Abstract
The successful applications and diversified popularity of the Internet of Things (IoT) present various advantages and opportunities in broad characteristics of our lives. However, unfortunately, the IoT is allied with several types of defenseless attacks and illegitimate exploits. Security specialists specify voluminous threats imposed by the IoT devices in various aspects. Therefore, security and intrusion detection have constantly been growing areas of distress for any field of IoT research. This paper introduces a new hybrid ensemble hyper-tuned model (i.e., Catboost) that efficiently recognizes IoT sensor attacks and anomalies. The hyper-parameters are optimized with Bayesian optimization to develop security-based models effectively. The significant contributions of this work are the design of an intelligent model-based security framework based on the advanced ensemble learning Catboost model for detecting malicious IoT activities in the IoT network, the use of a Bayesian optimization approach to find an optimal set of Catboost hyper-parameters, and evaluate the model with a new real dataset (DOS2DOS) from a large-scale IoT network. The performance of the proposed model is compared with other state-of-the-art approaches, and the experimental results are evident towards a high detection rate of 99.9%.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call