Feature selection for interest flooding attack in named data networking

摘要

Interest flooding attack (IFA) is one of the most severe attacks in named data networking (NDN). Due to IFA, the pending interest table (PIT) of NDN routers get filled with entries of malicious requests making it unavailable for legitimate users. Statistical approaches detect IFA using one or two features, but they cannot handle the variation of more than two features. To overcome this limitation a machine learning-based model has been proposed for the detection of IFA. First, we model IFA and collect features. Next, we select the most prominent features based on information gain-based ranking. Last, we use these features for the detection of IFA using machine learning approaches. Experimental results show that machine learning-based approaches perform better than statistical approaches regarding accurate detection.