Intelligent Informatics

In this paper, a novel algorithm called KFCP(top K Frequent Correlated subgraph Pairs mining) was proposed to discover top-k frequent correlated subgraph pairs from graph databases, the algorithm was composed of two steps: co-occurrence frequency matrix construction and top-k frequent correlated subgraph pairs extraction.We use matrix to represent the frequency of all subgraph pairs and compute their Pearson’s correlation coefficient, then create a sorted list of subgraph pairs based on the absolute value of correlation coefficient. KFCP can find both positive and negative correlations without generating any candidate sets; the effectiveness of KFCP is assessed through our experiments with real-world datasets.

The main goal of Biomedical Natural Language Processing (BioNLP) is to capture biomedical phenomena from textual data by extracting relevant entities, information and relations between biomedical entities (i.e. proteins and genes). Most of the previous works focused on extracting binary relations among proteins. In recent years, the focus is shifted towards extracting more complex relations in the form of bio-molecular events that may include several entities or other relations. In this paper we propose a classifier ensemble based on an evolutionary approach, namely differential evolution that enables extraction, i.e. identification and classification of relatively complex bio-molecular events. The ensemble is built on the base classifiers, namely Support Vector Machine, nave-Bayes and IBk. Based on these individual classifiers, we generate 15 models by considering various subsets of features. We identify and implement a rich set of statistical and linguistic features that represent various morphological, syntactic and contextual information of the candidate bio-molecular trigger words. Evaluation on the BioNLP 2009 shared task datasets show the overall recall, precision and F-measure values of 42.76%, 49.21% and 45.76%, respectively for the three-fold cross validation. This is better than the best performing SVM based individual classifier by 4.10 F-measure points.

The present research proposes a paradigm for the clustering of data in which no prior knowledge about the number of clusters is required. Here shape based similarity is used as an index of similarity for clustering. The paper exploits the pattern identification prowess of Hidden Markov Model (HMM) and overcomes few of the problems associated with distance based clustering approaches. In the present research partitioning of data into clusters is done in two steps. In the first step HMM is used for finding the number of clusters then in the second step data is classified into the clusters according to their shape similarity. Experimental results on synthetic datasets and on the Iris dataset show that the proposed algorithm outperforms few commonly used clustering algorithm.

Associate classification is a scientific study that is being used by knowledge discovery and decision support system which integrates association rule discovery methods and classification to a model for prediction. An important advantage of these classification systems is that, using association rule mining they are able to examine several features at a time. Associative classifiers are especially fit to applications where the model may assist the domain experts in their decisions. Cardiovascular deceases are the number one cause of death globally. An estimated 17.3 million people died from CVD in 2008, representing 30% of all global deaths. India is at risk of more deaths due to CHD. Cardiovascular disease is becoming an increasingly important cause of death in Andhra Pradesh. Hence a decision support system is proposed for predicting heart disease of a patient. In this paper we propose a new Associate classification algorithm for predicting heart disease for Andhra Pradesh population. Experiments show that the accuracy of the resulting rule set is better when compared to existing systems. This approach is expected to help physicians to make accurate decisions.

In the present work, we explore an extensive applications of Gabor filter and

K

-means clustering algorithm in detection of text in an unconstrained complex background and regular images. The system is a comprehensive of four stages: In the first stage, combination of wavelet transforms and Gabor filter is applied to extract sharpened edges and textural features of a given input image. In the second stage, the resultant Gabor output image is grouped into three clusters to classify the background, foreground and the true text pixels using

K

-means clustering algorithm. In the third stage of the system, morphological operations are performed to obtain connected components, then after a concept of linked list approach is in turn used to build a true text line sequence. In the final stage, wavelet entropy is imposed on an each connected component sequence, in order to determine the true text region of an input image. Experiments are conducted on 101 video images and on standard ICDAR 2003 database. The proposed method is evaluated by testing the 101 video images as well with the ICDAR 2003 database. Experimental results show that the proposed method is able to detect a text of different size, complex background and contrast. Withal, the system performance outreaches the existing method in terms of detection accuracy.

The Discriminative Multinomial Naive Bayes classifier has been a center of attention in the field of text classification. In this study, we attempted to increase the prediction accuracy of the Discriminative Multinomial Naive Bayes by integrating global and local application of Discriminative Multinomial Naive Bayes classifier. We performed a large-scale comparison on benchmark datasets with other state-of-the-art algorithms and the proposed methodology gave better accuracy in most cases.

