About admin

This author has not yet filled in any details.
So far admin has created 849 blog entries.

A CASE STUDY ON TENSORFLOW AND ARTIFICIAL NEURAL NETWORKS

Vivekanandan B 1, Hemadarshini M V 2

 1AVIN Systems Private Limited Bangalore, India

 2Department of Mathematics Mount Carmel College Bangalore, India

 

ABSTRACT: 

 

There are a number of machine learning software in use. In this paper, we look at the implementation of artificial neural networks using TensorFlow – A machine learning software developed by Google. TensorFlow has been used to solve two image classification problems – Classification of digits of MNIST database and classification of traffic signs of GTSRB database. Classification of MNIST digits could be done with 99% accuracy after 20000 iterations with every 500 iterations taking about 25 seconds. Classification of GTSRB traffic signs could be done with about 70% accuracy with every iteration taking about 9 minutes. Both these neural networks have been trained and tested on a CPU at 2.3GHz and 4 GB RAM. It has been noted that CPUs are fast enough to implement neural networks on a small scale. For real world applications, neural networks should have considerably more parameters which can conveniently be trained and implemented on GPUs. It has been noted that TensorFlow is capable of handling convolutional neural networks (CNNs) in an efficient manner.

 

Keywords: TensorFlow, Artificial Intelligence, Neural Networks, Machine Learning, Image Classification

 

pdf

By | 2017-09-21T16:42:26+00:00 September 21st, 2017|General, Publications|0 Comments

SECURITY CONSIDERATIONS FOR IOT APPLICATION LAYERS USING MQTT

Vinutha M1, Prof. Sankar Dasiga2

M.Tech Scholar, Department of CNE, NMIT, Bangalore, India1

Professor, Department of ECE, NMIT, Bangalore, India2

 

ABSTRACT: 

 

The concept Internet of Things(IoT) is a major developmental area of interest in the computer, information science and communication technology. Since IoT is  used to the higher extent in soo many fields, using of IoT matters but the thing is how  secure IoT communication is, hence security will acquire great effect in the domain of IoT. IoT, features and its systemic structure, analyzes the security problem of IoT in each and every layer of IoT communication like perception layer ,network layer and an application layer in the system of IoT, It puts forward the construction of secure IoT using MQTT protocol, as it offers the corresponding secure strategies based on the existing problems in the field of IoT. Hence, presented ideas to form the consistent security system for IoT. Since IoT is broadly used in many fields. Therefore the secure communication of IoT construction must be concentrated using the corresponding secure strategies like MQTT. HTTP has been majorly used for data transfer. However, in the networks of IoT, this protocol causes a high overhead. To solve this issue, transfer protocols have been reviewed. When HTTP and MQTT has  observed and compared,  the performance of HTTP with that of MQTT is high and MQTT offers less overhead too. Hence, enhancements using MQTT has made to secure the IoT Network Technology for secure and better performance.Vinutha M

 

Keywords: IoT, MQTT,  HTTP,Raspberry Pi, Sensors, MD5

pdf

By | 2017-09-21T16:38:19+00:00 September 21st, 2017|General, Publications|0 Comments

COMPARISONS OF SEGMENTATIONS IN BRAIN TUMOR BY ANT COLONY OPTIMIZATION AND GENETIC ALGORITHM

Dr. T. Logeswari

 Associate Professor, New Horizon College, Bangalore, India

ABSTRACT: 

The brain tumor is detected through radiologist with help of MRI, which takes considerably a maximum time. Most of the brain tumor detection methods provide compound information about the brain tumor and they lack in providing an accurate result on existence of tumor. As a result, a formal consultation with a radiologist is mandatory, which becomes a surplus expenditure in case of a non-tumor patient.  The objective of this research work is to develop a supporting system that would aid the radiologist to have aforementioned result which reduces the time taken in brain tumor detection.  The proposed method consists of two stages. In  the First  stage Ant colony Optimization  method combined with unsupervised method Fuzzy C mean to detect  brain Tumor. In the second stage Genetic algorithm combined with unsupervised method Fuzzy C mean to detect brain tumor. the study compares the two stages and shows that the most effective method  Ant Colony Optimization (ACO) with FCM method reduces the time complexity for brain tumor detection which also includes more accuracy. T. Logeswari

Keywords  Ant Colony Optimization, Genetic Algorithm, MRI Image, Fuzzy C Mean

 

pdf

By | 2017-09-21T16:34:40+00:00 September 21st, 2017|General, Publications|0 Comments

A SURVEY PAPER ON METHODS USED FOR THE EXTRACTION OF RETINAL VESSELS FOR DIABETIC RETINOPATHY

Sandhya Soman1,Vani Chakraborty2

1Department of Computer Science, UG, Kristu Jayanti College, Bangalore,India

2Department of Computer Science,UG,Kristu Jayanti College, Bangalore, India

 

ABSTRACT:

Due to the sedentary lifestyles these days, majority of the population in the developing and developed countries suffer from diabetes-a common disease due to body’s inability to maintain its glucose level. A good count of them, have poor vision due to diabetic retinopathy. This disease, if diagnosed at its non-proliferative stage, can enable and make it easy for the ophthalmologists to find effective cure for it. Thus a lot of research has been done in order to automate imaging systems which can identify diseased retinal images from healthy ones. A very vital step in that is the segmentation of the blood vessels of retina from the other micro developments in the retina due to the diseased condition. Hence, in this paper, we have conducted a study of the various techniques available in the literature for the identification of retinal blood vessels.

