Byeongok Choi and Chae Y. Lee, KAIST, Korea.
As video traffic increases with plentiful multimedia services and the proliferation of mobile devices such as smartphones, stream mining to extract valuable information out of multimedia big data is garnering attention. By applying cloud computing to stream mining, resource-scarce mobile devices can offload the workloads of heavy applications to a remote cloud. However, resource provisioning for task scheduling is an inherent challenge of stream mining in cloud computing. In this paper we consider problem of resource provisioning and bit rate scaling for multimedia big data processing. We aim to minimize the virtual machine (VM) leasing cost and the classification error cost while satisfying the deadline constraints of workloads which is formulated as a mixed integer nonlinear programming. Deadline based task scheduling and bit rate scaling are developed to find near optimal solution of the NP-hard problem. The upper and lower bounds of the required number of VMs are obtained for infeasible and feasible schedules respectively. Scaling down the highest bit rate first in the bit rate set of a workload is suggested to guarantee the minimum increase of error cost. Our simulation results show the efficiency of bit rate scaling in task scheduling. 5-10% cost reduction is achieved by bit rate scaling in a cloud computing environment.
Stream mining, multimedia big data, cloud computing, task scheduling, bit rate scaling.
Ana Emilia Figueiredo de Oliveira, Katherine Marjorie Mendonça de Assis, Camila Santos de Castro e Lima, Carla Galvão Spinillo, Elza Bernardes Monier, Maria de Fatima Oliveira Gatinho, and Marcelo Henrique Monier Alves Junior, Federal University of Maranhão, Brazil.
The use of mobile technologies in the educational process has been generating positive results. In this context, institutions that work with Distance Education (DE) need to update and adapt their processes to these innovations. Considering this trend, UNA-SUS/UFMA has built, in partnership with the Saite Group, a virtual library that enables fast and free access to its contents using mobile devices. With the expansion of the tool's use, the institution invested in a new version to provide a better experience to its users. This paper aims to perform a detailing of the Saite Store new version, showing its operation and the technological implementations carried out. Finally, the importance of performing updates in the application is highlighted, considering its great potential for use by more than twenty thousand users.
Distance education, virtual library, e-learning, mobile.
Ramtin Ranji, Ali Mohammed Mansoor and Sakhr Ahmed Hatem, University of Malaya, Malaysia.
Mobile and Internet of Things (IoT) devices are limited in computation capabilities. Device to device (D2D) collaboration and mobile edge computing (MEC) proposed to overcome this problem. Several cooperative schemes designed to offload the computational-intensive tasks to a more powerful nearby device or MEC. However, most of them designed on the unrealistic hypothesis that devices are willing to cooperate. In this work, we first present our offloading scheme where MEC decides about proper offloading destination based on the energy efficiency and deadline of the tasks. Then we propose a novel incentive mechanism for mobile edge computing, based on the blockchain technology. In our scheme, devices must register to the blockchain and receive initial tokens. Since then, they can trade their tokens in order to use the network service and performing offloaded tasks of their neighbours. We leverage MEC capabilities to perform the computational-intensive tasks of devices.
mobile edge computing, blockchain, device to device communication, energy efficiency, task offloading.
Shurug Alkhalifa and Hind AlMahmoud, King Saud University, Saudi Arabia
Due to the advances in social media channels that help people interact and express themselves, more and more users are using these channels on daily basis. Emoji’s gained a lot of popularity since they help users express their emotions and state of mind. In this paper, we proposed five formulas that help in determining the tweet’s sentiment based on both text and emoji parts. These five formulas are different between each other based on the given weight for both Emoji and text. Moreover, we compared the accuracy of tweet’s sentiment using the five formulae versus using text only to decide the sentiment of the tweet. Our results show that Emoji’s have a strong effect on conveying the tweet's sentiment when compared to human assessment of the tweet sentiment. And hence, Emoji should be incorporated when determining the sentiment of the tweet because it significantly improves the accuracy of it.
Emoji, Sentiment Analysis, Social Media, Twitter.
Qiaoqiao Li, Guoyue Chen, Xingguo Zhang, Kazuki Saruta and Yuki Terata, Akita Prefectural University, Japan
Medical image fusion plays an important role in clinical application such as image-guided radiotherapy and surgery, and treatment planning. The main purpose of the medical image fusion is to fuse different multi-modal images, such as MRI and CT, into a single image. In this paper, a novel fusion method is proposed based on a fast structure-preserving filter for medical image MRI and CT of a brain. The fast structure preserving filter is a novel double weighted average image filter (SGF) which enables to smooth out high-contrast detail and textures while preserving major image structures very well. The workflow of the proposed method is as follows: first, the detail layers of two source images are obtained by using the structure-preserving filter. Second, compute the weights of each source image by calculating from the detail layer with the help of image statistics. Finally, fuse source images by weighted average using the computed weights. Experimental results show that the proposed method is superior to the existing medical image fusion method in terms of subjective evaluation and objective evaluation.
Multimodal image fusion, structure-preserving filter, weighted average.
