anomaly detection fog (adf): a federated approach for internet of things

Heterogeneous data models and resource constraints are the challenging issues of anomaly detection in internet of things. due to these issues and the complexity of conventional anomaly detection methods, it is necessary to design an anomaly detection approach with iot-specific concerns. this paper presents a framework for anomaly detection specially designed for iot called anomaly detection fog(adf). adf uses network slicing to present a federation of heterogeneous fog clusters. federated fog clusters collaborate with each other via anomaly directives (heterogeneous context abstracts) for context-aware and application-independent anomaly detection. evaluations show that adf has a higher detection accuracy by detecting 95% of false alarms in comparison to conventional anomaly detection methods. also, adf reduces energy consumption by 40%, moreover, it reduces communication overhead and detection latency by preventing cloud offloading.