 

Keywords: Segmentation, curvelet transform, thresholding, line tracking, bottom hat transformation, diabetic retinopathy.

 

pdf

By | 2017-09-21T16:26:54+00:00 September 21st, 2017|General, Publications|0 Comments

A STUDY ON THE ARTIFICIAL INTELLIGENCE IN TOP GENERATION AND MULTIDOCUMENT SUMMARIZATION

1S. Saravana Kumar

1 Assistant  Professor, Department of Computer Science, Kristu Javanti , College Bangalore

 

Abstract:  

Artificial intelligence has drawn a lot of attention in both academic and industrial fields as vast amount of Text materials available in the various data source without of proper information management. To alleviate the problem, we analyse the sentence level clustering in terms of artificial intelligence paradigms in the area of document clustering, topic generation and multi document generation. In this paper, we study artificial intelligence models in detail for extracting topic and summarizing the document. Much significant impact on the sentence level clustering has made in order to establish the knowledge representation with effectiveness and accuracy. Though Extensive study, Future model for topic generation based on Hierrachical clustering and frequent Itemset mining is presented which is capable of identifying the overlapping clusters of conceptually and semantically related sentences.


Keywords – Sentence Clustering, Document Summarization, Topic generation, Document Clustering, Artificial Intelligence 

 

pdf

By | 2017-09-21T16:22:17+00:00 September 21st, 2017|General, Publications|0 Comments

ACCURATE BRAIN TUMOR PREDICTION SYSTEM FROM THE LARGE VOLUME OF DATA USING NETWORK PARTITIONING AWARE ENSEMBLE CLASSIFIER METHOD

Mary Jacob 1, Aswin Herbert Sathish  2

 1Department of Computer Science, Kristu Jayanti College, Bangalore, India

 2Department of Computer Science, Kristu Jayanti College, Bangalore, India

 

ABSTRACT:

Brain tumor prediction is the most concerned research field in the bio medical research environment where the most of the people from all age categories caused by serious threats. Accurate brain tumour prediction requires analysis of large volume of health care data gathered from multiple users. However handling large volume of data would be more complex task which needs to be focused well for the accurate prediction rate. This is resolved in our previous research work by introducing the method namely Accurate Prediction of Brain Tumor Disease from Big Data Framework (APBTD-BDF). However this method might be reduced in its performance under presence of more inter communication between different parts of data. And also prediction accuracy might get reduced due to not providing preference on highly correlated features. These problems are resolved in the proposed research technique by introducing novel framework namely Network Cost aware Brain Tumour Prediction using Ensemble Classifier (NCBTP-EC). In this research method initially given input brain tumor data set is partitioned into multiple parts to make them execute in the multiple nodes with the concern of network communication cost. After allocating the data into multiple nodes, filtering is done by using Highly Correlated Filter which can remove the more repeated and similar data contents present in the system, thus the classification performance can be optimized. Finally Weighted Vote based Ensemble classifier involving Support Vector Machine (SVM), Adaboost and Random Forest classifier are applied to predict the brain tumour disease presence. The overall evaluation of the research method is conducted in the matlab simulation environment from which it is proved that the proposed research method leads to provide optimal outcome than the existing research techniques.

Keywords: Mary Jacob

 

pdf

By | 2017-09-21T16:17:14+00:00 September 21st, 2017|General, Publications|0 Comments

ANALYZING THE SAFETY IN TRAFFIC NAVIGATION

Manjusha Sreekumar1,  Dany Antony2, Arun N P3 Lakshmi Madhuri K4

 1234MCA Department, T.John College, Bangalore, India

 

ABSTRACT: 

 

In real time traffic situations a driver encounters a number of traffic signs and road signs, which may go unnoticed by the human eye due to lack of awareness. This paper highlights the challenges associated with automotive safety in the real time traffic navigation. In real time traffic scenario, driver should not only maintain traffic regulations but also observe the adjacent moving vehicles. A preliminary research has been carried out in assisting the driver with the information of other approaching vehicles. One of the parameters for alarming the driver is speed of the approaching vehicle. With an effective method for tracking and proper calibration of the camera for real time traffic scenarios, speed detection is possible. Some of the existing approaches use perspective or linear projections to model the camera for the purpose of calibration. The effective method of calibration to be used for a real time traffic scenario depends on the complexity of the system and assistance provided to the driver. The exact speed determination with multiple target tracking is still a crux in real time traffic navigation research. We, hereby, propose an image mining and computer vision based approach using single camera mounted on the moving vehicle to capture the front view.