Gálvez Arias, Pierina Xiomara Guzmán Ramos, Pedro Jesús Chipana Vila, Luis Antonio , Trigoso Valeriano and Carlos Alberto, ESAN University, Peru
The analysis on the impact generated by the companies'publications on social networks implies a great human ef ort, since the audience expresses their opinions through comments or responses, these must be processed one by one. This ef ort involves allocating resources to read each comment and extract characteristics that make possible to determine if the comments are, for example, positive reactions or negative reactions. In our paper we propose the use of Artificial Intelligence tools to simplify the processes of analysis of publications in social networks making possible the generation of objective reports, that are more ef ective than manual procedures and that their creation demands less resources; generating savings of time and money. In the solution we extract the answers to a publication on Twitter, which are called Tweets using web Scraping techniques and we generate a characteristic vector for each of the responses using Word2Vec, also we propose another vectorization method called Bag of Embedding Words (BoEW), but unlike Word2Vec, it treats the comment as the minimum unit of analysis. Then, to be able to analyze feelings, we use statistical classification models to determine if the comment is positive or negative. Finally, the impact that the publication has had in general is known and this solution will allow to evaluate the ef ectiveness of the publications, in an automated way and it will help the companies to make better decisions about how to interact with the audience through the chosen social network.
Amine Ganibardi and Chérif Arab Ali, University of Vincennes in Saint-Denis, France
This paper addresses the issue of Weblog Data cleaning within the scope of Web Usage Mining. Weblog data are information on end-user clicks and underlying user-agent hits recorded by webservers. Since Web Usage Mining is interested in end-user clicks, user-agent hits are referred to as noise to be cleaned before mining. The most referenced cleaning methods are the conventional and advanced cleaning. They are content-centric filtering heuristics based on the requested resource attribute of weblog databases. These cleaning methods are limited in terms of relevancy, workability and cost constraints, within the context of dynamic and responsive web. In order to overcome these constraints, this contribution introduces a clustering-based cleaning method focused on the genetic features of the logging structure. The introduced method mines clicks from hits on the basis of their underlying genetics features and statistical properties. The genetics clustering-based cleaning experimentation demonstrates significant advantages compared to the content-centric methods.
Web Usage Mining, Web Usage Data Preprocessing, Weblog Data Cleaning, Genetics Clustering, Genetics Clustering-based Cleaning.
Mosima Anna Masethe and Sunday Ojo, Tshwane University of Technology, South Africa
The advent of Educational Data Mining techniques presents an opportunity for educational management decision makers in Educational Institutions to improve on the accuracy of predicting student academic performance. However, the diversity cum complexity of different educational data mining techniques pose a huge challenge resulting in uncertainty in making educational management decision.. The velocity and volume of educational data mining techniques make the challenge intractable. The RS approach addresses the need for a decision support system to guide process of appropriate EDM technique. JCOLIBRI framework was used in building CBR-RS, which allows an interface for a non-expert user to define a query based on the problem domain. The CBR-RS was evaluated using F1 measure metrics to measure quality of known algorithms used. The research result on evaluation metrics obtainable shows three different classification algorithms named –IBK, J48 and Naïve Bayes, where Naïve Bayes outperforms other algorithms, giving best results of the F1-measure as 26.7%.
Ontology, Educational Data Mining, Recommender Systems, Case Based Reasoning.
Zachary Yannes and Gary Tyson, Florida State University, USA
Since the initial release in 2008, the Android operating system has fundamentally changed how applications are developed, installed, and executed. Developing benchmark suites for such an operating system can be challenging when the system architecture changes so rapidly. We introduce our analysis framework, Dexplore, which uses new techniques to evaluate the Android framework against a well established benchmark suite, SPECjvm2008. We collect open source applications from FDroid.org and analyze for commonly invoked library methods to which should be included in every effective Android benchmark. In this paper, we will address four primary points. First, Android applications rely heavily on third-party libraries which produce most of the applications functionality. Second, traditional microbenchmark suites such as SPECjvm2008 do not properly represent common Android applications. Third, application benchmark suites are inefficient and redundant representations of modern Android applications. Finally, we use our framework to identify and discuss commonly invoked core and third-party libraries.
Android Development, Effective Benchmarks, Library-Reliance, Common Android Libraries, Developer Constraints.
JunSik Kim1 and SeungWook Hong2 and Suhyun Park3, 1Dongseo University,Korea, 2Sasang-gu,Korea and 3Dongseo University,Korea.
The virtual Automatic Identification System (AIS) generation management system is a system for analyzing waterway risk by generating virtual AIS data which contains location information of a specific area. The system uses the data as an input data of the IALA Waterway Risk Assessment (IWRAP).
AIS, Shipwreck, waterway risk assessment.
Mohammad R. Hassan1, Nidal M. Turab2, Khalil H. Al-Shqeerat3 and Saleh El-Omar1,1,2Al-Ahliyya Amman University Jordan and 3 Qassim University, Saudi Arabia.
