Computer Science And Engineering

In the field of network security, the task of processing and analyzing huge amount of Packet CAPture (PCAP) data is of utmost importance for developing and monitoring the behavior of networks, having an intrusion detection and prevention system, firewall etc. In recent times, Apache Spark in combination with Hadoop Yet-Another-Resource-Negotiator (YARN) is evolving as a generic Big Data processing platform. While processing raw network packets, timely inference of network security is a primitive requirement. However, to the best of our knowledge, no prior work has focused on systematic study on fine-tuning the resources, scalability and performance of distributed Apache Spark cluster (while processing PCAP data). For obtaining best performance, various cluster parameters like number of cluster nodes, number of cores utilized from each node, total number of executors run in the cluster, amount of main-memory used from each node, executor memory overhead allotted for each node to handle garbage collection issue, etc., have been finetuned, which is the focus of the proposed work. Through the proposed strategy, we could analyze 85GB of data (provided by CSIR Fourth Paradigm Institute) in just 78 seconds, using 32 node (256 cores) Spark cluster. This would otherwise take around 30 minutes in traditional processing systems.

Support Vector Machines (SVM) were originally designed for binary classification. Later it was extended to multi-class classification. Various methods have been proposed to construct a multi-class aclassifier by combining binary classifiers. As it is computationally more expensive to solve multi-class problems comparison of these methods using large scale problems have not been considered seriously. Especially for methods solving multi-class SVM in one step, a much larger optimization problem is required. So up to now the experiments are confined to small datasets. In this paper we give decomposition implementations for two such all-together methods. We then compare their performance with three methods based on binary classifications: "One-against-all" , "One-against-one" and DAG SVM. Our experiments indicate that one-against-one and DAG methods are suitable for practical use than other methods. Results also show that for large problems methods by considering all data at once in general need fewer support vectors.

This paper deals with the optimized Multi class SVM classifier (OMSC) with Named entity extraction in cloud environment. The proposed OMSC handles with scheduling workflow in cloud computing where the data and files are transferred between the participants based on the different set of rules, with having the additional advantage of rules formation capability in which it is going to follow 22 rules templates. This has shown an improved performance against the traditional Multi class SVM classifier. The average f-score for the tested data sets is 81.04 % when compared to the existing classifier it is improved sign. The time complexity is decreased and as per the scheduling is concerned the execution time and response time is improved.

looking at the training computational time factor. This paper presents mathematical foundation of multi-class support vector machines and the evolution of the problem formulation aiming to reduce the training time by improving the optimization algorithms. We then have a look at a new cutting plane algorithm that accelerates the training process while achieving good out-of-the-box performance in linear time. The cutting plane algorithm is first implemented as a standalone java executable in order to study its effectiveness in reducing the computational time.

Cloud computing is a model which can be used in any domain expanding from large scale enterprise to start ups to educational institutions. In the traditional way of building a site, people go for external/shared web hosting, these have few issues. In this paper We will demonstrate a site on cloud and achieve high availability and low latency through load balancing on cloud servers. We used cloud computing for hosting the site and maintain the web traffic on our website. We are self hosting the site on a server which is on cloud. The actual role of cloud in our research is to make sure our website is available and always maintains low latency regardless of web traffic. To achieve this we will have a load balancer again on cloud which takes the traffic into account and Scales the servers to manage the load which is nothing but scalability. This was not possible in traditional web hosting and servers. The end product of my research is a highly available, scalable, complete college network site with approximately no investment.

Dealing with high dimensional data of the form p>n for multivariate analysis of missingness is very complicated. It arises in many fields mainly in social science, economics and medical study; genome is an example for that where is to mention that samples are very less compared to study elements nothing but variables. The analysis is a combination of large covariate vectors with response and non-response effects of unknown functional form related to response variable of interest. Thus, there is a need for regularized regression models, with effect of smoothing parametric method to do this in this work combine regularization by incorporating different types of covariates. Although regularization approaches fits to framework but the computation high demands in high dimensional analysis they also rely on penalized estimation. The solution is to implement regularization in iteration based smoothing approaches to fit such analysis. The proposed algorithm called Iterative Bayesian Addi...