 

M.Manasa ManjunathKeywords:  Image Mining, Object Detection, Tracking, Computer Vision, Traffic Navigation

 

pdf

By | 2017-09-21T16:10:35+00:00 September 21st, 2017|General, Publications|0 Comments

INCREASED PRODUCTIVITY AND REDUCED WATER WASTAGE THROUGH AUTOMATIC COMMUNICATING DEVICES WITH STORED THRESHOLD VALUES IN AGRICULTURE

 M.Manasa Manjunath

 Department of Computer Science, Kristu Jayanti College, Bangalore, India

ABSTRACT: 

 

The growing population has increased the need for the increase in food production and the need for reduced water wastage. These days we can find the devices which are used for testing the dampness in the soil. But all these existing systems may just give us an alarm about the reading or just notify us about the same. I would like to propose an idea where we can integrate the dampness checking devices or the moisture sensing sensor device with the communicating device so that this application would open the valves to pump the water to the fields. The communicating device is set up with a predefined value called threshold value. The dampness level reading when read by the sensor is compared with the threshold value in the communicating device. If the value read by the sensor is less than the threshold value then the valve gets opened automatically and water will be pumped to the fields. If the value read by the sensor is greater than or equal to threshold value, then the valve closes automatically and then the pumping of water should be stopped. . With this proposed ideology one can save the water not flowing too much to the fields and the system makes sure that water wastage is completely seized.  Thus the moisture level is properly maintained and this can lead to proper food yield and enhanced profits for the famers.

 

Keywords: Dampness checking sensors or devices, Information and Communication Technology(ICT), Bioinformatics

 

pdf

By | 2017-09-21T16:06:04+00:00 September 21st, 2017|General, Publications|0 Comments

A MODIFIED PROBABILISTIC K-D TREE APPROACH TO IMPROVE THE DATA QUALITY OF SENSOR NETWORK

Krishnendu Saha1, Chandrani Ray Chowdhury2

 1Department of Computer Science Brainware University Barasat, India

 2Department of Computer Science & Engineering Brainware Group of Institutions Barasat, India

 

ABSTRACT: 

 

Sensor network is the collection of sensor nodes, which sends the live data (sensed data) to the base station or the nearest neighbors. As the sensor nodes are driven by limited power battery, an efficient utilization of energy is essential to use the network for a long time period and prevent data losses as much as possible. Data quality is directly related to a sensor node’s life. Limited database memory, limited power and complex network structure are the main characteristics of a sensor node and it’s a challenge to ensure data quality for the nodes. For all such nodes it is essential to ensure that each time the nodes form an updated route path and continue data sensing even if any failure occurs. The main goal of this paper is to improve data quality by implementing a mobile agent based algorithm, backup database approach, Tiny DB SQL filter query which will work for every node that has a probability of failure due to low energy. This paper mainly focuses on the improvement of data collection by using modified k-d tree approach.

Keywords: wireless sensor nodes; k-d tree, probability; data collection

 

pdf

By | 2017-09-21T16:03:33+00:00 September 21st, 2017|General, Publications|0 Comments

ANALYSIS OF NSS IMAGE CHARACTERISTICS

Gopika S 1, D. Malathi 2

 1Department of Computer Science , SRM University , Chennai

  

ABSTRACT: 

 

Human society always demands for a tool that helps in analyzing the quality of the visual content getting transferred in the internetwork. This paper aims in analyzing the visual quality of NSS images by means of statistical analysis. Mathematical processing is done on the image and results reveal the quality value possessed by the given image. Objective analysis is then compared with subjective score to state the versatility of the method. Images from LIVE dataset is taken for experiment.

 

Keywords: MSCN ; Gaussian distribution; Quality Assessment; distortion.

 

pdf

By | 2017-09-21T15:59:34+00:00 September 21st, 2017|General, Publications|0 Comments

ENHANCED DISTRIBUTED SECURITY ARCHITECTURE FOR AUTHENTICATION AND ATTACK PREVENTION IN 4G NETWORKS

Niranjani1 and Dr. M. Ganaga Durga2

    1Research Scholar, Bharathiyar University, Coimbatore, 

2Research Supervisor, Bharathiar University,Coimbatore,

 

ABSTRACT: D. Niranjani

Fourth Generation (4G) network is the highly advanced wireless network which aids broadband as well as multimedia applications. The increased set of features available at the user terminal also leads to the maximization of the security risks. It is susceptible to malicious attacks like corruption of the charge, Distributed Denial of Service (DoS) etc. Authentication techniques in 4G networks should provide integrity as well as should avoid this kind of attacks.  Hence, to overcome the security problems in the 4G network, in this paper, an enhanced distributed security architecture for authentication and attack prevention is developed. In this proposed architecture, the Elliptic Curve Diffie–Hellman (ECDH) protocol is used for authenticating the mobile nodes within the network through hop by hop authentication and neighbor authentication. In order to prevent DDoS attack, shared authentication information approach is used.

 

Keywords: 4G Networks; Authenticatio; Distributed Security Architecture; Elliptic Curve Diffie–Hellma;, Distributed Denial of Service (DDoS).

 

pdf

By | 2017-09-21T15:55:04+00:00 September 21st, 2017|General, Publications|0 Comments

A REVIEW ON EDUCATIONAL DATA MINING TECHNIQUES AND RECOMMENDATION MODEL IN ANALYZING STUDENT’S PERFORMANCE

Aswin Herbert Sathish 2, Mary Jacob 1, Riju K Sajee  3

 1,2Department of Computer Science, Kristu Jayanti College, Bangalore, India

 2Department of Computer Science, Kristu Jayanti College, Bangalore, India

                                3BCA Student, Department of Computer Science, Kristu Jayanti College, Bangalore India

ABSTRACT: 

 

The Mining of education data is an emerging trend in the learning analytics as it is time-consuming to analyze the data and to identify the hidden information automatically. In this paper, adetailed investigation of educational data mining technique is carried out. The application of the data mining includes storage and retrieval of student data in large repositories such as mark sheets, attendance sheets, student profiles, etc. The importance analysis is carried out on the retrieval of large data using machine learning algorithm in data mining. Along the retrieval of the data, these days, deep focus is made on prediction and recommendation models which provide more effectiveness to educational applications in terms of suggesting and extracting the correlation among students. However handling of large data from repositories leads to performance bottleneck, hence it is resolved by employing Map Reduce Paradigm from big data analytics. Through extensive study, classification and clustering provide more value for the management, hence Semantic and Opinion Mining is presented as the future research solution.

