Network attacks detection using fuzzy logic

Abstract

The aim of research is to increase the network attack detection accuracy by means of fuzzy logic. This paper considers an approach to intrusion detection using fuzzy logic. The approach is based on network monitoring of the variables characteristic of different network anomalies, such as ratio of the incoming traffic to the outgoing, packet size, etc. Every type of menace is characterized by a vector of fuzzy values describing the network state when this menace is present. These vectors constitute the fuzzy rule matrix. This article proposes computation of the integral indicator of the presence of any menace using the rule matrix. Keywords-computer networks, distributed denial of service, network security, fuzzy logic, intrusion detectionWorld-wide spread of networking technologies leads to the fact that network attacks are one of the most efficient means of malicious activity with the purpose of unauthorized access to the information, leading information systems out of operation, blocking or obstructing the access to network resources for legitimate users. In the conditions of active technology progress and increasing of malefactor skills ensuring of network security should be a continuous process.

Intrusion detection based on fuzzy logic for wireless body area networks: review and proposition

“…These parameters are used for calculating a jamming index (JI) to indicate the jamming level (high, medium or low). In the same way, Levonevskiy et al have proposed a method that based on a vector of fuzzy values to detect DDoS attacks [28] by monitoring the network characteristics as packet size and ratio of received traffic to outgoing traffic. The theory of fuzzy logic is implemented also [29] to predict exhaustion attacks over the IEEE 802.15.4 MAC layer, using FCM clustering.…”

“…However, some method as [14], [19], [27], [28] require a complicated calculation on the nodes, which is costly in terms of energy. The technique [27] makes the member of nodes send periodically data for checking the values of the four inputs parameters (BPR, PDR, RSS, and CCA).…”