Protein structure prediction is an important component in understanding protein structures and functions. Accurate prediction of protein secondary structure helps in understanding protein folding. In many applications such as drug discovery it is required to predict the secondary structure of unknown proteins. In this paper we report our first attempt to secondary structure predication, and approach it as a sequence classification problem, where the task is equivalent to assigning a sequence of labels (i.e. helix, sheet, and coil) to the given protein sequence. We propose an ensemble technique that is based on two stochastic supervised machine learning algorithms, namely Maximum Entropy Markov Model (MEMM) and Conditional Random Field (CRF). We identify and implement a set of features that mostly deal with the contextual information. The proposed approach is evaluated with a benchmark dataset, and it yields encouraging performance to explore it further. We obtain the highest predictive accuracy of 61.26% and segment overlap score (SOV) of 52.30%.

With the impact of both the authors of scientific articles and also scientific publications depending upon citation count, citations play a decisive role in the ranking of both researchers and journals. In this paper, we propose a model to refine the citations in a research article verifying the authenticity of the citation. This model will help to eliminate author-centric and outlier citations thereby refining the citation count of research articles.

In this paper, we compare and analyze the novelty of a scientific paper (text document) of a specific domain. Our experiments utilize the standard Latent Dirichlet Allocation (LDA) topic modeling algorithm to filter the redundant documents and the Ontology of a specific domain which serves as the knowledge base for that domain, to generate cognitive maps for the documents. We report results based on the distance measure such as the Euclidean distance measure that analyses the divergence of the concepts between the documents.

This paper presents an intelligent framework that combines case-based reasoning (CBR), fuzzy logic and particle swarm optimization (PSO) to build an intelligent decision support model. CBR is a useful technique to support decision making (DM) by learning from past experiences. It solves a new problem by retrieving, reusing, and adapting past solutions to old problems that are closely similar to the current problem. In this paper, we combine fuzzy logic with case-based reasoning to identify useful cases that can support the DM. At the beginning, a fuzzy CBR based on both problems and actors’ similarities is advanced to measure usefulness of past cases. Then, we rely on a meta-heuristic optimization technique i.e. Particle Swarm Optimization to adjust optimally the parameters of the inputs and outputs fuzzy membership functions.

Data Cleaning is a process for determining whether two or more records defined differently in database, represent the same real world object. Data Cleaning is a vital function in data warehouse preprocessing. It is found that the problem of duplication /redundancy is encountered frequently when large amounts of data collected from different sources is put in the warehouse. Eliminating redundancy in the data warehouse resolves conflicts in making wrong decisions. Data cleaning is also used to solve problem of “wastage of storage space”. One way of eliminating redundancy is by retrieving similar records using tokens formed on prominent attributes. Another approach is to use Conditional Functional Dependencies (CFD’s) to capture the consistency of data by combining semantically related data. Existing work on data cleaning do not deal with the case of multi-valued attributes. This paper deals with nesting based weak multi-valued dependencies (n-wMVD) which can handle multi-valued attributes and redundancy removal. Our contributions are of two fold (i) An approach to convert the given database to wMVD (ii) Implementation of n-wMVD to eliminate redundancy. The applicability of our approach was tested. The results are encouraging and are presented in the paper.

Current Information retrieval systems like Google are based on keywords wherein the result is in the form of list of documents. The number of retrieved documents is large. The user searches these documents one by one to find the correct answer. Sometimes the correct or relevant answer to the searched keywords is difficult to find. Studies indicate that an average user seeking an answer to the question searches very few documents. Also, as the search is tedious it demotivates the user and he/she gets tired if the documents do not contain the content which they are searching for. Question-answering systems (QA Systems) stand as a new alternative for Information Retrieval Systems. This survey has been done as part of doctoral research work on “Medical QA systems”. The paper aims to survey some open and restricted domain QA systems. The surveyed QA systems though found to be useful to obtain information showed some limitations in various aspects which should resolved for the user satisfaction.

A simplified procedure for weight computations from the pair-wise comparison matrices of triangular fuzzy numbers in the fuzzy analytic hierarchy process is proposed. A transform T:R3→R1 has been defined for mapping the triangular fuzzy numbers to equivalent crisp values. The crisp values have been used for eigenvector computations in a manner analogous to the computations of the original AHP method. The objective is to retain both the ability to capture and deal with inherent uncertainties of subjective judgments, which is the strength of fuzzy modeling and the simplicity, intuitive appeal, and power of conventional AHP which has made it a very popular decision making tool.

