Intrusion Detection System based on time related features and Machine Learning

The analysis of the behavior of network communications over time allows the extraction of statistical features capable of characterizing the network traffic flows. These features can be used to create an Intrusion Detection System (IDS) that can automatically classify network traffic. But introducing an IDS into a network changes the latency of its communications. From a different viewpoint it is possible to analyze the latencies of a network to try to identifying the presence or absence of the IDS. The proposed method can be used to extract a set of phisical or time related features that characterize the communication behavior of an Internet of Things (IoT) infrastructure. For example the number of packets sent every 5 minutes. Then these features can help identify anomalies or cyber attacks. For example a jamming of the radio channel. This method does not necessarily take into account the content of the network packet and therefore can also be used on encrypted connections where is impossible to carry out a Deep Packet Inspection (DPI) analysis.