Securing Smart Cities: AI-Driven Video Injection Attack Detection for Enhanced Urban Surveillance

With the integration of video surveillance sensors into smart city applications, video content security becomes more important. This research uses a combination of convolutional neural networks (CNNs) and long short-term memory (LSTM) networks to provide a detailed framework to detect video injection attacks. The method is evaluated using the Kitsune Network Attack Dataset, which includes a variety of network traffic scenarios for an in-depth examination. The proposed model achieves excellent outcomes with an accuracy of $98.0 \%$, precision of $99.8 \%$, recall of $99.7 \%$, and F1-score of $99.6 \%$. Moreover, extensive examination using a confusion matrix and metrics such as False Discovery Rate (FDR), False Negative Rate (FNR), False Omission Rate (FOR), and False Positive Rate (FPR) illustrate the model’s reliability in identifying between genuine and injected video and audio frames. The successful implementation of the proposed video injection attack detection technique has been demonstrated in this paper, which is facilitating the improvement of smart city security. The results of this research paper provide significant perspectives for implementing forefront safety procedures and preserving the reliability and integrity of video frames in the context of smart cities.