 

Keywords: Educational data mining, Clustering, Classification, Data Prediction, Recommendation, Map Reduce, Semantic and Opinion Mining

 

pdf

By | 2017-09-21T15:50:49+00:00 September 21st, 2017|General, Publications|0 Comments

A STUDY ON THE ELECTROENCEPHALOGRAPHY CONTROL SCHEMES FOR PATTERN RECOGNITION

Amjad Hassan Khan Mk 1, Nagendra S

1Asst.professor, Kristu Jayanti College, Bangalore.

2Asst.professor, Kristu Jayanti College, Bangalore.

ABSTRACT:

The research proposes an Exploratory study of simple and efficient movement classification technique for Electroencephalography control schemes on brain fingerprinting. The pattern recognition using Electroencephalography is analysed in detail in this work. Most brain fingerprinting using Electroencephalography control studies on brain waves have shown good performance. The Control generated can be acceptable or unacceptable. As an analysis, in this work, focus is made on efficient pattern recognition on the Electromyography for the application (human) brain fingerprinting. The signal is neural signals which gathered from the sensor of Electroencephalography Recording site can be used as input to decide the brain signal. The Signals were segmented and features were extracted with time domain feature extraction methods. The feature considered is various gestures. The control scheme is modelled with supervised and unsupervised learning mechanism for muscle configurations. In this work, detailed analysis various control mechanism for pattern recognition and classification carried with merits and demerits using fuzzy logic control. The pattern recognition through control scheme will be capable distinguishing the source to improve the classification performance in controlling functioning in the brain fingerprinting. The outcome of this study encourage in modelling the new control scheme with novel ensemble classification technique for brain fingerprinting application to any king brain waves.

 

Keywords: Electroencephalography, Pattern recognition techniques, signal processing, Fuzzy logic

 

pdf

By | 2017-09-21T15:46:58+00:00 September 21st, 2017|General, Publications|0 Comments

AN EFFICIENT MINING FOR MAXIMAL FREQUENT SEQUENCE PATTERN USING BINARY DIGIT REPRESENTATION AND SAME SUPPORT VALUE

Ramesh1 N. Jayaveeran2

Research Scholar1, Assistant Professsor2

Department of Computer Science, Khadir Mohideen College, Adhirampattinam, Tamilnadu1

Department of Computer Science, Khadir Mohideen College, Adhirampattinam, Tamilnadu3

 

ABSTRACT:

Mining Sequential Frequent Pattern gives more patterns to user. It is perplex for decision making in business and other applications in Data mining. Because of that the Maximal Closed Frequent Sequential Pattern Mining is proposed by many researchers. However, the Maximal Pattern is mined from vast sequence database which gives more number of patterns. This research paper is proposed the Efficient Maximal Closed Frequent Sequence Pattern (EMaxSPAN) to reduce the processing time and the Patterns by same support threshold value by user given minimum support value. The efficiency is experimented in real time sequence databases.

Keywords: Pattern Mining, Same Support Value, Maximal, Closed Sequential, Frequent pattern

 

pdf

By | 2017-09-20T16:58:37+00:00 September 20th, 2017|General, Publications|0 Comments

MANAGEMENT SYSTEM OF SCIENTIFIC INSTRUMENTS INVENTORY IN REGULATORY AUTHORITY

Kh. Abdul Aziz, R. M. A. Lashin, and N. S. Mahmoud.

1 Operation Safety and Human Factors Dept. Nuclear Installations Safety Division,

Nuclear & Radiological Regulatory Authority, Cairo, Egypt.

 

ABSTRACT

 

The inventory of scientific instruments plays a big role in developing the capabilities of regulatory and technical support organizations. It has a significant effect in minimizing costs, increasing the organization efficiency, and promoting the employee efforts. This paper demonstrates a software package for management of the scientific instrumentation inventory in the regulatory and technical support organization based on Oracle 8.0 under Windows XP operating system. It provides an Arabic user interface, formulates a model for describing the behavior of the inventory system, uses of the computerized information processing system, the maintained records for the current inventory levels. This is essential for identifying the optimum inventory policy.

