, to validate the performance of the proposed method. Moreover, we comprehensively analyze the potential problems by introducing CP, including group convolution, dilated convolution, and modeling efficiency. Experimental results demonstrate that the proposed method can efficiently achieve an impressive compression rate with negligible performance reduction.

The accuracy of existing information system security situation assessment methods is affected by expert evaluation preferences. This paper proposes an Information System Security Situation Assessment Model (ISSSAM), which is based on the Modified Interval Matrix-Entropy Weight based Cloud (MIMEC). Based on the system security situation assessment index system, the interval number judgment matrix reflecting the relative importance of different indicators is modified to improve the objectivity of the indicator layer weight vector. Then, the entropy weight based cloud is used to quantify the criterion layer and the target layer security situation index, and the security level of the system is graded. The feasibility and effectiveness of this model are verified by the security situation assessment of the Departure Control System (DCS). Through the comparison and analysis of the evaluation results based on entropy weight coefficient method and traditional AHP method, it is shown that the model we proposed has good stability and reliability.

Gait, i.e., the movement pattern of human limbs during locomotion, is a promising biometric for the identification of persons. Despite significant improvement in gait recognition with deep learning, existing studies still neglect a more practical but challenging scenario -- unsupervised cross-domain gait recognition which aims to learn a model on a labeled dataset then adapts it to an unlabeled dataset. Due to the domain shift and class gap, directly applying a model trained on one source dataset to other target datasets usually obtains very poor results. Therefore, this paper proposes a Transferable Neighborhood Discovery (TraND) framework to bridge the domain gap for unsupervised cross-domain gait recognition. To learn effective prior knowledge for gait representation, we first adopt a backbone network pre-trained on the labeled source data in a supervised manner. Then we design an end-to-end trainable approach to automatically discover the confident neighborhoods of unlabeled samples in the latent space. During training, the class consistency indicator is adopted to select confident neighborhoods of samples based on their entropy measurements. Moreover, we explore a high-entropy-first neighbor selection strategy, which can effectively transfer prior knowledge to the target domain. Our method achieves state-of-the-art results on two public datasets, i.e., CASIA-B and OU-LP., Comment: Accepted by ISCAS 2021. 5 pages, 2 figures

To improve network security situation prediction accuracy, an adaptive network security situation prediction model based on alarm elements was proposed. Firstly, we used the entropy correlation method to generate the network security situation time series according to Alarm Frequency (AF), Alarm Criticality (AC) and Alarm Severity (AS). Then, the initial situation predicted value is calculated through sliding adaptive cubic exponential smoothing. Finally, based on the error state, we built the time-varying weighted Markov chain to predict the error value and modify the initial predicted value. The experimental results show that the network security situation prediction results of this model have a better fit with the real results than other models.

To improve the prediction accuracy and stability of neural networks with random weights (NNRWs), we propose a novel ensemble NNRWs (E-NNRW) in this paper, which initializes its base learners by different distributions to improve their diversity. The final prediction results of the E-NNRW model are determined by these base learners through a voting mechanism, which minimizes the specific "blind zone" of a single learner, thus achieving higher prediction accuracy and better stability. Taking the random vector functional link network (RVFL), one of the most representative algorithms in NNRWs, as an example, we fully evaluate the performance of the proposed algorithm on nine benchmark classification problems. Extensive experimental results fully demonstrate the effectiveness of our method.

Adversarial examples are vital to expose the vulnerability of machine learning models. Despite the success of the most popular substitution-based methods which substitutes some characters or words in the original examples, only substitution is insufficient to uncover all robustness issues of models. In this paper, we present AdvExpander, a method that crafts new adversarial examples by expanding text, which is complementary to previous substitution-based methods. We first utilize linguistic rules to determine which constituents to expand and what types of modifiers to expand with. We then expand each constituent by inserting an adversarial modifier searched from a CVAE-based generative model which is pre-trained on a large scale corpus. To search adversarial modifiers, we directly search adversarial latent codes in the latent space without tuning the pre-trained parameters. To ensure that our adversarial examples are label-preserving for text matching, we also constrain the modifications with a heuristic rule. Experiments on three classification tasks verify the effectiveness of AdvExpander and the validity of our adversarial examples. AdvExpander crafts a new type of adversarial examples by text expansion, thereby promising to reveal new robustness issues., Work in progress