Machine learning approaches based on decision trees (DTs) have been proposed for classifying networking traffic. Although this technique has been proven to have the ability to classify encrypted and unknown traffic, the software implementation of DT cannot cope with the current speed of packet traffic. In this paper, hardware architecture of decision tree is proposed on NetFPGA platform. The proposed architecture is fully parameterizable to cover wide range of applications. Several optimizations have been done on the DT structure to improve the tree search performance and to lower the hardware cost. The optimizations proposed are: a) node merging to reduce the computation latency, b) limit the number of nodes in the same level to control the memory usage, and c) support variable throughput to reduce the hardware cost of the tree.

Due to their quite complex nature, Dynamic Hybrid Systems represent a constant challenge for their diagnosing. In this context, this paper proposes a general multiple faults model-based diagnosis methodology for hybrid dynamic systems characterized by slow discernable discrete modes. Each discrete mode has a continuous behavior. The considered systems are modeled using hybrid bond graph which allows the generating of residuals (Analytical Redundancy Relations) for each discrete mode. The evaluation of such residuals (detection faults step) extends previous works and is based on the combination of adaptive thresholdings and fuzzy logic reasoning. The performance of fuzzy logic detection is generally linked to its membership functions parameters.Thus, we rely on Particle Swarm Optimization (PSO) to get optimal fuzzy partition parameters. The results of the diagnosis module are finally displayed as a colored causal graph indicating the status of each system variable in each discrete mode. To make evidence of the effectiveness of the proposed solution, we rely on a diagnosis benchmark: The three-tank system.

Lightning locating system is very useful for the purpose of giving exact coordinates of lightning events. However, such a system is usually very large and expensive. This project attempts to provide instantaneous detection of lightning strike using the Time of Arrival (TOA) method of a single detection station (comprises of three antennas). It also models the whole detection system using suitable mathematical equations. The measurement system is based on the application of mathematical and geometrical formulas. Several parameters such as the distance from the radiation source to the station and the lightning path are significant in influencing the accuracy of the results (elevation and azimuth angles). The signals obtained by all antennas were analysed using the LabVIEW software. Improvements in the lightning discharge locating system can be made by adopting a multi-station technique instead of the currently adopted single-station technique.

Ferroresonance is a complex non-linear electrical phenomenon that can make thermal and dielectric problems to the electric power equipment. Ferroresonance causes overcurrents and overvoltages which is dangerous for electrical equipment. In this paper, ferroresonance investigation will be carried out for the 33kV/110V VT at PMU Kota Kemuning, Malaysia using ATP-EMTP simulation. Different preconditions of ferroresonance modes were simulated to ascertain possible ferroresonance conditions in reality compare with simulated values. The effect of changing the values of series capacitor is considered. The purpose of this series of simulations is to determine the range of the series capacitance value within which the ferroresonance is likely to occur.

The aim of paper is to design SCFDMA system using SFBC and receiver diversity which provides satisfactory performance over fast fading channel environment. The performance evaluation will be checked through MATLAB R2009b simulator. There are comparisons of performance among 1x1 SCFDMA system, 2x1 SCFDMA system using SFBC, 1x2 SCFDMA system using receiver diversity and 2x2 SCFDMA system using SFBC and receiver diversity. We describe design of SCFDMA system using transmitter diversity technique and receiver diversity technique. We have compared the performance of these systems with the conventional SCFDMA system. The main focus is on design of SCFDMA system using SFBC and receiver diversity which enables desired system to combat detrimental effects of fast fading.

Agent Oriented Software Engineering(AOSE) methodologies are meant for providing guidelines, notations, terminologies and techniques for developing agent based systems. Several AOSE methodologies were proposed and almost no methodology deals with testing issues, stating that the testing can be carried out using the existing object-oriented testing techniques. Though objects and agents have some similarities, they both differ widely. Role is an important mental attribute/state of an agent. The main objective of the paper is to propose a role based testing technique that suits specifically for an agent based system. To demonstrate the proposed testing technique, an agent based E-novel system has been developed using Multi agent System Engineering (MaSE) methodology. The developed system is tested using the proposed role based approach and found that the results are encouraging.

The detection of the regions with mutations associated with different pathologies is an important step for selecting relevant genes. The corresponding information of the mutations and genes is distributed in different public sources and databases, so it is necessary to use systems that can contrast different sources and select conspicuous information. This work proposes a virtual organization of agents that can analyze and interpret the results from Array-based comparative genomic hybridization, thus facilitating the traditionally manual process of the analysis and interpretation of results.