 

  Keywords: Database, Inventory System, Nuclear & Radiological Regulatory Authority

 

pdf

By | 2017-09-20T16:53:43+00:00 September 20th, 2017|General, Publications|0 Comments

SECURITY AFFAIRS OF CLOUD COMPUTING

[13] Mohini GourMohini Gour 1, Abhigyan Tiwari 2, Dr. Amit Shrivastava 3 , Dr. Manish Manoria 4

 1Department of Computer Science  Engineering,  Sagar Institute Of Research Technology & Science Bhopal, India

 2Department of Computer Science  Engineering, Sagar Institute Of Research Technology & Science Bhopal, India

3Department of Computer Science  Engineering,  Sagar Institute Of Research Technology & Science Bhopal, India

4Department of Computer Science  Engineering,  Sagar Institute Of Research Technology & Science Bhopal, India

 

ABSTRACT:

 

Cloud Computing is a distributed architecture that centralizes server resources on a scalable platform so as to provide on demand computing resources and services Due to the unprecedented success of internet in last few years, computing resources is now more ubiquitously available. And it enabled the realization of a new computing concept called Cloud Computing. Cloud Computing environment requires the traditional service providers to have two different ways. These are infrastructure and service providers. Infrastructure providers manage cloud platforms and lease resources according to usage. Service providers rent resources from infrastructure providers to serve the end users. Cloud Computing has attracted the giant companies like Google, Microsoft, and Amazon and considered as a great influence in today’s Information Technology industry. This paper is basically suggested the basic idea of the secured-cloud-computing. In this paper, we are showing a future work for cloud’s security and few solutions for the issues of cloud security. This paper may provide the brief overview of the several techniques for the security of cloud data also with its limitations.

Keywords: Cloud Computing, Security Issues, encryption

pdf

By | 2017-09-20T16:43:15+00:00 September 20th, 2017|General, Publications|0 Comments

PARTICLE SWARM OPTIMIZATION BASED VOICE ACTIVITY DETECTION FOR STUTTERED TAMIL SPEECH

Manjutha1, Dr.P.Subashini2

 1,2Department of Computer Science

1,2Avinashilingam Institute for Home Science and Higher Education for Women, Coimbatore, Tamilnadu, India

 

ABSTRACT: 

 

Speech is essential to communicate between humans, as it enables to share information. Stuttering is fluency disorder that affects the prolongation of sounds, syllables, phrase or words, flow of speech, an instinctive repetition and silent pause or blocks in communication. This instinctive speech disorder involves significant problems with the normal fluency and flow of speech. An important front end pre-processing method called Voice Activity Detection (VAD) is used to detect the presence or absence of speech in short segments of stuttered speech. The conventional VAD is desirable to extract the speech signal, based on Frame Energy and Zero Crossing Rate, which contain Voiced, Unvoiced or Silence (VUS) signals. In order to improve the VAD, Particle Swarm Optimization (PSO) is proposed. The main objective of this paper is to update the energy threshold based on conventional method and PSOVAD method. The performance of the proposed PSOVAD is tested with objective parameters like Front End Clipping (FEC), Mid-Speech Clipping (MSC), Over Hang (OVER) and Noise Detected as Speech (NDS). The experimental results suggest that PSOVAD segments the Tamil continuous speech signal into voiced, unvoiced and silence efficiently under poor SNR conditions.

Keywords: Voice Activity Detection, Short-Time Energy, Zero Crossing Rate, Threshold and Particle Swarm Optimization,

 

pdf

By | 2017-09-13T17:57:57+00:00 September 13th, 2017|General, Publications|0 Comments

PERFORMANCE ANALYSIS OF MULTIVERSION ONCURRENCY CONTROL ALGORITHMS

Sonal Kanungo1, Rustom. D Morena2

1 Smt.Z.S.PatelCollege of Computer Application, Surat, 2 Department of Computer Science

Veer Narmad South Gujarat University, Surat

ABSTRACT:

Concurrency control protocols based on multiversion have been used to give the serialized execution of transactions. This paper portrays a simulation study of the execution of different multiversion concurrency control algorithms. Different multiversion techniques are analyzed through simulations. A prototype dynamic environment is presented here for the evaluation of database for specific workload for each algorithm. We are comparing algorithms on the basis of a number of Commits, Rollbacks and Deadlocks in the mix predefine Read-only and Update transactions.

 

Keywords: Concurrency control, Multiversion, Rollback, Deadlock

 

pdf

 

By | 2017-09-13T17:53:49+00:00 September 13th, 2017|General, Publications|0 Comments

INTELLIGENT MACHINE LEARNING FOR HUMAN RECOGNTION USING PERIOCULAR BIOMETRCIS & SCNNS.

Nirgish Kumar¹, Dr. Vivek Srivastava ²

Research Scholar, Faculity of Engineering, Rama University, Kanpur ,

Dean, Faculity of Engineering, Rama University, Kanpur,

ABSTRACT:

 Periocular biometric for human recognition identification is one of the most critical and challenging tasks to meet growing demand for higher security. This paper proposes a new framework to efficiently and accurately match periocular images that are automatically acquired under the intelligent machine learning for human recognition. Our framework, referred to as semantics-assisted convolutional neural networks (SCNNs) in this paper, incorporates explicit semantic information to automatically recover comprehensive periocular features. This strategy enables superior matching accuracy with the usage of relatively smaller number of training samples, which is often an issue with several biometrics. Our reproducible experimental results on four different publicly available databases suggest that the SCNN-based periocular recognition approach can achieve outperforming results, both in achievable accuracy and matching time, for less-constrained periocular matching. Without increasing the volume of training data, the SCNN is able to automatically extract more discriminative features from the input data than a single CNN, therefore can consistently improve the recognition performance. The experimental results presented in this paper validate such an approach to enable faster and more accurate periocular recognition under an intelligent machine learning.