Communication in a Mobile ad hoc network is accomplished in a multi-hop fashion wherein the nodes themselves act as routers as well as source. MANET is vulnerable to a variety of insider as well as outsider attacks because of it's inherent characteristics like dynamically changing topology, limited power and no centralized authority to monitor the behaviour of the nodes . Most of the secure routing protocols focus on the control plane attacks targeting the different elements of the routing protocol but data plane attacks are more difficult to handle for which we need to ensure secure data forwarding since certain adversarial nodes may launch a number of attacks on the data transmission by simply dropping the packets without forwarding them. The current paper is an extension to our earlier work where we addressed one of the most difficult data plane attacks namely, the packet dropping attack. In our earlier work, we had proposed a secure hybrid routing protocol which combats the packet dropping attack carried on by an individual / colluding adversaries and elaborated upon the first two phases of Secure Least Cost route establishment and Detection of individual node / colluding nodes maliciously dropping packets. Once a node has been identified as an adversary, we need to work upon the remedial action to prevent the future routes from involving adversarial nodes. In the current paper, we elaborate upon the last two phases namely: Punishing the adversarial nodes upon misbehaviour detection and Propagation of information about node misbehaviour and good behaviour. The current paper extends our earlier work by considering an additional colluding adversarial model consisting of consecutive nodes on the route from source to destination acting as colluding adversaries. Finally the protocol results in the establishment of a route including those nodes with good packet forwarding behaviour.

Most of the existing security mechanisms for detecting the packet droppers in a mobile ad hoc network generally detect the adversarial nodes performing the packet drop individually wherein false accusations upon an honest node by an adversarial node are also possible. In this paper, we propose a security mechanism to detect those nodes performing packet dropping either on their own individually or in collusion such that they cannot evade detection and no false accusations are possible. The detection of adversarial nodes is done by the source node through the analysis of the reports submitted by all the intermediate nodes on the source to destination paths. The composition of the report from each of the intermediate nodes involves certain pre-computed hash values which act as acknowledgments from each receiver node (successor) to the forwarder node (predecessor) and also a secure proof through which each intermediate node claims the packets which have been received within a communication session. The proposed mechanism has minimum communication and computational overhead since the secure proof is based upon a hash computation and report submission is secured through symmetric cryptographic primitives. The report analysis process ensures that evading the detection is not possible even in collusive adversarial model.

Trust management frameworks play a very important role in securing the mobile ad hoc networks against various insider attacks that could occur during data forwarding. The success of a trust management framework greatly depends upon the proper design of each of its major components including the direct trust computation component as well as the indirect trust computation component. Specifically, the indirect trust computation component should be robust to handle the dishonest recommendations. The current paper shows the application of a trust model involving a robust indirect trust computation component called as RecommFilter which has been proposed in our earlier work. It can overcome the various attacks caused by dishonest recommenders. The application involves the integration of the trust model with a routing protocol based upon a reliability measure called as Path Allegiance metric (PAM) which is a cumulative value obtained through the trust values of the on-path nodes upon each other. Experimental results show that the proposed scheme along with PAM routing protocol is robust to different dishonest recommendation attacks and accurate in the detection of dishonest recommenders.

Malicious packet drop attack over the data plane in a MANET involves malicious nodes dropping the data packets after the route formation. In this paper, a security mechanism has been proposed to detect those nodes which exhibit the malicious behavior by dropping the data packets during the data transmission phase after participating in the route establishment phase without exhibiting any malicious behavior. The detection is done based upon a trust management framework employing the Dempster Shafer Theory to represent the trust. The design of the trust management framework has been covered in earlier works and the current work focuses upon its application for the design of a novel security mechanism. Trust is computed based upon the forwarding behavior represented by acknowledgement reports submitted to the source node. The composition of the report ensures that the source node can verify its authenticity. Trust updates upon intermediate nodes are done by the source node at the end of a session which facilitates the secure route formation through the proposed mechanism. The efficiency and accuracy of the proposed security mechanism is validated using the network simulator ns2 and the experimental results show that the proposed mechanism outperforms the other schemes.

Mobile ad hoc networks are open networks which are infrastructure less, decentralized and formed spontaneously resulting in a lot of security vulnerabilities. Reliable data delivery from a source node to a destination node requires a path involving nodes which collaborate to forward the data packets which requires the nodes to assess the forwarding behavior of each other. The paper presents an uncertainty reasoning based trust management framework to compute a trust metric which is integrated into a routing protocol based upon a reliability measure called as Path Allegiance metric (PAM) which is a cumulative value obtained through the trust values of the on-path nodes upon each other. Performance analysis through simulations proves the efficiency of the proposed Path allegiance metric based routing protocol (PAMRP) in forming the trusted/secure route.

