Deep Learning for Botnet Attack Detection

An effective technique for identifying botnet attacks is deep learning (DL). Yet, the amount of memory space and network traffic that is needed is typically substantial. Therefore, it is extremely difficult to implement the DL approach in IoT devices with little memory. Using dimensionality reduction approaches, we lower the feature dimensionality of large-scale IoT network traffic data. The most pertinent publicly accessible dataset for network-based botnet attack detection in IoT networks is the Bot-IoT dataset. With Bot-IoT, millions of samples of botnet attack traffic were included. The four IoT botnet scenarios that these attack traffic samples fall under are DDoS, DoS, reconnaissance, and information theft. The most common method for reducing the dimensionality of a feature set is to apply a linear or non-linear transformation approach.