Index TermsPeriocular biometrics recognition, machine learning, convolution neural network, training data augmentation

 

pdf

By | 2017-09-13T17:50:17+00:00 September 13th, 2017|General, Publications|0 Comments

WEIGHTED SIMILARITY PAGE RANK: AN IMPROVEMENT IN WPR AND WSR

Naresh Kumar 1, Prachi Dahiya 2

1Associate Professor, 2B. Tech. Student

Department of Computer Science Engineering

Maharaja Surajmal Institute of Technology, Janakpuri, New Delhi, India.

 

ABSTRACT: 

The  information on the  World  Wide  Web  has been increasing gradually. This increasing information  has  become  a  challenge  for  the  search engine(s) to use the  web  page  ranking   algorithms   in  an   efficient manner. The  existing  algorithms  try to place   the   most  relevant   web   pages  at   the top of the results. But   due   to    this increased    information,    these   algorithms are not  working  up  to  the mark. So,  the search  engines  require   a  new  algorithm to provide the best rankings of     these   web   pages  with    the  most  relevant information. This paper provides    an    analysis of  the previous derived algorithms namely PageRank     algorithm  ,    Weighted     PageRank algorithm,   HITS,  SALSA, Weight and Similarity PageRank   algorithm,   Content   Based   ranking, Topic    Sensitive     ranking  ,   Distance    ranking algorithm,   Eigen   Rumor   algorithm,  etc. It also uses Weighted PageRank, Weight  and Similarity PageRank  to   propose a new       algorithm     called  “Weighted Similarity PageRank” to rank the web pages. This  new  algorithm  not  only improves   the   web     page     rankings      but    also    improve the efficiency of    the search engine(s) which will use this     algorithm.   Thus,   this    paper provides a better  solution   to  rank  web    pages than   the   existing   algorithms.

 

 

pdf

By | 2017-09-13T17:41:55+00:00 September 13th, 2017|General, Publications|0 Comments

A DEFENSIVE APPLICATION TO IDENTIFY THE WEB ATTACKS USING HADOOP

S S Pavani Prathyusha1, S. Suhasini 2, A. Vijaya Banu3

 1M. Tech Scholar, Dept. of Information Technology, V.R. Siddhartha Engineering College, Vijayawada, India

 2Associate Professor, Dept. of Information Technology, V.R. Siddhartha Engineering College, Vijayawada, India

3Manager, BTIO, NRSC, Hyderabad, India

 

ABSTRACT: 

 Web applications these days have increased dependency extending from people to large organizations. Along with the web-based application market growing fast, the data that is being communicated through the network is not secure.  Attackers aim to attack a website or internet server by means of web application queries. Queries are created with the help of properly defined strings and parameters. These are registered in the web server log file. The proposed methodology identifies the basic web application security faults by investigating log records from the web server on Hadoop system. By investigating each record of server log file, the proposed technique recognizes the SQL Injection and Cross Site Scripting (XSS) attacks using Regular Expressions and Pattern Matching algorithms. The Regular Expressions are formed for each pattern of attacks using corresponding string characteristics of the attack. The Pattern list consists of an Anomaly Pattern list that is preserved for each attack. Then Pattern matcher matches, the server log entries to the Anomaly Pattern List. If the server logs pattern is accurately matched with any of the saved patterns in the Anomaly Pattern List then it is said that the Query gets affected by either SQL Injection or Cross Site Scripting attacks.

Keywords: Web application, Malicious, SQL Injection, Cross Site Scripting (XSS), Server log file, Hadoop framework

 

pdf

By | 2017-09-13T17:36:50+00:00 September 13th, 2017|General, Publications|0 Comments

PCB DEFECT DETECTION USING GENETIC ALGORITHM BASED DATA MINING APPROACH

Vaddin Prathiba1, Prof. M Nagendra2 , Prof. M Hanumantappa3

1(Research Scholar, Department of Computer Science , Rayalaseema University, AP, India)

2(Department of Computer science & Technology, Sri Krishnadevaraya University, AP, India)

3(Department of Computer science & Applications, Bangalore  University, Bangalore, India)

 

ABSTRACT: 

 In the field of PCB manufacturing, increasing complexity is becoming an obstacle for manufacturing industries for PCB manufacturing. This complexity issue leads to the manufacturing cost increment and yields a low quality product. In this field of manufacturing industrial application, various techniques and researches have been carried out for improving the manufacturing process and product quality. Soft-computing and computer vision techniques have attracted researchers due to significant nature of these techniques of efficient feature and parameter analysis. In order to deal with manufacturing process and PCB defect detection, we develop a data mining scheme for defect detection in PCBs. Proposed approach is formed by developing a hybrid model where three main stages are present such as (i) data pre-processing (ii) feature selection and reduction and (iii) Classification. To improve the feature selection genetic algorithm is implemented and finally, neural network based classification scheme is implemented. The complete model is developed using MATLAB tool where different types of classifiers are compared in terms of accuracy.

 

Keywords: PCB defect, data mining, feature selection, classification

 

pdf

By | 2017-09-13T17:29:11+00:00 September 13th, 2017|General, Publications|0 Comments

AN ARCHITECTURAL DESIGN FOR CLASSIFICATION OF MULTI RELATIONAL DATA USING SUPPORT VECTOR MACHINE

Thangaraj 1, C.R. Vijayalakshmi 2

 1AssociateProfesssor, Department of Computer Science, Madurai Kamaraj University

           2Department of Computer Science, Madurai Kamaraj University College, Aundipatti, Madurai Tamil Nadu, India

 

ABSTRACT:

  Support Vector Machine (SVM) is a popular pattern classification method with many application areas. SVM is mainly used for high-dimensional data classifications such as raw DNA sequences, Protein sequences/ structure, Genome data, Gene Expression data etc. The main objective of this work is to present an architectural design for classification of multi relational data using support vector machine. It uses Gaussian radial basis function kernel to map nonlinear samples into a higher dimensional space. The empirical results show that the proposed method outperforms well on real biological databases in terms of efficiency and accuracy.