Group key management is an important functional building block for any secure distributed communication environment. Key distribution techniques are key factors in secure communication in grid computing. Once the secure key management is performed, messages can be securely exchanged among the grid entities. In any group communication systems supporting GRID and cloud-based applications, the communication parties need to communicate with each other as members of a group in a secure manner. Several protocols have been proposed to support secure group key management. This paper presents a new secure password-based group key management protocol for Grid computing environment, which is composed of two dynamic servicing tiers: the grid application that requests grid services, and the grid services that work on behalf of the user.
Group Key, Grid computing, password-based, Grid Security.
Toan Nguyen Mau and Yasushi Inoguchi, JAIST, Japan.
Locality-sensitive hashing(LSH) is a significant algorithm for big-data hashing. The original LSH uses a static hash-table as a reduce mapping for the data. Which make LSH challenging to apply on real-time information retrieval system. The database of a realtime system needs to be scalably updated over time. In this research, we concentrate on increasing the accuracy, searching speed and throughput of the nearest neighbor searching problem on big dynamic database. The dynamic Locality-sensitive hashing(DLSH) is proposed for facing the static problem of original LSH. DLSH is targeted for deploying on main memory or GPGPU’s global memory, which can increase the throughput searching by parallel processing on multiple cores. We analyzed the efficiency of DLSH by building the big dataset of structured audio fingerprint and comparing the performance with original LSH. To achieve the dynamics, DLSH requires more memory space and takes slightly slower than the LSH. With DLSH’s advantages, it can be improved and fully applied in practice in a real-life information retrieval system.
Locality-sensitive hashing, Structured dataset, GPGPU Memory, Similarity Searching, Parallel Processing.
Hongsong Chen 1,2and Zhongchuan Fu3,1 University of Science and Technology Beijing, China. 2 Beijing Key Laboratory of Knowledge Engineering for Materials Science,China3Harbin Institute of Technology,China.
Routing protocol is the critical component of wireless sensor network(WSN). As the WSN may be attacked by all kinds of attack behaviours, the survivability of WSN is difficult to be evaluated quantitatively. Traditional survivability evaluation methods, such as stochastic process model, are difficult to precisely calculate the systematic survivability by experimental method. To precisely evaluate the systematic survivability ability under the collective action between external attack and internal security mechanism. A novel survivability entropy-based quantitative evaluation metric is proposed to calculate the systematic survivability ability of WSN routing protocol. Numerical analysis and simulation experiments are firstly combined to precisely calculate the novel survivability entropy metric.NS2(Network simulator) is used to simulate the DoS attack and security mechanism in WSN routing layer. Experimental results show that the novel survivability evaluation metric and approach can precisely evaluate the systematic survivability ability of routing protocol in wireless sensor network.
survivability entropy; quantitative evaluation; systematic survivability; wireless sensor network;routing protocol.
Hongsong Chen 1,2 and Zhongchuan Fu3,1 University of Science and Technology Beijing, China. 2 Beijing Key Laboratory of Knowledge Engineering for Materials Science, China 3 Harbin Institute of Technology, China.
Low-rate Denial of Service(LDoS) attack is a serious threat to the security of Wireless Sensor Netwokr(WSN).LDoS attack is a special DoS attack type in recent development. The LDoS attack traffic is simulator to normal traffic,so it is difficult to be detected by traditional detection method. Hilbert-Huang Transform(HHT) time-frequency analysis method can be used to analyze non-linear signal generated by LDoS attack. However, false IMF components are the challenge problems to analyze the LDoS attack. Correlation coefficient and Kolmogorov-Smirnov (KS) test approaches are united to recognize the false intrinsic mode function(IMF) components. Ad Hoc On-demand Multi-path Distance Vector (AOMDV) routing protocol and random Routing REQuest (RREQ) flooding attack are simulated to implement LDoS attack in wireless sensor network by Network Simulator (NS2). Experiment results show that the improved HHT methods are effective to detect LDoS attack.When the similarity probability of the IMF component to the original traffic is less than 0.4, and the correlation coefficient values of IMF to that of original traffic is less than 0.3, the IMF component is recognized as false IMF components, that will not be used to analyze LDoS attack. To our knowledge, this is the first quantitative experiment research on LDoS attack detection in WSN.
HHT, Correlation coefficient, KS test, LDoS attack, WSN,intrusion detection.
Rudy Agus Gemilang Gultom and Tatan Kustana, Indonesia Defense University, Indonesia.
This paper proposes a network security framework concept, so called the Six-Ware Network Security Framework (SWNSF). The SWNSF aim is to increase a Local Area Network (LAN) security readiness or awareness in a network security environment. This SWNSF proposal is proposed in order to enhance an organization’s network security environment based on cyber protect simulation experiences. Strategic thoughts can be implemented during cyber protect simulation exercises. Brilliant ideas in simulating an network security network environment become good lesson learned. The implementation for proper security strategy could secure an organization LAN from various threats, attacks and vulnerabilities in concrete and abstract levels. Countermeasure strategy, which is implemented in this simulation exercise is presented as well. At the end of this paper, an initial network security framework proposal, so called the Six-Ware Network Security Framework has been introduced.
Network security environment; cyber protect simulation; cyber threats, attacks and vulnerabilities; countermeasures strategy, LAN, SWNSF framework.
