IMPROVED PREDICTION STRATEGY USING PARTICLE SWARM OPTIMIZATION BASED ARTIFICIAL NEURAL NETWORK (IPS-ANN) CLASSIFIER FOR MALICIOUS NODE DETECTION FOR HYBRID P2P NETWORKS

E Banumathi

doi:10.26483/ijarcs.v9i1.5423

Abstract

Mining data from the hybrid P2P networks is an ever demanding task for detecting malicious behavior of nodes. This research work aims to propose an improved prediction strategy using particle swarm optimization based artificial neural network (IPS-ANN) classifier. From the extensive literature study it is identified that only very few literatures are there for detecting malicious node activities. The simulation are carried out using MATLAB by configuring 1000 nodes with three different attack scenarios namely collusion attack, Sybil attacks and file polluter attacks. Detection accuracy, false positive rate and false negative rate are taken as the performance metrics for evaluating the efficiency of the proposed classifier and also compared with the existing mechanisms. Results proved that the proposed IPS-ANN classifier better than that of PeerMate , SMART , Outlier mining mechanisms.

Highlights

The Peer-to-Peer (P2P) systems are able to rapidly supply information exchange and sharing service by utilizing resources from participating peers, so have of late gained noteworthy recognition [2]
Based on the finding that each malicious peer has the specific characteristic of outlier [1], this research work aims to present an improved prediction strategy using particle swarm optimization based artificial neural network (IPS-ANN) classifier for malicious node detection for hybrid P2P networks
Without the loss of generality, commonly used false positive rate (FPR, i.e. the ratio of peers that are normal but considered as malicious to all the normal peers) and false negative rate (FNR, i.e. the ratio of peers that are malicious but considered as normal to all the malicious peers) as the criterion [20, 21] to assess the performance of our model.True positive is the normal activity correctly identified as normal activity

Summary

Introduction

The Peer-to-Peer (P2P) systems are able to rapidly supply information exchange and sharing service by utilizing resources from participating peers, so have of late gained noteworthy recognition [2]. The open and anonymous nature of these systems often leads to a lack of responsibility for the content a peer puts on the network, opening the door to abuses by malicious peers [2]. A major challenge for a large-scale P2P system is to establish a scalable trust framework without any preexisting trust relationship among peers and without involving trusted third parties or authorities In a hybrid P2P network, all the nodes can be broadly grouped as super node and ordinary node [3]. There are several ordinary nodes with which the super node forms a subnet. The super node is responsible for managing the interaction data among the ordinary peers. All the super nodes are in charge of managing the interaction data among the subnets. Based on the finding that each malicious peer has the specific characteristic of outlier [1] , this research work aims to present an improved prediction strategy using particle swarm optimization based artificial neural network (IPS-ANN) classifier for malicious node detection for hybrid P2P networks

Objectives

Results

Conclusion