When Person Re-identification is considered as a retrieval task, re-ranking becomes a critical part of improving the re-identification accuracy. Most of the existing re-ranking methods focus on k -nearest neighbors, which requires a lot of queries and memory. In this paper, we propose a Feature Relation Map based Similarity Evaluation (FRM-SE) model to tackle this problem. The Feature Relation Map is utilized to automatically mine the latent relation between the k -neighbors through convolution operation. The re-ranking distance is learned through the FRM-SE model with metric learning. Further, we optimize the existing re-ranking method to utilize the advantage of the FRM-SE model for maintaining a balance between accuracy and complexity.The proposed approach is validated on two benchmark datasets, Market1501 and CUHK03. Results show that our re-ranking method is superior to the state-of-the-art re-ranking methods. Furthermore, in the transfer learning setting, the model trained on either Market1501 or CUHK03 can achieve a comparable accuracy improvement on the DuekMTMC dataset, which validates the generalization of our SE model.

Most of the Zero-Shot Learning (ZSL) algorithms currently use pre-trained models as their feature extractors, which are usually trained on the ImageNet data set by using deep neural networks. The richness of the feature information embedded in the pre-trained models can help the ZSL model extract more useful features from its limited training samples. However, sometimes the difference between the training data set of the current ZSL task and the ImageNet data set is too large, which may lead to the use of pre-trained models has no obvious help or even negative impact on the performance of the ZSL model. To solve this problem, this paper proposes a biologically inspired feature enhancement framework for ZSL. Specifically, we design a dual-channel learning framework that uses auxiliary data sets to enhance the feature extractor of the ZSL model and propose a novel method to guide the selection of the auxiliary data sets based on the knowledge of biological taxonomy. Extensive experimental results show that our proposed method can effectively improve the generalization ability of the ZSL model and achieve state-of-the-art results on three benchmark ZSL tasks. We also explained the experimental phenomena through the way of feature visualization.

Online sequential extreme learning machine (OS-ELM) and its variants provide a promising way to solve data stream problems, but most of them do not take the timeliness of the problems into account, which may degrade the performance of the model. The main reason is that these algorithms are unable to adapt to the latest data accordingly when the distribution of the data stream changes. To mitigate this limitation, the forgetting factor is introduced into the relevant models, which is used to balance the relative importance of past data and new data when necessary. However, there is no efficient way to set the forgetting factor properly so far. In this paper, we have developed a novel updating strategy for setting the forgetting factor and proposed a dynamic forgetting factor based OS-ELM algorithm (DOS-ELM). In the sequential learning phase of DOS-ELM, the forgetting factor can be adjusted dynamically according to the change degree of the model accuracy in each learning epoch. This updating process does not require setting any parameters artificially and thus greatly improves the flexibility of the model. The experimental results on ten classification problems, five regression problems, one time-series problem show that DOS-ELM can deal well with both stationary and non-stationary data stream problems. In addition, we have extended DOS-ELM to an online deep model named ML-DOS-ELM, which can handle more complex tasks such as the face recognition problem and the handwritten digit recognition problem. Our experimental evaluations show that both DOS-ELM and ML-DOS-ELM can achieve higher prediction accuracy compared to the other similar algorithms.

Existing assessment models have inconvenient statistics of fuzzy factors and lack of intuitive graphical implementation methods. This paper proposes an information system security situation assessment model based on fuzzy Petri nets (ISSSAF). Firstly, a general hierarchical assessment index system is established. According to the assessment system, a fuzzy Petri net (FPN) model is built and a fuzzy inference algorithm is designed. Secondly, the credibility of each proposition is calculated by inference, and the security situation of the system is graded by grey assessment and inference algorithm. Finally, an assessment experiment of a domestic departure control system (DCS) is carried out. The experimental results verify the effectiveness of ISSSAF. Compared with the traditional methods, it is more objective and accurate, which can help the relevant personnel to formulate effective security protection strategies for the information system.

With the wide application of network technology, the Internet of Things (IoT) systems are facing the increasingly serious situation of network threats; the network threat situation assessment becomes an important approach to solve these problems. Aiming at the traditional methods based on data category tag that has high modeling cost and low efficiency in the network threat situation assessment, this paper proposes a network threat situation assessment model based on unsupervised learning for IoT. Firstly, we combine the encoder of variational autoencoder (VAE) and the discriminator of generative adversarial networks (GAN) to form the V-G network. Then, we obtain the reconstruction error of each layer network by training the network collection layer of the V-G network with normal network traffic. Besides, we conduct the reconstruction error learning by the 3-layer variational autoencoder of the output layer and calculate the abnormal threshold of the training. Moreover, we carry out the group threat testing with the test dataset containing abnormal network traffic and calculate the threat probability of each test group. Finally, we obtain the threat situation value (TSV) according to the threat probability and the threat impact. The simulation results show that, compared with the other methods, this proposed method can evaluate the overall situation of network security threat more intuitively and has a stronger characterization ability for network threats.