All the scientific community cares about is understanding the complex systems, and explaining their emergent behaviors. We are interested particularly in Multi-Agent Systems (MAS). Our approach is based on three steps : observation, modeling and explanation. In this paper, we focus on the second step by offering a model to represent the cause and effect relations among the diverse entities composing a MAS. Thus, we consider causal reasoning of great importance because it models causalities among a set of individual and social concepts. Indeed, multiagent systems, complex by their nature, their architecture, their interactions, their behaviors, and their distributed processing, needs an explanation module to understand how solutions are given, how the resolution has been going on, how and when emergent situations and interactions have been performed. In this work, we investigate the issue of using causal maps in multi-agent systems in order to explain agent reasoning.

A novel learning algorithm is proposed for non linear modeling and identification by the use of the beta basis function neural network (BBFNN). The proposed method is a hierarchical particle swarm optimization (HPSO). The objective of this paper is to optimize the parameters of the beta basis function neural network (BBFNN) with high accuracy. The population of HPSO forms multiple beta neural networks with different structures at an upper hierarchical level and each particle of the previous population is optimized at a lower hierarchical level to improve the performance of each particle swarm. For the beta neural network consisting n particles are formed in the upper level to optimize the structure of the beta neural network. In the lower level, the population within the same length particle is to optimize the free parameters of the beta neural network. Experimental results on a number of benchmarks problems drawn from regression and time series prediction area demonstrate that the HPSO produces a better generalization performance.

In ant colony optimization technique (ACO), the shortest path is identified based on the pheromones deposited on the way by the traveling ants and the pheromones evaporate with the passage of time. Because of this nature, the technique only provides possible solutions from the neighboring node and cannot provide the best solution. By considering this draw back, this paper introduces a fuzzy integrated ACO technique which reduces the iteration time and also identifies the best path. The proposed technique is tested for travelling sales man problem and the performance is observed from the test results.

Handwritten Gestures may vary from person to person. Moreover, they may vary for same person, if taken at different time and mood. Various rule-based automatic classifiers have been designed to recognize handwritten gestures. These classifiers generally include new rules in rule set for unseen inputs, and most of the times these new rules are distinguish from existing one. However, we get a huge set of rules which incurs problem of over fitting and rule base explosion. In this paper, we propose a self adaptive gesture fuzzy classifier which uses maximum entropy principle for preserving most promising rules and removing redundant rules from the rule set, based on interestingness. We present experimental results to demonstrate various comparisons from previous work and the reduction of error rates.

Speech recognition has been developed from theoretical methods practical systems. Since 90’s people have moved their interests to the difficult task of Large Vocabulary Continuous Speech Recognition (LVCSR) and indeed achieved a great progress. Meanwhile, many well-known research and commercial institutes have established their recognition systems including via Voice system IBM, Whisper system by Microsoft etc. In this paper we have developed a simple and efficient algorithm for the recognition of speech signal for speaker independent isolated word recognition system. We use Mel frequency cepstral coefficients (MFCCs) as features of the recorded speech. A decoding algorithm is proposed for recognizing the target speech computing the cepstral distance of the cepstral coefficients. Simulation experiments were carried using MATLAB here the method produced relatively good (85% word recognition accuracy) results.

The present research proposes a paradigm which combines the Wavelet Packet Transform (WPT) with the distinguished Mel Frequency Cepstral Coefficients (MFCC) for extraction of speech feature vectors in the task of text independent speaker identification. The proposed technique overcomes the single resolution limitation of MFCC by incorporating the multi resolution analysis offered by WPT. To check the accuracy of the proposed paradigm in the real life scenario, it is tested on the speaker database by using Hidden Markov Model (HMM) and Gaussian Mixture Model (GMM) as classifiers and their relative performance for identification purpose is compared. The identification results of the MFCC features and the Wavelet Packet based Mel Frequency Cepstral (WP-MFC) Features are compared to validate the efficiency of the proposed paradigm. Accuracy as high as 100% was achieved in some cases using WP-MFC Features.

The goal of research in computer vision is to impart and improvise the visual intelligence in a machine i.e. to facilitate a machine to see, perceive, and respond in human-like fashion(though with reduced complexity) using multitudinal sensors and actuators. The major challenge in dealing with these kinds of machines is in making them perceive and learn from huge amount of visual information received through their sensors. Mimicking human like visual perception is an area of research that grabs attention of many researchers. To achieve this complex task of visual perception and learning, Visual Attention model is developed. A visual attention model enables the robot to selectively (and autonomously) choose a “behaviourally relevant” segment of visual information for further processing while relative exclusion of others (Visual Attention for Robotic Cognition: A Survey, March 2011).The aim of this paper is to suggest an improvised visual attention model with reduced complexity while determining the potential region of interest in a scenario.

