A privacy-preserving botnet detection approach in largescale cooperative IoT environment

With the development of Internet-of-Things (IoT), our modern life has been greatly facilitated, while an exponentially growing number of vulnerable devices also breed a wonderful ground for botnet controllers,. However, existing detection approaches developed for individual traditional network area neglect cross-area privacy issue and resource restraint nature of IoT network and therefore impede their effectiveness of mitigating IoT botnet. In this work, we present a lightweight and privacy-preserving system, namely PPBotHunter, to detect botnet across multiple network areas. PPBotHunter implements a fuzzy matrix algorithm to retrieve effective bot similarity computation while ensuring a high privacy degree. This algorithm is designed based on a privacy-preserving scalar product computation technique (PPSPC) which enables PPBotHunter to be lightweight yet efficient. We utilize only time series feature to build the fuzzy matrices, which further improve the compatibility, energy-efficacy and resistance against heterogeneity. The theoretical analysis and detailed simulations illustrate the efficacy and effectiveness of our proposed botnet detection system.