Dishonest recommenders can have an impact upon the trust management framework of a mobile ad hoc network through the launch of an on-off attack pattern which involves periodic good behaviors so as to evade detection and avoid the trust degradation. The current paper presents the design of a recommendation trust update module which aims to achieve robustness against such on-off dishonest recommenders by considering the consistency of behaviors to dynamically decide the fading/aging factor which determines the weight to be given to past behaviors and current behaviors so as to compute an overall trust estimate which reflects the true nature of the node. The robustness of the scheme and the accuracy of trust computation in the presence of on-off dishonest recommenders are demonstrated through simulation results.

—A mobile ad hoc network is a spontaneous self-organized infrastructure-less network wherein the networking activities like routing and data transmission are carried on by the nodes themselves in a collaborative manner. However, since nodes are resource-constrained with limited battery power, few nodes may be selfish which expect services from other neighbouring nodes but refuse to provide any service to it's neighbours. More specifically, the selfish nodes drop the packets belonging to some other node instead of forwarding them to the next hop on the route. A number of mechanisms have been proposed to defend against packet drop attacks carried out by an individual malicious node. Such mechanisms are random audit based which cannot detect collaborative packet drop attack wherein the attack is carried out cooperatively by colluding adversaries for which the defense mechanism becomes still complicated. We propose a mechanism to detect colluding adversaries which collectively carry out packet drop attack. Keywords—Mobile Ad hoc Networks (MANETs), Colluding adversaries, Packet Drop Attack, Audit based detection.

Machine learning (ML) is a rising field. Machine learning is to find patterns automatically and reason about data.ML enables personalized care called precision medicine. Machine learning methods have made advances in healthcare domain. This paper discuss about application of machine learning in health care. Machine learning will change health care within a few years. In future ML and AI will transform health care, but quality ML and AI decision support systems (DSS) Should Require to address the problems faced by patients and physicians in effective diagnosis. 

With the technological advances in ICT people are migrating to urban areas. Urbanization becomes a global phenomenon. By 2030, 60% of the population will live in urban. Development of smart cities is inevitable. Cyber Physical Systems (CPS) supports our day to day activities. With the emergence of CPS is becoming more important to smart cities. Smart cities can be viewed as large scale implementations of CPS. This paper throws some light on the concepts of CPS, Smart and the various challenges posed by CPS in implementing in smart cities .This paper also describes research directions for CPS .  

The applications of a mobile ad hoc network (MANET) range from military and disaster response to commercial and civilian environments. The intrinsic characteristics like lack of infrastructure, self-organizing nature and deployment flexibility contributes to its usefulness but few other features like open media for communication and topological changes on-the-fly makes it vulnerable to security attacks. The focus of the current work is security mechanisms for a MANET against the various insider attacks targeting the data forwarding operations. It is accomplished through a trust management framework and a secure routing protocol relying upon the TMF to form the routes with most trusted nodes. The protocol computes a novel reliability metric termed as Path Allegiance metric (PAM) which acts as a selection criteria for route selection. The performance evaluation involves simulation results which prove the robustness of the proposed TMF and the effectiveness of the PAM routing protocol ...

The early diagnosis of diabetes helps in avoiding the major risks associated with the disorder. The proposed research involves the design of a machine learning pipeline which generates the most representative feature subset of minimal size that predicts the onset of Diabetes with highest accuracy. It employs a novel diabetes dataset which is gender-neutral and representative enough unlike the well-known PID dataset. The machine learning pipelines involve multiple feature engineering pipelines to generate a reduced feature subset which is fed into multiple heterogeneous classifiers. The feature engineering involves feature selection as well as feature extraction. The former uses the ANOVA filter and Crow Search Optimization algorithm. The latter employs the Singular Value Decomposition. The classification is performed on the preprocessed dataset using a wide range of heterogeneous classifiers like Naive Bayes', Logistic Regression, K-Nearest Neighbor, Decision Trees, Support Vector Machine, Random Forest, AdaBoost, and GradientBoost as base learners followed by their stacking ensemble. The performance evaluation of each machine learning pipeline is done through Repeated Stratified K-fold Cross Validation using the metrics of accuracy, precision, recall, F1 Score and area under Receiver Operating Characteristic curve. For each pipeline, the number of features in the preprocessed dataset varies and the highest accuracy of 98.4% is achieved with Crow Search algorithm through a stacking ensemble of multiple heterogeneous classifiers. A comparative analysis with a recent related work on the same dataset shows that the proposed feature engineering pipelines with the same set of classifiers outperform with improved accuracy using a feature set of reduced size.