Jianwen Xiang | Wuhan University of Technology (original) (raw)

Papers by Jianwen Xiang

With the rapid development of knowledge bases(KBs),question answering(QA)based on KBs has become ... more With the rapid development of knowledge bases(KBs),question answering(QA)based on KBs has become a hot research issue. In this paper,we propose two frameworks(i.e.,pipeline framework,an end-to-end framework)to focus answering single-relation factoid question. In both of two frameworks,we study the effect of context information on the quality of QA,such as the entity's notable type,out-degree. In the end-to-end framework,we combine char-level encoding and self-attention mechanisms,using weight sharing and multi-task strategies to enhance the accuracy of QA. Experimental results show that context information can get better results of simple QA whether it is the pipeline framework or the end-to-end framework. In addition,we find that the end-to-end framework achieves results competitive with state-of-the-art approaches in terms of accuracy and take much shorter time than them.

2017 IEEE International Conference on Software Quality, Reliability and Security Companion (QRS-C)

This paper proposes a knowledge creation model for graduate students and introduces a checklist o... more This paper proposes a knowledge creation model for graduate students and introduces a checklist of evaluation of research ability and research environment based on this model. The model is derived from three academic knowledge creation models, which were proposed with reference to a famous organizational knowledge creation model. A questionnaire survey, using the checklist at a graduate school of computer technology, proves the effectiveness of the proposed model and reveals some interesting facts, such as the great differences in the evaluation of research environment between teachers and students and in the self-evaluation of research ability between female students and male students.

2018 IEEE International Symposium on Software Reliability Engineering Workshops (ISSREW)

Android users are sometimes troubled by slow UI responses or even application/OS crashes. These i... more Android users are sometimes troubled by slow UI responses or even application/OS crashes. These issues are typically caused by software aging, a phenomenon characterized by progressive degradation of performance and functionality observed in long-running software systems. A practical and widely used approach to combat software aging is software rejuvenation, i.e. manual or scheduled restart of an application or a device. To reduce service outages, proactive rejuvenation is preferred, which strives to balance application downtime and performance level. However, traditional rejuvenation models cannot be directly applied to Android applications or system, as they do not address user experience, such as avoiding rejuvenation during high activity phases. In this work we exploit the fact that the usage time of mobile phones is typically fragmented in daily life, with periodic switches between active and sleep modes. We propose proactive rejuvenation strategies, which consider both usage and age factors. In particular, we model the usage behavior and aging process as individual Stochastic Petri-Nets, and then compose them into Continuous Time Markov Chains. We evaluate our models via numerical experiments and demonstrate the effectiveness and advantages of the proposed rejuvenation approach.

2016 IEEE International Conference on Software Quality, Reliability and Security Companion (QRS-C)

Knowledge management theories and tools have been developed rapidly in the field of business admi... more Knowledge management theories and tools have been developed rapidly in the field of business administration. However, since their application in the academic field is still very limited, this paper focuses on knowledge creation in the academic field. It is usually difficult for young graduate students to imagine their research process, which proceeds from a small idea to a research paper. Wierzbicki and Nakamori [1] introduced three knowledge creation models for three different academic fields, which are normative or hypothetical models to be verified in applications. This paper modified these models by interviewing graduate students and their supervisors, who are doing research in four fields in knowledge science. This study made clear the knowledge gaps between graduate students and their supervisors.

Neural Information Processing

Data mining has become very popular with the arrival of big data era, but it also raises privacy ... more Data mining has become very popular with the arrival of big data era, but it also raises privacy issues. Negative database (NDB) is a new type of data representation which stores the negative image of data and can protect privacy while supporting some basic data mining operations such as classification and clustering. However, the existing clustering algorithm upon NDB s is based on Hamming distance, when facing datasets which have many categories for each attribute, the encoded data will become very long and resulting in low computational efficiency. In this paper, we propose a privacy-preserving k-means clustering algorithm based on Euclidean distance upon NDB s. The main step of k-means algorithm is to calculate the distance between each record and cluster centers, in order to solve the problem of privacy disclosure in this step, we transform each record in database into an NDB and propose a method to estimate Euclidean distance from a binary string and an NDB. Our work opens up new ideas for data mining upon negative database.

2020 7th International Conference on Dependable Systems and Their Applications (DSA), 2020

Searching homologous vulnerabilities in firmware from different platforms is key for protecting t... more Searching homologous vulnerabilities in firmware from different platforms is key for protecting the Internet of Things (IoT) devices. Due to the hard work to acquire the source code, the existing works mainly focused on binary code, and many methods have been used to find vulnerabilities in binary code, like 1) numerical features are used to achieve efficient prefiltering, however, a large number of experiments showed that this method is which is lack of adaptability in some platforms; 2) structural features are used to achieve the accurate matching, but it will bring low efficiency. Based on the shortcomings of the existing methods, we proposed a lightweight method to search vulnerabilities of firmware in a cross-platform model. The main idea of this method is to exploit readable string literals inside functions combined with local call graph and use these string literals as a feature of each function in form of text, then MinHash LSH and MinHash LSH Forest algorithms are used to a...