Keywords: Support vector machine, Gaussian radial basis function, Multi-Relational Data Mining, K-Nearest Neighbors

 

[06] C.R. Vijayalakshmi

By | 2017-09-13T17:23:37+00:00 September 13th, 2017|General, Publications|0 Comments

A FAULT TOLERANCE BASED DATA AGGREGTION AND ROUTE OPTIMIZATION USING SWARM INTELLIGENCE APPROACH IN WIRELESS SENSOR NETWORKS

Vinod Kumar Menaria 1, S.C. Jain 2, A. Nagaraju 2

 1,2Department of Computer Engineering, Rajasthan Technical University Kota, India

 3Department of Computer Science, Central University of Rajasthan Ajmer, India

 

ABSTRACT: 

 In the rapid increasing usage of wireless sensor networks, fault tolerance is an exigent task to improve the overall performance. In this work we made an attempt using artificial bee colony approach to find data aggregation with fault tolerance in WSN to make an effective use of the existing resources. In this paper it is tried to apply the performance of the Q-MST based fault tolerance mechanism with the existing ant colony algorithms for data aggregation in WSN. We applied Q-MST for fault tolerance purpose; Ant Colony algorithm and PRIMS algorithm are used for generating data aggregation tree (MST). The quadratic minimum spanning tree (Q-MST) is an improved version of minimum spanning tree where ordered pair cost of distinct edges would be considered for implementing an alternate edge for the existing edge failure in MST.

 

Keywords: WSN, Data aggregation, Fault Tolerance, MST, Q-MST, Swarm Intelligence  

 

pdf

By | 2017-09-13T16:51:47+00:00 September 13th, 2017|General, Publications|0 Comments

IMPROVED TASK SCHEDULING MODEL WITH TASK GROUPING FOR COST AND TIME OPTIMIZATION

Shiva, Dr Anurag Jain

Department of Computer science and Engineering CEC Landran

 

ABSTRACT:

The existing models are based upon t[04] Shivahe quality of service (QoS) integration for the purpose of task prioritization and cost based scheduling for the client’s convenience, i.e. quick and minimized time based scheduling. The existing models are not found efficient enough in the different ways, which states the various shortcomings of the existing models. The existing model (Anton Belogazov, 2012) does not incorporates the concept of the task group aggregation (also known as task clustering) to segregate the tasks on the basis of their dependency. The dependent tasks can be grouped altogether to prepare the task chains which are scheduled sequentially, and in the parallel processing manner with other task groups belonging to the one task batch. In such way, the unnecessary burden can be reduced from the machines lying in the sleeping state to minimize the complexity of scheduling for processing of target task group (or cluster).

Keywords: Cloud scheduling, task scheduling, QoS, cloud cluster management

 

pdf

By | 2017-09-13T16:47:56+00:00 September 13th, 2017|General, Publications|0 Comments

SPAM E-MAIL DETECTION USING CLASSIFIERS AND ADABOOST TECHNIQUE

Nilam Badgujar 1, Namrata Chaudhari2, Ronit Chougule3, Shraddha Malve4

1,2,3,4Pune Institute of Computer Technology, Information Technology Department, Pune, India.

 

ABSTRACT:

The Internet has turned our existence upside down and has dramatically revolutionalized many different fields. Emails, social networking sites, surfing all have become a part and parcel of our lives. Emails are the fastest way to send information from one place to another. Whatever may be the work on the Internet, emails are involved everywhere. But, nowadays, emails are getting more prone to exploitation due to malicious attacks which include spam mails. Spam is flooding the whole internet with many copies of the same messages so as to force people who would not otherwise choose to receive such messages. So, in this study, Analyzing the performance of certain classifiers like Naive Bayes, Support Vector Machine, K-Nearest Neighbor and Decision Tree with the help of a dataset is done. These classifiers help in detecting spam emails with the help of keywords. Boosting grants power to these classifiers to improve their accuracy of prediction in most of the cases. Adaptive boosting also known as Adaboosting is used after the result of every classifier. This is implemented with the help of WEKA interface. The results have shown a comparison of all the three algorithms with boosting on the same dataset.

 

Keywords: Spam, Naive Bayes, SVM, KNN, Decision Tree, Adaboosting.

 

pdf

By | 2017-09-13T16:44:18+00:00 September 13th, 2017|General|0 Comments

PREDICTIVE ANALYTICS: AN APPLICATION PERSPECTIVE

Razeef Mohmmad1, Muheet Ahmed Butt2, Majid Zaman3

1PG Department of Computer Science, University of Kashmir, India

 2PG Department of Computer Science, University of Kashmir, India

3Directorate of Information Technology and Support Systems, University of Kashmir, India

 

ABSTRACT: 

 Predictive Analytics is a data science that helps in predicting the things after deep analysis of huge amount of data using predictive analytic techniques. Predictive analytics is a set of technologies and methods that uncovers relationships and patterns within large amount of data that can be used to predict behavior and events. Predictive analytics has wide application domain almost in every industry where the data is generated and stored. This research essay is aimed to provide a comprehensive overview of predictive analytics and its real world applications in various fields.