Malena Cecilia Castro Caro1, Luz Enith Márquez Cantillo2, José Duván Márquez Díaz3, 1Institución Universitaria ITSA, Colombia, 2Institución Universitaria ITSA, Colombia and 3Universidad del Norte, Colombia
This paper explains some results from the research project "Comparative statistical analysis of the performance of data network for different algorithms for quality of service, routing protocols, and equipment technologies", which a set of testwere implementedto measure parameters such as: packet loss, delay and jitter using static routing and different packet scheduling algorithms (CQ, PQ and WFQ), different services (voice, data and video) and connection speeds using Cisco 2800 Routers and traffic generator D-ITG in conjunction with the NTP synchronizer. Based on the results of those tests a comparative analysis about sensitive parameters to the Quality of Service was conducted to describe the behavior of the network for each service.
Routing protocols, scheduling algorithms, equipment technologies, quality of service, delay, jitter, packet loss.
Masum Celil Olgun, Kadir Metin Akpolat, Zakir Baytar, Ozgur Koray Sahingoz, Istanbul Kultur University, Turkey
Lane detection and vehicle tracking algorithms are important keystones for many autonomous vehicle systems. The navigational process of those systems is mainly focused on the output of detection and classification algorithms. However, these algorithms need more pre-processing time and computational effort before executing in a real environment. They are also affected by environmental conditions and must regularly be improved. In this paper, to solve the autonomous control problem of vehicles with lane and vehicle tracking properties, we used Deep Learning techniques on vision get from cameras on the vehicle. With the nature of deep learning algorithm, the proposed system can handle complex image problems. For the experimental environment, we used an autonomous Remote Control (RC) car that tracks lane line, follows other RC car and detect objects to stop in necessary conditions. For that, one of the primary purposes is image-based lane tracking methodology by using learning algorithms. Data augmentation is applied to create diversity for the dataset. Application in this methodology has been discussed. For lane tracking Convolutional Neural Network architecture which is based on NVIDIA’s PilotNet is preferred. For detecting objects and vehicles, the system is trained on the faster region-based convolutional neural network (Faster R-CNN) to identify traffic light and stop sign are by Haar Cascade Classifier. All these learning models are trained on GPU to reduce training time. Experimental results showed that the proposed system gives an outstanding result to autonomously control vehicles for lane and vehicle tracking purposes by vision.
Controlling Autonomous Car, Deep Learning, Image Processing, Haar Cascade, Lane Tracking, Vehicle Tracking, Object Detection, GPU
Saide Işılay Baykal, Deniz Bulut, and Ozgur Koray Sahingoz, Istanbul Kultur University, Turkey
In recent years, Internet technologies are grown pervasively not only in information based web pages but also in online social networking and online banking, which made people’s lives easier. As a result of this growth, computer networks encounter with lots of different security threats from all over the world. One of these serious threats is “phishing”, which aims to deceive their victims for getting their private information such as username, passwords, social security numbers, financial information, and credit card number by using fake e-mails, webpages or both. Detection of phishing attack is a challenging problem, because it is considered as a semantics-based attack, which focuses on users’ vulnerabilities, not networks’ vulnerabilities. Most of the anti- phishing tools mainly use the blacklist/whitelist methods; however, they fail to catch new phishing attacks and results a high false-positive rate. To overcome this deficiency, we aimed to use a machine learning based algorithms, Artificial Neural Networks(ANNs) and Deep Neural Networks(DNNs), for training the system and catch abnormal request by analysing the URL of web pages. We used a dataset which contains 37,175 phishing and 36,400 legitimate web pages to train the system. According to the experimental results, the proposed approaches has the accuracy in detection of phishing websites with the rate of 92 % and 96 % by the use of ANN and DNN approaches respectively.
Phishing Detection System, Artificial Neural Networks, Deep Neural Networks, Big Data, Machine Learning, Tensorflow, Feature Extraction
Shay Horovitz, Alon Ben-Lavi, Refael Auerbach, Bar Brownshtein,Chen Hamdani and Ortal Yona, College of Management Academic Studies, Israel
As modern applications and systems are growing fast and continuously changing, back-end services in general and database services in particular are being challenged with dynamic loads and differential query behaviour. The traditional best practice of designing database – creating fixed relational schemas prior to deployment - becomes irrelevant. While newer database technologies such as document based and columnar are more flexible, they perform better only under certain conditions that are hard to predict. Frequent manual modifications of database structures and technologies under production require expert skills, increase management costs and often ends up with sub-optimal performance. In this paper we propose AdaptaBase - a solution for performance optimization of database technologies in accordance with application query demands by using machine learning to model application query behavioural patterns and learning the optimal database technology per each behavioural pattern. Experiments present a reduction in query execution time of over 25% for the relational-columnar model selection, and over 30% for the relation-document based model selection.
