Federated IoT Botnet Detection
Event Website
https://documentcloud.adobe.com/link/track?uri=urn:aaid:scds:US:1591efe4-e33b-49d6-b469-4aa3996c0df7
Document Type
Event
Start Date
3-12-2020 5:00 PM
Description
Internet of Things (IoT) devices are mass produced, often heterogeneous in nature, updated infrequently, and can remain out of sight on a home or office network for extended periods. Security and privacy are two of the many ongoing research and operational challenges in IoT. To address potential threats to IoT devices, deep learning-based solutions have been largely utilized in recent years. However, as big data is transferred under traditional centralized methods, it is at risk for data privacy issues because the data can be captured. In this paper, we propose a federated-based solution employing a deep autoencoder to ensure that data privacy is maintained and provides a way to detect botnet attacks using on-device decentralized data. Through the suggested federated option, privacy is addressed as data is made more secure since it is not transferred out of the device at all. Instead, the computation itself is brought to where data is born (e.g. the edge layer) providing the sought after results of a traditionally centralized machine learning technique, with the added benefit of data security. The results from our proposed model give up to 98\% accuracy of anomaly detection when using many features such as source IP, MAC-IP, and destination IP channel for training. The comparison analysis of the results for a traditionally centralized format, versus our decentralized approach, justifies a significant improvement in the accuracy rate of attack detection.
Federated IoT Botnet Detection
Internet of Things (IoT) devices are mass produced, often heterogeneous in nature, updated infrequently, and can remain out of sight on a home or office network for extended periods. Security and privacy are two of the many ongoing research and operational challenges in IoT. To address potential threats to IoT devices, deep learning-based solutions have been largely utilized in recent years. However, as big data is transferred under traditional centralized methods, it is at risk for data privacy issues because the data can be captured. In this paper, we propose a federated-based solution employing a deep autoencoder to ensure that data privacy is maintained and provides a way to detect botnet attacks using on-device decentralized data. Through the suggested federated option, privacy is addressed as data is made more secure since it is not transferred out of the device at all. Instead, the computation itself is brought to where data is born (e.g. the edge layer) providing the sought after results of a traditionally centralized machine learning technique, with the added benefit of data security. The results from our proposed model give up to 98\% accuracy of anomaly detection when using many features such as source IP, MAC-IP, and destination IP channel for training. The comparison analysis of the results for a traditionally centralized format, versus our decentralized approach, justifies a significant improvement in the accuracy rate of attack detection.
https://digitalcommons.kennesaw.edu/cday/Fall/graduateresearch/3