The technological advancement of the Internet of Things (IoT) devices in our world today has become beneficial to many users but has security issues that are left unattended. IoT devices have the ability to connect to other devices on the internet to transmit and share data from anywhere hence the need to secure these devices as they make our lives comfortable and easy. Research shows that about70% of IoT devices are easy to hack. Therefore, an efficient mechanism is extremely needed to safeguard these devices as they are connected to the internet. Therefore, this thesis proposal proposes a novel deep based anomaly detection model to predict cyber attacks on the IoT devices and learn new outliers as they occur over time. The Long Short- Term Memory(LSTM) is one of the efficient deep learning architectures that addresses the spatial and temporal information. Therefore, it could perform effectively in an anomaly detection model for IoT security. An intrusion detection dataset is fed into the LSTM model to train and test the anomaly detection model. The model performance to detect an anomaly will be based on accuracy, precision and recall. The proposed LSTM based model performance will be analyzed and compared to the state-of-the-art deep learning-based anomaly detection for IoT devices. The proposed model promises a significant accuracy improvement due to the LSTM learning methodology that perfectly fits to solve the anomaly detection for IoT devices.
Author: Sylvia Azumah