Data Mining and Knowledge Engineering

P. Rukmini Devi ¹, N. Mohan ², V. Praveen Kumar ³, A. Nageswara Rao ⁴

Affiliations
1 PVP Siddartha Institute of Engineering and Technology, University of JNTUK, IN
2 Pragati Engineering College, University of JNTUK, IN
3 Central University of Hyderabad, IN
4 Dept. CSE, Pragati Engineering College, University of JNTUK, Surampalem, IN

Abstract

This paper presents a variation of Fuzzy c-Means (FCM) algorithm that provides image clustering. The proposed algorithm incorporates the local spatial information and gray level information in a novel fuzzy way. The new algorithm is called C-Means with Fuzzy Local Information (CMFLI). CMFLI can overcome the disadvantages  of known fuzzy c-means algorithm and at the same time enhances the clustering performance. The major characteristic of CMFLI is the use of fuzzy local information (both spatial and gray level) similarity measure, aiming to guarantee noise insensitiveness and image detail preservation. Furthermore, the proposed algorithm is fully free of empirically adjusted parameters (a, λg , λs etc.) incorporated into all other fuzzy c-means algorithms proposed in the literature. Experiments performed on synthetic and real world images show that CMFLI algorithm is efficient,providing robustness to noisy images.

Keywords

Clustering, Fuzzy C-Means, Fuzzy Constraints, Graylevel Constraints, Image Segmentation, Spatial Constraints.

Full Text

     

L. Ramesh ¹, V. Praveen Kumar ¹, K. Vijay Kumar ¹

Affiliations
1 Department of Computer Science & Engineering, P.V.P. Siddhartha Institute of Technology, Vijayawada, IN

Abstract

Peer-to-peer (P2P) networks are vulnerable to peers who cheat, propagate malicious code, leech on the network, or simply do not cooperate. The traditional security techniques developed for the centralized distributed systems like client-server networks are insufficient for P2P networks by the virtue of their centralized nature. The absence of a central authority in a P2P network poses unique challenges for reputation management in the network. These challenges include identity management of the peers, secure reputation data management, Sybil attacks, and above all, availability of reputation data. In this paper, we present a cryptographic protocol for ensuring secure and timely availability of the reputation data of a peer to other peers at extremely low costs. The past behavior of the peer is encapsulated in its digital reputation, and is subsequently used to predict its future actions. As a result, a peer's reputation motivates it to cooperate and desist from malicious activities. The cryptographic protocol is coupled with self-certification and cryptographic mechanisms for identity management and countering Sybil attack. We illustrate the security and the efficiency of the system analytically and by means of simulations in a completely decentralized Gnutella-like P2P network.

Keywords

Like A Certificate Peer-To-Peer Networks, Distributed Systems, Security, Reputations and Identity Management.

Full Text