Artificial Intelligence Internet of things

• In order for the anomaly detection system to be able to characterize normal patterns and create a model of the normal behavior, wide-ranging training sets of the normal system activities are needed. Any change in the system’s normal patterns must lead to necessary update of the knowledge base. • If the detection and prevention system inaccurately classifies a legitimate activity as a malicious one, the results can be very unfortunate since it will attempt to stop the activity or change it. • An intrusion detection system, no matter how efficient, may be disabled by attackers if they can learn how the system works. • In heterogeneous environments there is also an issue of integrating information from different sites. • Another problem involves supplying intrusion detection systems that will conform to legal regulations, security requirements and/or service-level agreements in real world.