Keywords: Predictive Analytics, Predictor, Modeling, Customer Relationship Management (CRM)

 

pdf

By | 2017-09-13T16:41:10+00:00 September 13th, 2017|General, Publications|0 Comments

BUILDING SEMANTIC CONTEXT: AN APPROACH FOR LEARNING CONTEXT FROM THE WEB FOR FACILITATING TWITTER ANALYSIS

Nazura Javed1, B. L. Muralidhara2

1 Department of Computer Science and Applications, Bangalore University,Bangalore, India

2 Department of Computer Science and Applications, Bangalore University,Bangalore, India

 

ABSTRACT: 

 Tweets are cryptic and often laced with insinuation. Interpretation of user tweets cannot be done in isolation. Human beings can interpret the tweets because they possess the ‘Background/Contextual knowledge’. This knowledge enables them to understand the context of tweets and interpret the text. Emulating this interpretation ability in machines requires the machine to acquire ‘Contextual Knowledge’. In this paper we propose a novel technique of harnessing the contextual knowledge from the online sources and building a Labeled Context Corpus (LCC). Our contribution in this paper is a methodology that discovers trending topics and automates the construction of a LCC. Since most of the user tweets do not contain hashtags and associating them with a topic label is a challenge; LCC can be leveraged for analysis of tweets even when they lack reference URLs or hashtags. Mining of tweets can provide valuable insight into societal sentiment.

Keywords: [01] Nazura Javed

 

pdf

By | 2017-09-13T15:47:30+00:00 September 13th, 2017|General, Publications|0 Comments

A NEW APPROACH FOR SECURE OUT SOURCING OF KEY UPDATES IN CLOUD STORAGE

Mohd Mujaheed1, Md Ateeq Ur Rahman2

1M.Tech Scholar, CSE Department, SCET, Hyderabad

2Professor and Head, Department of CSE, SCET Hyderabad,

ABSTRACT:

In this paper, we concentrate on the best way to make the key overhauls as straightforward as could be expected under the circumstances for the customer and propose another worldview called distributed storage reviewing with certain outsourcing of key redesigns, In this world key plans can be securely outsourced to some approved gathering and along these lines the key-upgrade trouble on the customer will be kept insignificant. In particular, we influence the outsider inspector (TPA) in great to open examining outlines, let it assume the part of approved gathering for our situation and make it accountable for both the capacity reviewing and secure key upgrades for key-presentation resistance. In this worldview, key redesigns can be securely outsourced to some approved gathering, and subsequently the key-overhaul load on the customer will be kept insignificant. In particular, we influence the outsider evaluator (TPA) in numerous current open examining plans, let it assume the part of approved gathering for our situation, and make it accountable for both the capacity inspecting and the safe key upgrades for key introduction resistance. Recently, key exposure problem in the settings of cloud storage auditing has been proposed and studied. Existing solutions all require the client to update his secret keys in every time period, which may inevitably bring in new local, burdens to the client, especially those with limited computation resources such as mobile phones. In this Concept, we focus on how to make the key updates as transparent as possible for the client and propose a new paradigm called cloud storage auditing with verifiable outsourcing of key updates, the key updates can be safely outsourced to some the authorized party, and thus the key-update burden on the client will be kept minimal we formalize the definition and security model of this paradigm. The security proof of the performance simulation show that our detailed design instantiations are secure and efficient.

 

Keywords: Cloud storage, cloud storage auditing, key update, verifiability outsourcing computing,

 

pdf

By | 2017-09-20T10:29:34+00:00 September 13th, 2017|General, Publications|0 Comments

A STUDY OF DEDUPLICATION CLOUD STORAGE VIA DYNAMIC OWNERSHIP MANAGEMENT TECHNIQUE

Heba Salam1, Md Ateeq Ur Rahman2

1M.Tech Scholar, CSE Department, SCET, Hyderabad

2Professor and Head, Department of CSE, SCET Hyderabad,

ABSTRACT:

 Deduplication Cloud storage services is technique to reduce the amount of storage space an organization needs by saving just one copy of the data and pointing, commonly used to reduce the space and bandwidth requirements of services by eliminating redundant data and storing only a single copy of them. De-duplication is frequently effective when multiple users involve to use the same data to store on cloud storage, but it raises issues relating to security and ownership. However, many users are likely to encrypt their data before outsourcing them to the cloud storage to preserve privacy, but this hampers deduplication because of the randomization property of encryption. However, most of the times suffer from security flaws, since they do not consider the multiple changes in the ownership of outsourced data that occur frequently in a practical cloud storage service. In this paper, we propose a factious prose server-side de-duplication scheme for encrypted data This prevents data leakage not only to put an end to the validity or operation users though they previously owned that data, but also to true-but-curious cloud storage server. In addition, the proposed scheme guarantees data integrity against any tag inconsistency attack. Thus, security is enhanced in the proposed scheme. The efficiency analysis results demonstrate that the proposed scheme is almost as efficient as the previous schemes, while the additional computational overhead is negligible.

Keywords: [22] Heba Salam

 

pdf

By | 2017-09-05T17:07:25+00:00 September 5th, 2017|General, Publications|0 Comments
Load More Posts