Most of the researchers have widely explored the use of link-context to determine the theme of target web-page. Link-context has been applied in areas such as search engines, focused crawlers, and automatic classification. Therefore, extraction of precise link-context may be considered as an important parameter for extracting more relevant information from the web-page. In this paper, we have proposed a rule-based approach for the extraction of the link-context from anchor-text (AT) structure using bottom-up simple LR (SLR) parser. Here, we have considered only named entity (NE) anchor-text. In order to validate our proposed approach, we have considered a sample of 4 ATs. The results have shown that, the proposed LCEA has extracted 100% actual link-context of each considered AT.

One of the most important topics in pattern recognition is text recognition. Especially offline handwritten recognition is a most challenging job due to the varying writing style of each individual. Here we propose offline Malayalam handwritten character recognition using probabilistic simplified fuzzy ARTMAP (PSFAM). PSFAM is a combination of SFAM and PNN (Probabilistic Neural Network). After preprocessing stage, scanned image is segmented into line images. Each line image is further fragmented into words and characters. For each character glyph, extract features namely cross feature, fuzzy depth, distance and Zernike moment features. Then this feature vector is given to SFAM for training. The presentation order of training patterns is determined using particle swarm optimization to get improved classification performance. The Bayes classifier in PNN assigns the test vector to the class with the highest probability. Best

n

probabilities and its class labels from PSFAM are given to SSLM (Statistical Sub-character Language Model) in the post processing stage to get better word accuracy.

The advances in Science and Technology made it possible for people with hearing impairments and deaf to be more involved and get better chances of education, access to information, knowledge and interaction with the large society. Exploiting these advances to empower hearing-impaired and deaf persons with knowledge is a challenge as much as it is a need. Here, we present a part of our work in a national project to develop an environment for automatic translation from Arabic to Saudi Sign Language using 3D animations. One of the main objectives of this project is to develop a bilingual parallel corpus for automatic translation purposes; avatar-based 3D animations are also supposed to be built. These linguistic resources will be used for supporting development of ICT applications for deaf community. Due to the complexity of this task, the corpus is being developed progressively. In this paper, we present a first part of the corpus by working on a specific topic from the Islamic sphere.

This work presents the development of a software-based Malaysian Sign Language recognition system using Hidden Markov Model. Ninety different gestures are used and tested in this system. Skin segmentation based on YCbCr colour space is implemented in the sign gesture videos to separate the face and hands from the background. The feature vector of sign gesture is represented by chain code, distance between face and hands and tilting orientation of hands. This work has achieved recognition rate of 72.22%.

With the progression of digitization it is very necessary to archive the Bangla newspaper as well as other Bangla documents. The first step of reading Bangla Newspaper is to detect headlines and column from multi column newspaper. But there is no such algorithm developed so far in Bangla OCR that can fully read Bangla Newspaper. In this paper we present an algorithmic approach for multi column & headline detection from Bangla newspaper as well as Bangla magazine. It can separate headlines from news and also can detect columns from multi column. This algorithm works based on empty space between headline- columns, column-column.

In this paper, we present an approach for building translation memory for American Sign Language (ASL) from parallel corpora between English and ASL, by identifying new alignment combinations of words from existing texts. Our primary contribution is the application of several models of alignments for Sign Language. The model represents probabilistic relationships between properties of words, and relates them to learned underlying causes of structural variability within the domain. We developed a statistical machine translation based on generated translation memory. The model was evaluated on a big parallel corpus containing more than 800 millions of words. IBM Models have been applied to align Sign Language Corpora then we have run experimentation on a big collection of paired data between English and American Sign Language. The result is useful to build a Statistical Machine Language or any related field.

In this paper, we propose a Selective Parameters based Image Denoising method that uses a shrinkage parameter for each coefficient in the subband at the corresponding decomposition level. Image decomposition is done using the wavelet transform. VisuShrink, SureShrink, and BayesShrink define good thresholds for removing the noise from an image. SureShrink and BayesShrink denoising methods depend on subband to evaluate the threshold value whereas the VisuShrink is a global thresholding method. These methods remove too many coefficients and do not provide good visual quality of the image. Our proposed method not only keeps more noiseless coefficients but also modifies the noisy coefficients using the threshold value. We experimentally show that our method provides better performance in terms of objective and subjective criteria i.e. visual quality of image than the VisuShrink, SureShrink, and BayesShrink.

The mere existence of natural living thing can be studied and analyzed efficiently only by Ontology, where each and every existence are concern as entities and they are grouped hierarchically via their relationship. This paper deals the way of how an image can be represented by its feature Ontology though which it would be easier to analyze and study the image automatically by a machine, so that a machine can visualize an image as human. Here we used the selected MPEG 7 visual feature descriptor and Texton parameter as entity for representing different categories of images. Once the image Ontology for different categories of images is provided image retrieval would be an efficient process as through ontology the semantic of image is been defined.

