S. H. Manjula

Association rule sharing model for privacy preservation and collaborative data mining efficiency

The disclosure of information and its misuse in Privacy Preserving Data Mining (PPDM) systems is a concern to the parties involved. In PPDM systems data is available amongst multiple parties collaborating to achieve cumulative mining accuracy. The vertically partitioned data available with the parties involved cannot provide accurate mining results when compared to the collaborative mining results. To overcome the privacy issue in data disclosure this paper describes a Key Distribution-Less Privacy Preserving Data Mining (KDLPPDM) system in which the publication of local association rules generated by the parties is published. The association rules are securely combined to form the combined rule set using the Commutative RSA algorithm. The combined rule sets established are used to classify or mine the data. The results discussed in this paper compare the accuracy of the rules generated using the C4.5 based KDLPPDM system and the C5.0 based KDLPPDM system using receiver operating characteristics curves (ROC).

EESOR: Energy Efficient Selective Opportunistic Routing in Wireless Sensor Networks

Opportunistic Routing in wireless sensor networks is a multi-hop routing. In this routing neighbors of a node overhear the transmission and form multiple hops from source to the destination for transfer of information. The set of neighbor nodes participating in the routing are included in the forwarder list in the order of priority. A node with highest priority is allowed to forward the packet it hears. This paper implements Energy Efficient Selective Opportunistic Routing (EESOR), reduces the size of forwarder list by applying a condition that the forwarding node is nearer to the destination. The path followed by acknowledgment packet follows opportunistic routing, assuring reliability of transmission and energy balancing. The simulated results obtained in NS2 simulator show that proposed EESOR protocol performs better than existing Energy Efficient Opportunistic Routing (EEOR) protocol in terms of average End-to-End delay, maximum End-to-End delay and Network Lifetime.

Elimination of Black Hole and False Data Injection Attacks in Wireless Sensor Networks

Wireless Sensor Networks (WSNs) are currently being used in a wide range of applications that demand high security requirements. Since sensor network is highly resource constrained, providing security becomes a challenging issue. Attacks must be detected and eliminated from the network as early as possible to enhance the rate of successful transactions. In this paper, we propose to eliminate Black Hole and False Data Injection attacks initiated by the compromised inside nodes and outside malicious nodes respectively using a new acknowledge scheme with low overhead. Simulation results show that our scheme can successfully identify and eliminate 100 % black hole nodes and ensures more than 99 % packet delivery with increased network traffic.

Secure Reputation Update for Target Localization in Wireless Sensor Networks

Wireless Sensor Networks (WSNs) are by its nature more prone to security attacks and data losses. Security and data privacy have become need of the day. The most challenging area in WSNs which needs security is target localization. In addition to this complexity we are here concentrating on acoustic sensor nodes which uses Particle Swarm Optimization (PSO) algorithm for Location Estimation. We propose Secure Reputation Update Target Localization (SRUTL) algorithm which addresses target localization and security issues viz., bad-mouth attack, Sybil attack, on-off attack and malicious node attack at different levels of target localization. Simulation results shows positive response in attack detection hence contribute in building a secure wireless sensor network.

ACTM: Anonymity Cluster Based Trust Management in Wireless Sensor Networks

Wireless Sensor Networks consists of sensor nodes that are capable of sensing the information and maintaining security. In this paper, an Anonymity Cluster based Trust Management algorithm(ACTM) is proposed which enhances the security level and provides a stable path for communication. It is observed that the performance of the network is better than existing schemes through simulation.

Chain Routing for Convergecast Small Scale Wireless Sensor Networks

Wireless sensor networks have many applications involving autonomous sensors transmitting their data to a sink placed in the network. A protocol by name Chain Routing for Convergecast Small Scale (CRCSS) Wireless sensor networks is proposed in this paper. The set of sensor nodes in the network send the data periodically to the sink located in the area of interest. The nodes who cannot reach sink in one hop choose one of the neighbours for forwarding the data to the sink by forming a chain of links. The selection of forwarding node and the waiting period before forwarding plays an important role in the protocol. The proposed CRCSS protocol exhibits improvement in energy spent per packet and latency per packet for a wireless sensor network as compared to ConverSS protocol for small scale wireless sensor networks. In CRCSS protocol energy spent per packet is independent of the network radius.

Improving the precision and recall of web people search using hash table clustering

Searching the people names on the web is a challenging query types to the web search engines today on the web. When a person name is queried, the returned result often contains web pages related to several distinct keywords that have the queried name. In this paper we have proposed a new approach to improve the precision, recall and F-measure metrics of the web search engine by removing the unwanted resulting web pages and the proposed method outperforms by 15.9 percent of F-measure improvement as compared to WWW2005 algorithm.