As smart cities evolve, they integrate various applications such as intelligent transportation systems, energy management, healthcare, and public safety, all of which depend on interconnected networks. These applications rely on massive data exchanges between sensors, devices, and cloud services, making the system more efficient but also exposing it to cybersecurity challenges. Cyber threats, including data breaches, denial of service (DoS) attacks, and malware, can disrupt essential services, compromise privacy, and endanger lives. The complexity of smart city infrastructure amplifies vulnerabilities, making real-time detection and localization of attacks a critical necessity. In this paper, we propose a novel framework for attack detection and localization specifically designed for smart city environments. The framework integrates machine learning-based intrusion detection systems (IDS) with packet analysis techniques. Upon detection of an anomaly, detailed packet analysis is performed to extract crucial information, such as IP addresses, GPS coordinates, and other metadata. This enables precise localization of the attack's source, facilitating rapid response and mitigation. The combination of machine learning for anomaly detection with packet-level analysis ensures a comprehensive approach, significantly improving detection accuracy and localization precision. Extensive evaluations on real-world datasets demonstrate the efficacy of the proposed method in enhancing the security of smart city networks, while reducing false positives and improving real-time response capabilities. This framework represents a critical advancement in protecting smart cities from evolving cyber threats.

In order to improve the quality of human existence, comfort and efficiency are key objectives in smart environments. It is now possible to construct smart cities due to the latest advancements in Internet of Things (IoT) technology. Privacy and security are major concerns in IoT-based smart objects. Smart environments are at risk for safety from IoT-based technologies. Intrusion detection systems (IDSs) created for IoT environments are essential for preventing IoT-related security threats. Many cyber security systems use IDSs to find intrusions. Anomaly-based IDS learns the typical pattern of system activity and alerts on anomalous events as they happen as opposed to analyzing monitored events against a database of known intrusion events, as is the case with signature-based IDS. The installation of IDS on the IoT network is the main topic of this paper. Key design approach presented in this paper must be taken into consideration when developing an intrusion detection system for the Internet of Things. In this study, we use the Convolutional Neural Network (CNN) to identify attacks on nine commercial IoT devices. Using an actual N-BaIoT dataset that was taken from a real system and included both benign and harmful patterns, extensive empirical research was conducted. The testing results demonstrated a good accuracy of the CNN model in identifying botnet assaults from security cameras with accuracies of 90.25% and 91.76%. Overall, the CNN model was effective in accurately identifying botnet attacks from a variety of IoT devices.

Smart cities are made up of various components that are interconnected. These components exchange data on an ongoing basis and they facilitate the lives of citizens. Its use of Information and Communication Technology (ICT) was a key factor in its sustainable development. Nonetheless, this development contributed to a rise of safety threats, criminal use of information and several other security and privacy challenges. As a result, security and privacy concerns have emerged as a significant problem for smart cities. Safety factors for smart cities have become a concern for all those involved in this field. In this study, we deeply examine and review and the concept of smart cities and the challenges it faces at first. In a second phase, we mainly address the research gap of security in smart cities and present an analysis of associated security challenges. In the last section, we introduce our approach that aims to: (i) capture the processes of penetration attempts, alterations and cyber attacks; (ii) truck malicious behaviors and locate their sources; and (iii) finally setup controls to repel and prevent them. To illustrate the applicability and efficiency of our solution, We refer to a case of study to demonstrate the efficacy of our detection method, using different machine learning algorithms and the dataset CICIDS2017.