In recent years, with the problem of network security is getting worse, the network threat situation assessment becomes an important approach to solve these problems. Aiming at the traditional methods based on data category tag that has high modeling cost, low efficiency, and a long period in the network threat situation assessment, this paper proposes a Variational-Generative (V-G) network assessment method. Firstly, we design the V-G network which is composed of VAE’s encoder and GAN’s discriminator and obtain the reconstruction error of each layer network by training the network collection layer of the V-G network with normal network traffic. Then, conduct the reconstruction error learning by the 3-layer variational autoencoder of the output layer and calculate the abnormal threshold of the training. Moreover, carry out the group threat testing with the test dataset contains abnormal network traffic and calculate the threat probability of each test group. Finally, obtain the Threat Situation Value (TSV) according to the threat probability and the threat impact. The simulation results show that compared with the other methods, this proposed method can evaluate the overall situation of network security threat more intuitively and has a stronger characterization ability for network threats.

In recent years, deep neural networks have been widely applied on recommender systems. Most research efforts are put on modeling the side information such as textual information, contextual information and social network information, but the core part, i.e., interaction relationship between users and items are relatively less explored by neural networks, in particular, when multiple types of implicit feedbacks, e.g., click, browsing, add-to-cart, etc. are available in the system. In this paper, we propose an end-to-end learning framework, which systematically and comprehensively models multiple implicit feedback between users and items. Firstly, for each type of implicit feedback, we apply matrix factorization and Multi-Layer Perception (MLP) to capture both linearity and nonlinearity of user-item interactions. Then we fuse the effects of multiple implicit feedback through neural networks to boost the quality of recommendation. Experiments on Alibaba real production dataset with over two million interactions demonstrate the effectiveness of proposed approaches, which achieve superior performance compared with state-of-the-art methods.

Online learning for imbalanced streaming data is an important and challenging problem for many classification tasks in the machine learning research field. Traditional online learning algorithms are mainly focused on classification tasks with balanced data, and with little consideration about the characteristics of imbalanced streaming data. In this paper, we propose a novel online learning algorithm called Truncated Gradient Confidence-Weighted (TGCW), which integrate the truncated gradient algorithm with the confidence weighted algorithm together to improve the feature selection ability while reducing the dimensions of imbalanced streaming data effectively. We study a number of classification tasks with various imbalance data ratio including the pedestrian detection application and compare the performance of the TGCW algorithm with traditional online learning algorithms, and empirical results show that the TGCW algorithm can achieve better performance consistently than other baseline approaches.

Natural language generation (NLG) is an essential component of task-oriented dialogue systems. Despite the recent success of neural approaches for NLG, they are typically developed for particular domains with rich annotated training examples. In this paper, we study NLG in a low-resource setting to generate sentences in new scenarios with handful training examples. We formulate the problem from a meta-learning perspective, and propose a generalized optimization-based approach (Meta-NLG) based on the well-recognized model-agnostic meta-learning (MAML) algorithm. Meta-NLG defines a set of meta tasks, and directly incorporates the objective of adapting to new low-resource NLG tasks into the meta-learning optimization process. Extensive experiments are conducted on a large multi-domain dataset (MultiWoz) with diverse linguistic variations. We show that Meta-NLG significantly outperforms other training procedures in various low-resource configurations. We analyze the results, and demonstrate that Meta-NLG adapts extremely fast and well to low-resource situations., Accepted as a full paper at IJCAI 2019

Contexts have been proven to be an important source of information that can significantly improve the performance of collaborative filtering (CF), e.g., for recommendation. Most context-aware approaches that are basing on latent factor models assume that contexts share the same latent space with users and items. However such a strategy does not always make sense, e.g., the influence of contextual information may be overestimated. In this paper, we propose a generic framework to learn context-aware latent representations for context-aware collaborative filtering without imposing contexts into latent space of users and items. Contextual contents are combined via a function to produce the contextual influence factor, which is then combined with each latent factor to derive latent representations. We instantiate the generic framework using biased Matrix Factorization for rating prediction task and Bayesian Personalized Ranking (BPR) for item recommendation tasks. Stochastic Gradient Descent (SGD) based optimization procedures are developed to fit the two context-aware models by jointly learning the weight of each context and latent factors of users and items. Experiments conducted on three real-world datasets demonstrate that our context-aware CF model significantly outperforms not only the base models but also the representative context-aware models.

