Detection of CIFA using SMOTEBoost and LSTM in NDN

Abstract

The efficient forwarding mechanism of Named Data Networking (NDN) has attracted many scholars' attention. However, the threat of network attacks still exists in NDN, just like other networks. Among them, the Collusive Interest Flooding Attacks (CIFA) has an extremely significant attack effect in NDN. CIFA attackers send malicious interests in the form of periodic pulses with the help of collusive producers, which will affect the quality of NDN network services. Through simulating CIFA in ndnSIM, the network traffic features under CIFA and normal network state are extracted, including PIT occupancy rate, throughput, satisfaction of Interests and received data packets. Furthermore, a detection method using SMOTEBoost and LSTM is proposed by making in-depth analysis of the impact of CIFA based on the CIFA attack features. Finally, experiments show that the proposed detection method achieves 99.2 % detection rate, 0.5 % false alarm rate and 0.6 % missed alarm rate, which is far superior to other methods.