DEVELOPMENT OF AN IOT-BASED DEEP LEARNING SYSTEM FOR EARLY DETECTION OF CYBERSECURITY ATTACKS IN SMART CITY NETWORKS
Keywords:
Internet of Things, Smart City, Cybersecurity, Deep Learning, CNN-LSTMAbstract
The growing use of Internet of Things (IoT) devices in smart cities has created security challenges in cyber-physical systems because of the complicated and diverse IoT environment. The current research trains an IoT-based deep learning model to identify cybersecurity attacks in smart city systems at an early stage. Hybrid convolutional neural network (CNN) and long-short term memory (LSTM) networks with CICIoT2023 data detection were used. This research selected and prepared a balanced data subset of 15200 network observations, including benign traffic and seven attack types, for analysis. The data were cleaned, categorical variables were encoded, and data were normalized. The data subset was split into 70% training, 15% validation, and 15% test sets. The CNN-LSTM model was trained and tested using Adam optimization with a learning rate of 0.001, batch size of 64, and early stopping. Overall multiclass accuracy, precision, recall, and F1 score were 54.61% (Attack detection: 85.56%), 61.56%, 54.61%, and 54.67%, respectively. Binary detection yielded 85.56% attack detection and 49.47% false-positive rates. The study revealed that the CNN-LSTM model outperformed traditional machine learning models and was effective against distributed denial of service (DDoS), denial of service (DoS), Mirai, and reconnaissance attacks but was less efficient against brute force, spoofing, and web attacks. The results showed that the CNN-LSTM model can detect malicious traffic for smart city IoT systems at an early stage with reasonable accuracy. However, further improvements are required to enhance multiclass attack detection before deploying the model in production.
