Transfer Learning Approach for Detecting Keylogging Attack on the Internet of Things

The Internet of Things is an innovation that brings together an imagined space and the real world on one platform. With the rapid growth of IoT devices, the lack of standards has resulted in many unsecured devices connecting to networks, leading to an increase in cyberattacks on IoT, especially keylogging attacks. Intrusion detection systems (IDS) have traditionally relied on machine learning techniques, particularly deep learning, to detect attacks. However, these methods often require a large amount of labeled data to train, which is often unavailable for IoT networks. In this paper, we propose a transfer learning approach (TL) based on keylogging attacks where labeled data is sparse and unbalanced. We evaluated the proposed approach on the keylogger _detection dataset to show its effectiveness and compare it to current IDS approaches. The main findings of the experiments are as follows: The proposed approach has a high level of accuracy and a low level of false predictions (FPR). It performs better than traditional deep learning (DL) based IDS.

Перевод пока недоступен