Aerial and satellite images are projected images where clouds and cloud-shadows cause interferences in them. Detecting the presence of clouds over a region is important to isolate cloud-free pixels used to retrieve atmospheric thermodynamic information and surface geophysical parameters. This paper describes an adaptive algorithm to reduce both effects of clouds and their shadows from remote sensed images. The proposed method is implemented and tested with remote sensed RGB and monochrome images and also for visible (VIS) satellite imagery and infrared (IR) imagery. The results show that this approach is effective in extracting infected pixels and their compensation.

In this research paper, a system capable of taking as input multiple 2-dimensional images of the surrounding environment from a particular position and creating a 3-dimensional model from them, with navigation possible inside it, for the 360 degree view is developed. Existing approaches for image stitching, which use SIFT features, along with approaches for depth estimation, which use supervised learning to train a Markov Random Field (MRF), are modified in this paper in order to improve their efficiency. Also, an improvement in accuracy of depth estimation is suggested for the 3-dimensional model using matching SIFT features from the multiple input images. A method for enabling navigation in the 3D model through which we prevent motion in areas where movement is not possible is outlined, thus making the 3-dimensional model realistic and suitable for practical use. The proposed system is also an application of Neural Networks.

3D game engines are originally developed for 3D games. In combination with developing technologies we can use game engines to develop a 3D graphics based navigation system or 3D Animated Map Viewer (AMV). Visualizing geospatial data (buildings, roads, rivers, etc) in 3D environment is more relevant for navigation systems or maps rather than using symbolic 2D maps. As 3D visualization provides real spatial information (colors and shapes) and the 3D models resembles the real world objects. So, 3D view provides high accuracy in navigation. This paper describes the development of human interactive 3D navigation system in virtual 3D world space. This kind of 3D system is very useful for the government organization, school bodies, and companies having large campuses, etc for their people or employers for navigation purposes.

To design an efficient Wireless Sensor Network application one need to understand the behavior of the sensor nodes deployed. To understand the behavior of a systemwe need a very good model, that can represent the system in amore realistic manner. There is a vagueness that we can identify in defining the sensing coverage of any sensor node. The human like reasoning that best suits for this vagueness is the fuzzy reasoning. Here in this paper, we are proposing a fuzzy based model and inference system to best way represent the sensing behavior of sensor nodes. Also, we propose a measure that can be used to check or compare the WSN system performance.

Peer-to-Peer (P2P) detection using machine learning (ML) classification is affected by its training quality and recency. In this paper, a practical retraining mechanism is proposed to retrain an on-line P2P ML classifier with the changes in network traffic behavior. This mechanism evaluates the accuracy of the on-line P2P ML classifier based on the training datasets containing flows labeled by a heuristic based training dataset generator. The on-line P2P ML classifier is retrained if its accuracy falls below a predefined threshold. The proposed system has been evaluated on traces captured from the Universiti Teknologi Malaysia (UTM) campus network between October and November 2011. The overall results shows that the training dataset generation can generate accurate training dataset by classifying P2P flows with high accuracy (98.47%) and low false positive (1.37%). The on-line P2P ML classifier which is built based on J48 algorithm which has been demonstrated to be capable of self-retraining over time.

As distributed system software gain complexity owing to increasing user needs, monitoring and adaptations are necessary to keep system fit and running. These distributed applications are difficult to manage due to changing interaction patterns, behaviors and faults resulting from varying conditions in the environment. Also the rapid growth in Internet users and diverse services has highlighted the need for intelligent tools that can assist users and applications in delivering the required quality of services. To address these complexities, we introduce mobile agent based monitoring for supporting the self healing capabilities of such distributed applications. We present the novel mobile agent based monitoring technique where the monitor agents constantly collect and update the global information of the system using antecedence graphs. Updating weights of these graphs further help in evaluating host dependence and failure vulnerability of these hosts. These graphs help monitoring mobile agents to detect undesirable behaviors and also provide support for restoring the system back to normalcy.

The increasing transmission of portable devices with wireless connectivity enables new pervasive scenarios, where users require personalized service access according to their needs, position, and environment conditions (context-aware services). A fundamental requirement for the context-aware service discovery is the dynamic retrieval and interaction with local resources, i.e., resource discovery. The high degree of dynamicity and heterogeneity of mobile environments requires moving around and/or extending traditional discovery solutions to support more intelligent service search and retrieval, personalized to user context conditions. We have reviewed the research of context aware service discovery that based on semantic data representation and technologies; allow flexible matching between user requirements and service capabilities in open and dynamic deployment scenarios.

