Abstract

Since security threats in IoT-enabled smart cities may not appear clear and present to detection mechanisms, efforts have been made to use artificial intelligence methods for anomaly detection. Anomaly detection has been performed using unsupervised learning approaches (Autoencoders, GANs, One Class SVMs) in turn, with these instances considered security threats. In addition, an element for patches and traffic redirection in real time is included in the framework. Results show that the AI detection in general has much more security resilience, decreasing possible attack vectors. This makes the integration of various features like AI, blockchain, and IDS for a solid IoT security a must. Since security threats in IoT-enabled smart cities may not appear clear and present to detection mechanisms, efforts have been made to use artificial intelligence methods for anomaly detection. Anomaly detection has been performed using unsupervised learning approaches (Autoencoders, GANs, One-Class SVMs) in turn, with these instances considered security threats. In addition, an element for patches and traffic redirection in real time is included in the framework. Results show that the AI detection in general has much more security resilience, decreasing possible attack vectors. This makes the integration of various features like AI, blockchain, and IDS for a solid IoT security a must.