Database, Cross-Technology, Machine Learning, Adaptive
Norzanah Abd Rahman1 and Zamali Tarmudi2, 1University Teknologi MARA (Sabah Branch), Malaysia and 2University Teknologi MARA (Johor Branch), Malaysia
Intuitionistic fuzzy set (IFS) is a generalization of the fuzzy set that characterized by the membership and non-membership function. It is proven that IFS improved the drawbacks in fuzzy set since it is designed to deal with the uncertainty aspects. In spite of this advantage, the selection of the ranking approach is still one of the fundamental issues in IFS operations. Thus, this paper intends to compare three ranking approaches of the trapezoidal intuitionistic fuzzy numbers (TrIFN). The ranking approaches involve are; expected value-based approach, centroid-based approach, and score function-based approach. To achieve the objective, one numerical example in prioritizing the alternatives using intuitionistic fuzzy multi-criteria decision making (IF-MCDM) are provided to illustrate the comparison of these ranking approaches. Based on the comparison it was found that the alternatives MCDM problems can be ranked easily in efficient and accurate manner.
Intuitionistic fuzzy set, trapezoidal intuitionistic fuzzy numbers, multi-criteria decision making, ranking approach
Cecilia Sullca, Carlos Molina, Carlos Rodríguez and Thais Fernández, Universidad ESAN, Perú
This paper explains how image processing techniques and Machine Learning algorithms were used, such as Support Vector Machine (SVM), Artificial Neural Networks (ANN) and Random Forest, and the use of Deep Learning using the Convolutional Neural Network (CNN) was used. ) to determine which was the best for the construction of a recognition model that detects whether a blueberry plant is being affected by a disease or pest, or is healthy. The images were processed with different filters such as medianBlur and gaussianblur for the elimination of noise, the addWeighted filter was used for the enhancement of details in the images.
The results of the model showed an adequate recognition index with 84% accuracy using Deep Learning, this model was able to classify whether the blueberry plant was being affected or not. The result of this work provides a solution to a constant problem in the agricultural sector.
Artificial Vision, Deep Learning, Machine Learning, Random Forest, Support Vector Machine.
Lingling Mao1, Zaibin Chang1 and Jingqian Wang,2 1 Xi’an Traffic Enginering University, China 2 and Shaanxi University of Science & Technology, China
In this paper, we investigate a pair of rough approximation operators on matroids through circuits. Firstly, an equivalence relation is induced by circuits of a matroid, and some representations of the pair of approximation operators with respect to the equivalence relation are presented. These representations are presented from a family of circuits and a graph induced by circuits, respectively. Secondly, we illustrate that the pair of approximation operators are not the existing operators in matroids. Moreover, we investigate some characteristics of this pair of rough approximation operators on matroids. These characteristics are presented mainly through circuits and some operations in matroids. In a word, these results show an interesting view to investigate the combination between rough sets and matroids.
Rough set; Matroid; Approximation operator; Circuit; Granular computing.
Fang-Yi Chang1, 2Chia-Wei Tsai, 3Shu-Wei Lin and 3 Po-Chun Kuo,1,3Institute For Information Industry, Taiwan and 2,3Southern Taiwan University of Science and Technology, Taiwan.
Clustering is an useful tool in the data analysis to discover the natural structure in the data. The technique separates given smart meter data set into several representative clusters for the convenience of energy management. Each cluster may has its own attributes, such as energy usage time and magnitude. These attributes can help the electrical operators to manage their electrical grids with goals of energy and cost reduction. In this paper, we use principle component analysis and K-means as dimensional reduction and the reference clustering algorithm, respectively, and several choices must be considered: the number of cluster, the number of the leading principle components, and whether use normalized principle analysis schema or not. To answer these issues simultaneously, we use the stability scores as measured by dot similarity and confusion matrix as our evaluation decision. The advantage is that it is useful for comparing the performance under different decisions, and thus provides us to make these choices simultaneously.
Smart meter; Unsupervised; Nonparameter; Clustering; PCA; Stability; Smart Grid; Value-Add Electricity Services; Energy Saving; Energy management.
Kostandina Veljanovska, University St. Kliment Ohridski, Republic of Macedonia
The aim of this paper is to analyse performance of two machine learning algorithms, Naive Bayes and k-Nearest Neighbor in travel mode selection. The problem of choosing the car or public transportation for traveling from home to work was selected. The aim of this research is to help the local government in reducing air pollution by making influence of choosing the way of travel for the citizens. This way there are possibilities for environmental pollution reduction, fuel consumption reduction and for improving air quality. The results are promising for various dimensions of the cities since the machine learning algorithms are used and the proposed model is capable of learning from presented data even if they are not precise.
Soft Computing, Artificial Intelligence, Machine Learning, k-Nearest Neighbor algorithm, Naive Bayes algorithm
Karima Hadj-Rabah, Faiza Hocine, AssiaKourgli and AichoucheBelhadj-Aissa, University of Sciences and Technology Houari Boumediene (USTHB), Algeria
Given its efficiency and its robustness in separating the different scatterers present in the same resolution cell, SAR tomography (TomoSAR) has become an important tool for the reflectivity reconstruction of the observed complex structures scenes by exploiting multi-dimensional data. By its principle, TomoSAR reduces geometric distortions especially the layover phenomenon in radar scenes, and thus reconstruct the 3D profile of each azimuth-range pixel. In this paper, we present the results and the comparative study of six tomographic reconstruction methods that we have implemented. The analysis is performed with respect to the separability and location of scatterers by each method, supplemented by the proposal of a quantitative analysis using metrics (accuracy and completeness) to evaluate the robustness of each method. The tests were applied on simulated data with TerraSAR-X sensor parameters.