IEEE 802.15.4 is an emerging standard specifically designed for low-rate wireless personal area networks (LR-WPAN) with a focus on enabling the wireless sensor networks. It attempts to provide a low data rate, low power, and low cost wireless networking on the device-level communication. In low rate wireless personal area networks the position of each node changes over time. A network protocol that is able to dynamically update its links in order to maintain strong connectivity is said to be self-reconfiguring. In this paper, we propose a fuzzy logic system for detecting ping pong effect attack in low rate wireless personal area networks design method with self-reconfiguring protocol for power efficiency. The LR-WPAN is self-organized to clusters using an unsupervised clustering method, fuzzy clustering means (FCM). A fuzzy logic system is applied to master/controller selection for each cluster. A self-reconfiguring topology is proposed to manage the mobility and recursively update the network topology. We also modify the mobility management scheme with hysteresis to detect the ping-pong effect attack.

In this paper, we propose a global quality of service management applied to DiffServ environments and IEEE 802.11e wireless networks. Especially, we evaluate how the IEEE 802.11e standard for Quality of Service in Wireless Local Area networks (WLANs) can interoperate with the Differentiated Services (DiffServ) architecture for end-to-end IP QoS. An Architecture for the integration of traffic conditioner is then proposed to manage the resources availability and regulate traffic in congestion situation. This traffic conditioner is modelled as an agent based on reinforcement learning.

To achieve high reliability of biometric authentication using fusion of multimodal biometrics in authentication systems is a novel approach. In this paper we propose a method for the management of access control to ensure the desired level of security using the adaptive combination of multimodal matching scores. It uses a score normalization technique for multimodal biometric authentication using fingerprint, palmprint and voice. This technique is based on the individual scores obtained from each of the biometrics and then normalized to get a fused score. Training data sets are generated from genuine and impostor score distributions. Also this technique is compared with other score normalization techniques and the performance of the proposed system is analyzed. The proposed multimodal biometric authentication system overcomes the limitations of individual biometric systems and also meets the response time as well as the accuracy requirements.

Web Usage Mining is the extraction of information from web log data. The extended web log file contains information about the user traffic and behavior, the browser type, its version and operating system used. Mining these web logs provide the origin of visit or the referring website and popular keywords used to access a website. This paper proposes an indiscernibility approach in rough set theory to extract information from extended web logs to identify the origin of visits and the keywords used to visit a web site which will lead to better design of websites and search engine optimization.

Developed online social networks are recently being grown and popularized tremendously, influencing some life aspects of human. Therefore, privacy preservation is considered as an essential and crucial issue in sharing and propagation of information. There are several methods for privacy preservation in social networks such as limiting the information through community detection. Despite several algorithms proposed so far to detect the communities, numerous researches are still on the way in this area. In this paper, a novel method for community detection with the assumption of privacy preservation is proposed. In the proposed approach is like hierarchical clustering, nodes are divided alliteratively based on learning automata (LA). A set of LA can find min-cut of a graph as two communities for each iteration. Simulation results on standard datasets of social network have revealed a relative improvement in comparison with alternative methods.

During web search, we often end up with untrusted, duplicates and near duplicate search results which dilutes the focus of search query. Factors that may influence the trust of web search results shall be referred to as ’Provenance’. Provenance is basically the information about the history of data. In this paper, we propose a provenance model which uses both content based and trust based factors in identifying trusted search results. The novelty of our idea lies in attempting to construct a provenance matrix which encompasses 6 factors (who, where, when, what, why, how) related to the search results. Inferences performed over the provenance matrix leads to trust score which is then utilized to remove near-duplicates and retrieve trusted search results.

Cloud computing is an internet based pay as use service which provides three type of layered services (Software as a Service, Platform as a Service and Infrastructure as a Service) to its consumer on demand. These on demand service facilities is being provide by cloud to its consumers in multitenant environment but as facility increases complexity and security problems also increase. Here all the resources are at one place in data centers. Cloud uses public and private APIs (Application Programming Interface) to provide services to its consumer in multitenant environment. In this environment Distributed Denial of Service attack (DDoS), especially HTTP, XML or REST based DDoS attacks may be very dangerous and may provide very harmful effects for availability of services and all consumers may get affected at the same time. One other reason is that because the cloud computing users make their request in XML and then send this request using HTTP protocol and build their system interface with REST protocol (such as Amazon EC2 or Microsoft Azure) hence XML attack more vulnerable. So the threaten coming from distributed REST attacks are more and easy to implement by the attacker, but to security expert very difficult to resolve. So to resolve these attacks this paper introduces a comber approach for security services called filtering tree. This filtering tree has five filters to detect and resolve XML and HTTP DDoS attack.