This paper proposed a multi-keyword ciphertext search, based on an improved-quality hierarchical clustering (MCS-IQHC) method. MCS-IQHC is a novel technique, which is tailored to work with encrypted data. It has improved search accuracy and can self-adapt when performing multi-keyword ciphertext searches on privacy-protected sensor network cloud platforms. Document vectors are first generated by combining the term frequency-inverse document frequency (TF-IDF) weight factor and the vector space model (VSM). The improved quality hierarchical clustering (IQHC) algorithm then generates document vectors, document indices, and cluster indices, which are encrypted via the k-nearest neighbor algorithm (KNN). MCS-IQHC then returns the top-k search result. A series of experiments proved that the proposed method had better searching efficiency and accuracy in high-privacy sensor cloud network environments, compared to other state-of-the-art methods.

How to evaluate the importance of nodes in networks and detect the centrality has become a vital problem in improving the efficiency of telecommunication and making a disease immunity strategy. We consider the mechanisms of real networks, and define a cost function to describe different hierarchies of networks to measure node importance. This method takes up a node’s regional influence as well as its global influence to evaluate its importance. The results of simulation prove that this method is proper to describe effectively and detect node discrepancies in a network.

In order to protect the network and evaluate the network security risks automatically, a new multi-agents risk assessment model based on attack graph (MRAMBAG) is presented. First, a network risk assessment model with master-slave agents is established, especially the functional architecture of master-slave agents and the risk association relation analysis process are designed. Then, the attack path and the attack graph are constructed by using the Attract Graph Building algorithm with the input of the dynamic data information collected by components. Finally, risk indexes of attack path, components, hosts, vulnerabilities and association risk index of network nodes are calculated successively and consequently the security risk quantitative index of target networks are obtained. The experimental results demonstrate that the MRAMBAG is a more feasible and effective way for evaluate the network security risk.

Emergencies have been bringing huge damage, even worse when emergency event chain reaction is formed. Different emergency evolution models have been studied in the systematic perspective to find out common characters and correlation among four various emergency events. We propose an academic framework with four hierarchies including single event, event chain, event network, and event supernetwork and discuss how to response to the emergency in different hierarchies. We also prove its efficiency by discussing a real case. This paper can provide useful guidelines to prevent broadcasting of disaster and macro precaution.

Collaborative filtering (CF) is a successful approach for building online recommender systems. The fundamental process of the CF approach is to predict how a user would like to rate a given item based on the ratings of some nearest-neighbor users (user-based CF) or nearest-neighbor items (item-based CF). In the user-based CF approach, for example, the conventional prediction procedure is to find some nearest-neighbor users of the active user who have rated the given item, and then aggregate their rating information to predict the rating for the given item. In reality, due to the data sparseness, we have observed that a large proportion of users are filtered out because they don't rate the given item, even though they are very close to the active user. In this paper we present a recursive prediction algorithm, which allows those nearest-neighbor users to join the prediction process even if they have not rated the given item. In our approach, if a required rating value is not provided explicitly by the user, we predict it recursively and then integrate it into the prediction process. We study various strategies of selecting nearest-neighbor users for this recursive process. Our experiments show that the recursive prediction algorithm is a promising technique for improving the prediction accuracy for collaborative filtering recommender systems.

Preference-based search (PBS) is a popular approach for helping consumers find their desired items from online catalogs. Currently most PBS tools generate search results by a certain set of criteria based on preferences elicited from the current user during the interaction session. Due to the incompleteness and uncertainty of the user's preferences, the search results are often inaccurate and may contain items that the user has no desire to select. In this paper we develop an efficient Bayesian filter based on a group of users' past choice behavior and use it to refine the search results by filtering out items which are unlikely to be selected by the user. Our preliminary experiment shows that our approach is highly promising in generating more accurate search results and saving user's interaction effort. Copyright 2007 ACM.

In a service-oriented e-commerce environment, it is a crucial task to help consumers choose desired products efficiently from a huge amount of dynamically configured product candidates. Decision agents can be designed to provide interactive decision aids for end-users by eliciting their preferences and then recommending matching products. In reality the users’ preferences may keep changing along with the dynamic decision environment and may not be fully satisfied. As a result, the decision agent is required to be agile: it should allow decision making with an incomplete user’s preference model and should afford users to add, retract or revise their various preferences with little effort. In this paper we propose the general design of an agile decision agent to meet this need. We model users’ preferences with the soft constraint technique and elicit them by the example critiquing interaction paradigm.