Abstract:

The Internet of Things (IoT) is a network of millions of devices fitted with sensors and actuators which have either wired or wireless connectivity in order to communicate information between them. IoT has shown tremendous growth in the past ten years. By 2020, it is anticipated that there will be over 25 billion devices that would have been connected. In the coming years, these devices will emit much higher amount of data. In addition to volume, IoT devices are known to generate large data sets in various modalities with different qualities of data as defined by velocity regarding time and location dependency. Under such a scenario, the ML algorithms will play an important role in ensuring biotechnology-based security and authentication coupled with anomaly detection to further improve the usability and security of the IoT systems. On the other hand, attackers often exploit learning algorithms to exploit vulnerabilities in intelligent IoT-based systems. Motivated by this, in this article, we propose to improve the security of IoT devices through spam detection using ML. For this objective, spam detection in IoT using a machine learning framework is proposed. In this framework, five ML models are tested with different metrics, including a large collection of input feature sets. Each model computes a spam score by taking into account refined input characteristics. This score reflects the reliability of an IoT device under different parameters. The proposed technique is validated by using the REFIT smart home dataset. The achieved results show the efficiency of the proposed scheme in comparison with other existing schemes.

Description:

.