In this paper, the main target is to achieve distributed atmosphere of Cloud configuring heterogeneous structure framework with accessible common equipments, technologies, and configurable high-end servers concerning no additional expenditure to the system network. In proposed system, different categories of machines are being utilized to generate efficient diverse background. Server-side background mode of operation is to be conducted for accessing dedicated servers as well as all-purpose servers which are not only assigned for the user specific responsibilities. This is an additional challenge for this approach to make it happen using any kind of server machines. In this approach, unicast topology is going to be used for avoiding network congestion. Minimization of time and maximization of speed are the objectives of proposed system structure. Earlier version of this paper has been published in [1].

The advent of technology has led to the emergence of new technologies such as cloud computing. Evolution of IT industry has oriented towards the consumption of large scale infrastructure and development of optimal software products, thereby demanding heavy capital investment by the organizations. Cloud computing is one of the upcoming technologies that have enabled to allocate apt resources on demand in a pay-go approach. However, the existing techniques of load balancing in cloud environment are not efficient in reducing the response time required for processing the requests. Thus, one of the key challenges of the state-of- art of research in cloud is to reduce the response time, which in turn reduces starvation and job rejection rates. This paper, therefore aims to provide an efficient load balancing technique that can reduce the response time to process the job requests that arrives from various users of cloud. An enhanced Shortest Job First Scheduling algorithm, which operates with threshold (SJFST), is used to achieve the aforementioned objective. The simulation results of this algorithm shows the realization of efficient load balancing technique which has resulted in reduced response time leading to reduced starvation and henceforth lesser job rejection rate. This enhanced technique of SJFST proves to be one of the efficient techniques to accelerate the business performance in cloud atmosphere.

Security is one of the most challenging ongoing research area in cloud computing because data owner stores their sensitive data to remote servers and users also access required data from remote cloud servers which is not controlled and managed by data owners. This paper Proposed a new algorithm PASA (Privacy-Aware Security Algorithm) for cloud environment which includes the three different security schemes to achieve the objective of maximizing the data owners control in managing the privacy mechanisms or aspects that are followed during storage, Processing and accessing of different Privacy categorized data. The Performance analysis shows that the proposed algorithm is highly Efficient, Secure and Privacy aware for cloud environment.

During past few years the wireless network has grown in leaps and bounds as it offered the end users more flexibility which enabled a huge array of services. All these services are achieved due to the network which is the backbone. The concept of the wireless network and the wireless devices also brings a lot of challenges such as energy consumption, dynamic configuration and congestion. Congestion in a network occurs when the demand on the network resources is greater than the available resources and due to increasing mismatch in link speeds caused by intermixing of heterogeneous network technologies. It is not limited to a particular point in the network but it can occur at various points in the network and it results into high dropping and queuing delay for packets, low throughput and unmaintained average queue length. It is factor that affects a network in a negative manner. Queue management provides a mechanism for protecting individual flows from congestion. One of the technique which uses Active Queue Management technique is RED. The basic idea behind RED queue management is to detect incipient congestion early and to convey congestion notification to the end-hosts, allowing them to reduce their transmission rates before queues in the network overflow and packets are dropped. Carnegie Mellon University proposed a new queue based technique for wireless network called CMUQ. The basic philosophy behind CMU queue is to prevent congestion. This paper introduces a new range variable and priority queue for existing CMU queue and for RED algorithm.

Load balancing has been known as one of the most challenging problems in computer sciences especially in the field of distributed systems and grid environments; hence, many different algorithms have been developed to solve this problem. Considering the revolution occurred in the modern processing units, using mutli-core processors can be an appropriate solution. one of the most important challenges in multi-core-based grids is scheduling. Specific computational intelligence methods are capable of dealing with complex problems for which there is no efficient classic method-based solution. One of these approaches is multi-objective genetic algorithm which can solve the problems in which multiple objectives are to be optimized at the same time. CoreIIScheduler, the proposed approach uses NSGA-II method which is successful in solving most of the multi-objective problems. Experimental results over lots of different grid environments show that the average utilization ratio is over 90% whilst for FCFS algorithm, it is only about 70%. Furthermore, CoreIIScheduler has an improvement ratio of 60% and 80% in wait time and makespan, respectively which is relative to FCFS.