2020 IEEE Symposium Series on Computational Intelligence (SSCI)

Iris recognition is one of the mainstream biometric recognition methods. Protecting iris data to ... more Iris recognition is one of the mainstream biometric recognition methods. Protecting iris data to prevent personal privacy leakage is significant to the popularity of iris recognition. Negative database is a new type of privacy protection technique. We proposed a promising method (called NDBIris) of iris template protection based on negative databases in previous work. However, its unlinkability is vulnerable under typical parameter settings (e.g. p1=0.8$,p_{2}$=0.14) and it does not protect the privacy of real-time iris data from users for recognition. This paper proposes an improved version called NDBIris-II to achieve better unlinkability and protect the real-time iris data. Specifically, a noninvertible transform using local sorting is performed before converting iris data into negative databases. Moreover, a method for estimating the similarity between iris data from negative databases is proposed to support effective iris recognition. Finally, an iris template in the form of negative database is generated for each iris data, and it is stored and used during iris recognition instead of raw iris data for privacy protection. Experimental results on iris database CASIA-IrisV3-Interval demonstrate that the proposed method could maintain recognition performance while achieving better unlinkability and protecting real-time iris data.

2017 IEEE International Conference on Image Processing (ICIP), 2017

In this paper, a novel multiscale ellipse descriptor (MED) method is proposed for shape descripti... more In this paper, a novel multiscale ellipse descriptor (MED) method is proposed for shape description and matching. MED extracts the competitive features of shape contour by measuring the spatial location relationship between contour sample points and topology structure information of segmented multiscale zone. This method not only has the discriminative ability to describe the global and local information, but also is robustness to various linear (rotation, scale and translation transforms) and non-linear (irregular intra-class deformation) transforms. Experimental results on two public available databases consistently demonstrate that, our proposed method is effective and efficient when compared with other state-of-the-art shape retrieval benchmarks (such as 9.72% higher and 64 times faster than popular IDSC method on leaf 100 dataset).

2019 20th Asia-Pacific Network Operations and Management Symposium (APNOMS), 2019

The most valuable resource in Wireless Sensors Networks (WSNs) is power consumption since it dire... more The most valuable resource in Wireless Sensors Networks (WSNs) is power consumption since it directly influences the lifetime of micro-sensors. So, there are several techniques have been proposed to solve this issue, such as routing protocol (energy-efficient medium access control (MAC)) and routing methods. Recently, compression techniques are used to reduce the size of transmitted data (less storage cost) and transmission time (less bandwidth needed) over wireless channels, which is the main power consumer in wireless sensor networks. In this paper, a new lossless compression scheme for Wireless Sensor Networks (WSNs) is proposed. It built on the nature of the data we dropped from a real-world deployment (Sensorscope project), which take into account the integer and the float part of the samples and compressed them separately using different methods. Our method overcomes many traditional algorithms, and it performed 90% and 79% better in term of compression ratio for Temperature a...

2018 Tenth International Conference on Advanced Computational Intelligence (ICACI), 2018

Software aging is a common experience in Android operating system, as the gradual performance deg... more Software aging is a common experience in Android operating system, as the gradual performance degradation is usually complained by the users. However, the mathematical modelling and detection of such experience is still an emerging issue due to the complexity and relatively young age of Android. This paper applies and compares three machine learning algorithms, namely decision tree, Support Vector Machine (SVM), and Deep Belief Network (DBn), for the detection of software aging in Android. In addition to the traditional aging indicator of launch time (LT), this paper also investigates the effectiveness of page fault (PF) and multiple labels (i.e., the combination of LT and PF). Experimental results show that the accuracy of DBN is comparable to decision tree and SVM when the data volume increases to 5000, which means DBN and other similar algorithms suitable for high dimensional and large data may also play a role in software aging. The results also reveal that PF is a little more s...

2018 5th International Conference on Dependable Systems and Their Applications (DSA), 2018

Dynamic fault tree (DFT) is an extension of traditional static fault tree, in which several dynam... more Dynamic fault tree (DFT) is an extension of traditional static fault tree, in which several dynamic gates are introduced to model sequential dependency between fault events. As for the quantitative analysis of DFT, sequential binary decision diagram (SBDD) has been proposed by incorporating sequential nodes into traditional binary decision diagram (BDD). With SBDD, a DFT can be reduced into a sum of disjoint products (SDP) of basic events and sequential nodes, which can avoid the space explosion problem and exponential complexity caused by traditional Markov chain and inclusion/exclusion based solutions, respectively. However, the SBDD is not developed base on a well-defined temporal or sequential algebra. Rather, it is developed based on a precedence operator with specific and very limited number of reduction rules. The applications of SBDD is thus restricted to the DFTs whose dynamic gates consist of inputs of only basic events or some specific events. In this paper, we present an...