SAR Tomography (TomoSAR), Reconstruction Algorithms, Accuracy& Completeness
Zheng Zhu, Aiping Li*, Rong Jiang, Yulu Qi, Dongyang Zhao, Yan Jia, National University of Defense Technology, China
At present, many fingerprint recognition techniques are applied to public infrastructures. Their targets are mainly for normal-sized fingerprints. However, with the rise of small-sized fingerprint sensors, the acquired partial fingerprints containing only part of information of the finger, which causes that many researcherschange their research directions to partial fingerprint recognition.This paper proposes a SIFT-based pseudo-splicing partial fingerprint recognitionalgorithm. This algorithm uses the SIFT algorithm to pseudo-splice the input fingerprints during the fingerprint enrollment to increase the robustness of the fingerprint feature database. The comparisons of the accuracy of the recognitionamong this algorithm, the minutia-based fingerprint recognition algorithm and the fingerprint recognition algorithm based on image similarity,that showsthe first performs well.Moreover, this paper proposes an algorithm to evaluatethe quality of partial fingerprintby calculating the invalid blocks of fingerprint image. The result shows that the evaluation algorithm can effectively filterout low-quality fingerprints.
Fingerprint Recognition, SIFT, Partial Fingerprint,Pseudo-splicing
M.L.M. Fernando, W.S.M. Fernando, I.H. Dolawatta, N.T.C. Athukorala, Dr. Pradeepa Samarasinghe, Sri Lanka Institute of Information Technology New Kandy Road,Sri Lanka
Self-navigation and reading are two main challenges which visually impaired people all over the world encounter that limit them in social and educational capabilities. Even though many types of researches and products are presented today to assist the visually impaired it cannot be assisted both reading and navigation. As a result, visually impaired individuals have to use several devices/tools to overcome their day to day challenges.View Pods is a wearable device; a pair of sunglasses which a camera is setup aided with a Raspberry Pi and a headset connected to it that is enable the visually impaired person to capture the documents using the finger gestures based on a model trained on Convolutional Neural Networks and listen to the document content which is read using a TTS system where the text is extracted from the images of documents corrected and enhanced using image processing techniques. Also, it is provided with the ability to identify obstacles, get to know the distance to the obstacle and free space that they can walk through. View Pods will also provide them the guidance with rising and decreasing stairs and slopes giving the step count of stairs which are also based on image processing techniques and model trained on Convolutional Neural Networks.
Visual impairment, Image Processing, selfnavigation, Reading, Convolution Neural Network, Internet of things, Gesture recognition, Image enhancement
Dikshan N Shah1 and Dr. Harshad Bhadka2, 1Gujarat Technological University, India and 2C U Shah University, India
A morphological analyzer is a tool which makes syntactic analysis of a word and obtains root form of an inflected word form. The elementary step to identify a given sentence is the morphological analysis. We portray Morph analyzer for the Gujarati language. For a superior understanding of a language, word level, sentence level, context level and discourse level analysis has to be done. One of the tasks is the morphological level analyzing for various word-forms. In this paper, we have discussed a paradigm based approach for Morphological analysis for various POS tags. The algorithmic development gives an accuracy of analysis for Noun 84.50%, 81.50% for the verb and 80.50% for Adjectives.
Morphology, Paradigm based approach, Inflectional form, Morphological analysis
Dikshan N Shah1 and Dr. Harshad Bhadka2, 1Gujarat Technological University, India and 2C U Shah University, India
Indo-Aryan language is a vocabulary confection of Persian, Arabic, and Turkic elements with the grammar of the lingoes. The Bengali and Hindustani were the two substantial languages that formed from ‘Apabhramsa’ and other languages include Odia, Punjabi, Sindhi, Marathi, and Gujarati. Part of speech tagging is the practices of smearing each word of a corpus into its appropriate POS tag in terms of its definition and context. To develop various tags which esteem discrete part of speech is possibly the most emergent work of POS tagging. For European languages, the Penn Treebank tagset has manifest as the standard POS tagset, in contrast to Indian language perspective is much more impulsive. In this paper, we observed and decomposed of available tagset for POS tagging in Indian languages. We mainly focused on the operation of these tagset for tagging Indo-Aryan languages and proposed the expansion of the tagger for the Gujarati language which is a citizen language of Gujarat state. Furthermore, we have also mentioned the appraisal done. System attained the accuracy of 79.32%.
Indo-Aryan language, Part of Speech Tagging, Rule-Based approach
Primoz Podrzaj and Boris Kuster, University of Ljubljana, Slovenia.
Vision is probably the most important sense for human beings. As a consequence, our way of behaviour and thinking is also often based on visual information. When trying to perform complex information especially in situations where humans are involved, it is of great benet if some information can be obtained from images. This is the eld of image processing and computer vision. There are various libraries available for these tasks. Probably the best known one is OpenCV. It can also be used in Python programming language. Simple and more complex image processing algorithms are already available in the library. One of the more complex ones is face detection. In this paper it shown how face detection can be executed within Python with OpenCV library. This is the rst step needed in emotion recognition. When face is detected, we can determine the emotional state of the subject using a special purpose library.
