IoTInDet: Detecting Internet of Things Intrusions with Class Scatter Ratio and Hellinger Distance Statistics

Technology evolution has attracted hackers to launch cyberattacks on Internet of Things (IoT) devices. Smart devices must be protected against these attacks by identifying anomalies in IoT communications. Intrusions in IoT traffic can often be detected using statistical methods. The problem with current statistical intrusion detection methods is that these methods suffer from curse of dimensionality and are unable to identify intrusions with a low false positive rate when choosing to change the statistical assumptions. In order to deal with these issues, a method called IoT Intrusion Detection (IoTInDet) is proposed. The IoTInDet identifies intrusions based on how well a set of features correlates with one another after selecting intrusion relevant features. Class Scatter Ratio (CSR) selects the features by calculating the weights of the features and ranking them. Using the Hellinger Distance Chart (HDC), the correlation between the chosen features is calculated. By calculating the HDC of the new traffic with the produced IoT normal traffic description, the intrusion is discovered. It is clear from the experiments on benchmark UNSW Bot-IoT dataset that the proposed IoTInDet method is providing promising results.

Automated summary

The evolution of technology has led to an increase in cyberattacks on IoT devices. Statistical methods can be used to detect intrusions in IoT traffic, but current techniques suffer from the curse of dimensionality. To address this issue, a method called IoT Intrusion Detection (IoTInDet) is proposed, which identifies intrusions by selecting relevant features and calculating their correlation.