2020 25th International Conference on Pattern Recognition (ICPR), 2021

Face anti-spoofing is becoming increasingly indispensable for face recognition systems, which are... more Face anti-spoofing is becoming increasingly indispensable for face recognition systems, which are vulnerable to various spoofing attacks performed using fake photos and videos. In this paper, a novel “LDN-TOP representation followed by ProCRC classification” pipeline for face anti-spoofing is proposed. We use local directional number pattern (LDN) with the derivative-Gaussian mask to capture detailed appearance information resisting illumination variations and noises, which can influence the texture pattern distribution. To further capture motion information, we extend LDN to a spatial-temporal variant named local directional number pattern from three orthogonal planes (LDN- TOP). The multi-scale LDN- TOP capturing complete information is extracted from color images to generate the feature vector with powerful representation capacity. Finally, the feature vector is fed into the probabilistic collaborative representation based classifier (ProCRC) for face anti-spoofing. Our method is...

Negative survey is a promising method which can protect personal privacy while collecting sensiti... more Negative survey is a promising method which can protect personal privacy while collecting sensitive data. Most of previous works focus on negative survey models with specific hypothesis, e.g., the probability of selecting negative categories follows the uniform distribution or Gaussian distribution. Moreover, as far as we know, negative survey is never conducted with manual selection in real world. In this paper, we carry out such a negative survey and find that the survey may not follow the previous hypothesis. And existing reconstruction methods like NStoPS and NStoPS-I perform poorly on the survey data. Therefore, we propose a method called NStoPS-MLE, which is based on the maximum likelihood estimation, for reconstructing useful information from the collected data. This method also uses background knowledge to enhance its performance. Experimental results show that our method can get more accurate aggregated results than previous methods.

Fault Tree Analysis is a traditional deductive safety analysis technique that is applied during t... more Fault Tree Analysis is a traditional deductive safety analysis technique that is applied during the system design stage. However, traditional fault trees often suffer from a lack of formal semantics to check the correctness or consistency of the descriptions. This is especially a problem in safety-critical system analysis. To overcome this limitation, we propose a novel formal fault tree construction method, which is different from traditional methods that focus on providing the formal semantics for the fault tree constructs after the informal fault tree has been built. In our method, the correctness of the fault tree is proved by the construction process itself, and the time relationships among different events are guaranteed by introducing temporal logic notations. Furthermore, by the stepwise deduction process, the hidden domain rules and inattentive design deficiencies can be discovered at an earlier stage, which helps the designers and domain experts effectively check and revis...

In a phased-mission system (PMS), an uncovered component fault may lead to a mission failure rega... more In a phased-mission system (PMS), an uncovered component fault may lead to a mission failure regardless of the status of other components, and the reliability can be analyzed with traditional imperfect fault coverage model (IFCM). The IFCM, however, only considers the coverage of faulty components. Recently, an irrelevancy coverage model (ICM) is proposed to cover both faulty components and irrelevant components, but the analysis is limited to normal non-phased mission systems. This paper first demonstrates that, the coverage of irrelevant components is also important in PMSs, as an initially relevant component could also become irrelevant later due to the failures of other components, and an uncovered fault of irrelevant component may threaten the whole mission as well. A method to analyze the reliability of PMS in ICM is proposed using sum of disjoint products (SDP) technique. Experimental results demonstrate not only the effectiveness of the proposed reliability analysis method, ...

Nowadays, privacy protection has become an important issue in data mining. k-means algorithm is o... more Nowadays, privacy protection has become an important issue in data mining. k-means algorithm is one of the most classical data mining algorithms, and it has been widely studied in the past decade. Negative database (NDB) is a new type of data representation which can protect privacy while supporting distance estimation, so it is promising to apply NDBs to privacy-preserving k-means clustering. Existing privacy-preserving k-means clustering algorithms based on NDBs could effectively protect data privacy, but their clustering performance has a non-negligible degradation. In this paper, we propose a new NDB generation algorithm (named QK-hidden algorithm), and based on this algorithm, we propose a privacy-preserving k-means algorithm. The proposed algorithm can control the accuracy of distance estimation in a fine-grained manner, and thus it can control the clustering results granularly. Experimental results demonstrate the proposed algorithm has better clustering performance than exis...

With the security breaches in Internet of Things devices, the detection of firmware vulnerability... more With the security breaches in Internet of Things devices, the detection of firmware vulnerability is more crucial than ever. Presently, many methods for firmware vulnerability detection have been proposed, but there are still some room for improvement on the detection precision. In this paper, we propose to use clonal selection algorithm to detect vulnerability functions in firmware. Firstly, we use the Relief algorithm to select the features that are more suitable for clonal selection algorithm. Then, we utilize principal component analysis algorithm to calculate the weights of the features. In the process of detection, we establish a set of specific detectors for each vulnerability function. In the end, we detect the vulnerability functions through these specific detectors. The experimental results show that the precision of our approach on detecting real vulnerabilities is competitive to the typical algorithm VDNS which is based on the neural network.