A bootstrap algorithm for matrix factorization based network coordinate system

Internet Coordinate System, ICS is vulnerable while facing malicious attacks. Traditional works can mitigate the negative influence caused by malicious attacks but the computing processes take too much computation resource. Thus the security awareness method of ICS is taken into consideration. Through which nodes can aware whether it is suffering attack or not, and decide how to deal with the situations they faced on. To identify the current security situation, this paper explores an unsupervised learning method which utilizes the spectral radius of the estimation error matrix of the time series. Our method is evaluated on the Vivaldi algorithm against many attack models. In our simulations malicious nodes perform all known attack sequences against ICS. Experiments show that this method can recognize the situation of the networks swiftly with low computation cost.