Image processing, Python, OpenCV, face detection, emotion recognition.
Junnan Zhang and Hanyi Nie, NUDT, China.
Chronic wounds have a long recovery time, occur extensively, and are difficult to treat. They cause not only great suffering to many patients but also bring enormous work burden to hospitals and doctors. Therefore, an automated chronic wound detection method can efficiently assist doctors in diagnosis, or help patients with initial diagnosis, reduce the workload of doctors and the treatment costs of patients. In recent years, due to the rise of big data, machine learning methods have been applied to Image Identification, and the accuracy of the result has surpassed that of traditional methods. With the fully convolutional neural network proposed, image segmentation and target detection have also achieved excellent results. However, the accuracy of chronic wound image segmentation and identification is low due to the limitation of the deep convolution neural network. To solve the above problem, we propose a post-processing method based on fully connected CRFs with multi-layer score maps. The experiment results show that our method can be used to improve the accuracy of chronic wound image segmentation and identification.
Fully Connected CRFs, Chronic Wound Segmentation, Post-processing Method.
Igor Mishkovski1, Sanja Šćepanović2, Miroslav Mirchev1 and Sasho Gramatikov1, 1University Ss. Cyril and Methodius,Macedonia and 2Aalto University, Finland
Knowledge about the strength of the anti-virus engines (i.e. tools) to detect malware files on the Deep web is important for people and companies to devise proper security polices and to choose the proper tool in order to be more secure. In this study, using malware file set crawled from the Deep web we detect similarities and possible groupings between plethora of anti-virus tools (AVTs) that exist on the market. Moreover, using graph theory, data science and visualization we find which of the existing AVTs has greater advantage in detecting malware over the other AVTs, in a sense that the AVT detects many unique. Finally, we propose a solution, for the given malware set, what is the best strategy for a company to defend against malwares if it uses a multi-scanning approach.
Malware, Community detection, Anti-virus engines, data science, multi-scanning approach.
Julio Hernandez, Cinvestav, Mexico
The Web represents a valuable data source of information that is presented mainly as unstructured data. The extraction of structured and valuable information from sources such as the Web is an important challenge for the Semantic Web and Information Extraction areas, where elements representing real world objects and their relations need to be extracted from text and formally represented through RDF triples. Thus, extracting such information from the Web is manually unfeasible due to its large scale and heterogeneity of domains. In this sense, Open Information Extraction (OIE) is an independent domain task based on patterns to extract any kind of relation between real world things. Hence, one step further is to transform such relations into RDF triples. This paper proposes a method to represent relations obtained by an OIE approach into RDF triples. The method is based on the extraction of named entities, their relation, and contextual information from an input sentence and a set of defined rules that lead to map the extracted elements with resources from a Knowledge Base of the Semantic Web. The evaluation demonstrates promising results regarding the extraction and representation of information.
Open Information Extraction, Semantic Web, Named Entity Recognition, Named Entity Linking.
Damla Dinler and Ozgur Koray Sahingoz, Kultur University,Turkey
In recent years, the importance of the cyber-attack concept has started to increase in network environment. Therefore, companies need to take some certain measures to prevent these attacks and they employed some penetration tests (shortly pentests) which are performed to evaluate the security of the system by authorized attackers. Companies need to direct some employees who have a hacker point of view, to look for the vulnerabilities of their systems. These type employees are called as “pentester”. Penetration test is not only laborious but also time-consuming operation and pentesters use manually lots of software tools requires additional time for execution and follow-up. Therefore, generally, pentesters select several tools to make the test easier. These tools can find some vulnerabilities of the system; however, they can also miss some important points due to the lack of time or getting out of sight of the tester. The most important parts of the penetration test information gathering in which the targeted system is identified and the critical points where vulnerability can occur are detected. In this study, it is aimed to collect the manual scans of the pentester in a single software system to conduct them automatically, which leaves additional time for pentester to go a deeper analysis. After executing this type of test routine, pentester can also look back and see the collected information about the system, which are stored in the database of the system to compare with the previous tests and making some additional analysis. Additionally, we have surveyed information gathering tools in penetration test and propose a new-generation vulnerability management system for security professionals which automates the important tools of penetration testing via a web based solution. Experimental results show that the proposed system decreases the checking time of pentester and increase the efficiency of the testing process.
Penetration test, Information gathering, Nmap, Wfuzz, The Sublist3r, The Harvester, Whois
Hadjer Benkraouda, Ezedin Barka, and Khaled Shuaib,United Arab Emirates University,UAE
With the exponential growth in the digitalization of critical infrastructures such as nuclear plants and transmission and distribution grids, these systems have become more prone to coordinated cyberphysical attacks. One of the ways used to harden the security of these infrastructures is by utilizing UAVs for monitoring, surveillance and data collection. UAVs use data communication links to send the data collected to ground control stations (GCSs). The literature  suggests that there is a lack of research in the area of the cybersecurity of data communication from drones to GCSs. Therefore, this paper addresses this research gap and analyzes the vulnerabilities and attacks on the collected sensor data, mainly on: data availability, data integrity and data confidentiality, and will propose solutions for securing the drone’s data communication systems.
