Impact of an attack on a network executing distributed computing

Impact of an attack on a network executing distributed computing

Článek Scopus

In this paper, we examine the effect of an attack on a network executing the In this paper, we examine the effect of an attack on a network executing the distributed algorithm average consensus. We assume that only one attack is committed during the process of reaching the convergence. We examine how intensively an attack affects features of the average consensus algorithm. We define two parameters: the number of additional iterations and the change of the final value. Then we examine how they are changing when the following parameters are being changed: the number of iteration during which an attack is committed, an attacker's internal value, the initial values, the range of the initial values. At the end, we examine how the position of the attacked element affects the impact of an attack on the network. Firstly, we perform experiments in an example network containing 24 densely placed elements and whose topology is randomly generated. We decide to apply TDMA as a method to share a transmission medium. Then we execute another experiment in which we examine how intensive an attack is when the size of a network changes. This paper is motivated by the publications where a potential failure of a node such as a dead node, a misbehaving node etc. significantly affects the whole computation process. In contrast to the previous works, we assume the presence of an attacker who is aware of the weaknesses of distributed computing.

In this paper, we examine the effect of an attack on a network executing the In this paper, we examine the effect of an attack on a network executing the distributed algorithm average consensus. We assume that only one attack is committed during the process of reaching the convergence. We examine how intensively an attack affects features of the average consensus algorithm. We define two parameters: the number of additional iterations and the change of the final value. Then we examine how they are changing when the following parameters are being changed: the number of iteration during which an attack is committed, an attacker's internal value, the initial values, the range of the initial values. At the end, we examine how the position of the attacked element affects the impact of an attack on the network. Firstly, we perform experiments in an example network containing 24 densely placed elements and whose topology is randomly generated. We decide to apply TDMA as a method to share a transmission medium. Then we execute another experiment in which we examine how intensive an attack is when the size of a network changes. This paper is motivated by the publications where a potential failure of a node such as a dead node, a misbehaving node etc. significantly affects the whole computation process. In contrast to the previous works, we assume the presence of an attacker who is aware of the weaknesses of distributed computing.

parallel distributed computing, average consensus, wireless sensor networks

parallel distributed computing, average consensus, wireless sensor networks

KENYERES, M.; KENYERES, J.

2017

01.06.2016

1454-234X

UPB Scientific Bulletin, Series C: Electrical Engineering

78

2

Rumunsko

35

52

18

https://www.scientificbulletin.upb.ro/rev_docs_arhiva/fulle8b_932467.pdf