Abstract:
The Internet of Things (IoT) technology development speed, along with its integration into
smart city infrastructure, requires strong cybersecurity solutions to protect important systems. The research
introduces an extensive framework to protect IoT systems and wireless networks through the integration
of penetration testing method with anomaly detection and predictive modeling techniques. In this study,
we collected real-time network traffic data as part of our methodology before performing penetration tests
with Airmon-ng and Wireshark to create enriched attack scenario datasets. Anomalies were identified using
an optimized Isolation Forest model, revealing patterns such as unusual activity involving the Tenda_476300
WiFi network. The Tenda_476300 network exhibited frequent authentication packet anomalies, along
with other potential misconfiguration or vulnerability indicators. Predictive modeling utilized both logistic
regression (LR) and support vector machines (SVM) for binary classification to identify benign from
malicious traffic, resulting in high accuracy rates and precise results. XGBoost achieved better performance
than Random Forest (RF) across all metrics when performing as a multiclass classifier to identify Denial of
Service (DoS), Distributed Denial of Service (DDoS), and brute force attacks. The reliability and robustness
of the constructed models were tested using precision, recall, F1 scores, ROC curves, and precision-recall
curves during performance evaluation. The anomaly detection and predictive modeling analysis proves that
real-time surveillance systems should incorporate these techniques for proactive security threat discovery
and defense. The proposed framework delivers a flexible solution for protecting IoT and smart city wireless
networks, which helps create safer, resilient urban environments.