Information security, UAV Security, Critical Infrastructure Security.
Wai Weng Lo, Xu Yang and Yapeng Wang, Macao Polytechnic Institute, China
In this work, we applied a deep Convolutional Neural Network (CNN) with Xception model to perform malware image classification. The Xception model is a recent development and is a special CNN architecture that is more powerful with less over-fitting problems than the current popular CNN models such as VGG16. However only a few use cases of the Xception model can be found in literature, and it has never been used to solve this malware classification problem. The performance of our approach was compared with other methods including KNN, SVM, VGG16, and the complicated ensembled model. The experiments on two datasets (Malimg and Microsoft Malware Dataset) demonstrated that the Xception model achieved higher training and validation accuracy. Additionally we propose a novel method to combine the predictions between .bytes files and .asm files, showing a lower logloss can be achieved. The proposed method also proves that the Xception model can achieve a better performance than the VGG16 model. Although the champion on the Microsoft Malware Dataset achieved a bit lower logloss, our approach does not require any features engineering, making it more effective to adapt to any future evolution in malware, and very much less time consuming than the champion’s solution.
Malware classification, image classification, convolutional neural network (CNN), Xception, transfer learning
Reda El hachloufi, Imane Bachane and Habiba Chaoui ,ENSA Kenitra,Morocco
Digital forensics has gained much attention in the last few years, this is due mainly to the prevalence of cybercrime which often rely on malwares, and therefore the need of analyse and investigate the infected systems or devices. Many digital forensic investigation process models were established in order to meet this need, while there is some ne distinctions between them, all theses models share the same headlines; Acquisition, Identication, Evaluation, and Admission. However, the traditional forensic process suppose that the digital investi- gator or the incident responder have the convenient features of physical access, almost unlimited storage capacity, and sucient investigation time, which is not always the case, especially with the technical development and the appear- ance of external SOCs (security operation centers). A lot of frameworks have been designed with the intention of cover such problems, in particular by pro- viding remote and real-time forensics capabilities, even though the majority of them are commercials, there is much open-source tools which try to reach the same goal. The research focus of this paper is to analyse and compare the four most popular and widespread frameworks of remote digital investigation incident re- ponse; Google's GRR, Mozilla's MIG, Facebook's Osquery, and EnCase End- point Investigator. .
Remote forensics, Incident response, Real-time forensics, Digital investigation, Traditional forensics.
Michal Kedziora, Paulina Gawin, Michal Szczepanik and Ireneusz Jozwiak,Faculty of Computer Science and Management Wroclaw University of Science and Technology Wroclaw, Poland
This paper is focused on the issue of malware detection for Android mobile system by Reverse Engineering of java code. The characteristics of malicious software were identified based on a collected set of applications. Total number of 1958 applications where tested (including 996 malware apps). A unique set of features was chosen. Five classification algorithms (Random Forest, SVM, K-NN, Nave Bayes, Logistic Regression) and three attribute selection algorithms were examined in order to choose those that would provide the most effective malware detection.
Malware Detection, Android, Random Forest, SVM, K-NN, Naive Bayes, Logistic Regression.
Pranay Dugar and Rajesh Devaraddi M, International Institute of Information Technology, India
Cyber Physical Systems are increasingly becoming targets of different attacks as they become more and more common in today’s world. The attack behaviour is not completely predictable and hence a systematic approach to incorporate behaviour of the attackers has to be developed. This paper tries to model this very behaviour by proposing a mathematical model to capture the behaviour and then goes on to show how, after knowing this behaviour, can we safeguard the cyber physical systems against such attacks. We present a game theoretic approach to capture the behaviour of the attacker and how to defend against it and then do a Failure Mode Effects Analysis (FMEA) to come with actual possible damages to a system and which course of action is the best against which kind of attack based on the values generated through FMEA.
Game Theory, Cyber Physical Systems, FMEA, Automobile, Threat Analysis.
Manoj Athreya A, Ashwin A Kumar, Nagarajath S M and Gururaj H L, Vidyavardhaka College of Engineering, India
Blockchain is a growing list of records called blocks, that are linked using cryptography. It is a decentralized, distributed and an immutable ledger to store digital transactions. Its databases are managed using peer-topeer network where all the nodes in a network are equal and is the major concern in the types of network architecture. The decentralization of blockchain means that it won’t depend on a central point of control. With a lack of single authority, which makes the system equitable and more secure to validate transactions and to record data which makes it incorruptible. It uses consensus protocol which is a set of rules that describes the communication and transmission of data between nodes, to transact securely with other users without relying on the central authority. In this paper, we have proposed an innovative and efficient way of adopting pets using Blockchain technology, the adoption of pets and its payment method is made more easier and assures high security to data. Many approaches are made to the application but the approach using decentralized system is not used and it provides a new dimension to the application.
Blockchain, Decentralized, Distributed, peerto-peer network, Smart contracts.