Gulshan Saleem

Distributed Denial of Service attacks remain a critical threat to modern networked systems due to their scale, diversity and evolving attack strategies. Although machine learning and deep learning techniques have been widely explored for DDoS detection, many existing studies rely on inconsistent preprocessing pipelines, single-dataset evaluations and limited reproducibility. This work proposes a unified and resource efficient detection framework that addresses these challenges through systematic data handling and transparent model evaluation. The proposed pipeline integrates data cleaning, memory optimization, class balancing and hybrid feature engineering that combines linear, tree-based, statistical and information-theoretic selection methods. Classical machine learning models and a one-dimensional convolutional neural network (CNN) are evaluated on two widely used benchmark datasets, CIC-IDS2018 and CIC-DDoS2019, under a leakage-free experimental protocol. Principal Component Analysis is further examined as an optional dimensionality reduction technique. Experimental results show that Random Forest and the CNN achieve strong and consistent performance across both datasets, with hybrid feature selection improving accuracy while reducing dimensionality. The findings demonstrate that careful preprocessing and feature engineering enable classical models to perform competitively with deep learning approaches while maintaining lower computational cost. The study emphasizes reproducibility, efficiency and practical deployability, providing a robust baseline for future DDoS detection research and real-world intrusion detection systems.

Vision sensors are becoming more important in Intelligent Transportation Systems (ITS) for traffic monitoring, management, and optimization as the number of network cameras continues to rise. However, manual object tracking and matching across multiple non-overlapping cameras pose significant challenges in city-scale urban traffic scenarios. These challenges include handling diverse vehicle attributes, occlusions, illumination variations, shadows, and varying video resolutions. To address these issues, we propose an efficient and cost-effective deep learning-based framework for Multi-Object Multi-Camera Tracking (MO-MCT). The proposed framework utilizes Mask R-CNN for object detection and employs Non-Maximum Suppression (NMS) to select target objects from overlapping detections. Transfer learning is employed for re-identification, enabling the association and generation of vehicle tracklets across multiple cameras. Moreover, we leverage appropriate loss functions and distance measures to handle occlusion, illumination, and shadow challenges. The final solution identification module performs feature extraction using ResNet-152 coupled with Deep SORT based vehicle tracking. The proposed framework is evaluated on the 5th AI City Challenge dataset (Track 3), comprising 46 camera feeds. Among these 46 camera streams, 40 are used for model training and validation, while the remaining six are utilized for model testing. The proposed framework achieves competitive performance with an IDF1 score of 0.8289, and precision and recall scores of 0.9026 and 0.8527 respectively, demonstrating its effectiveness in robust and accurate vehicle tracking.

Surveillance video analytics encounters unprecedented challenges in 5G and IoT environments, including complex intra-class variations, short-term and long-term temporal dynamics, and variable video quality. This study introduces Edge-Enhanced TempoFuseNet, a cutting-edge framework that strategically reduces spatial resolution to allow the processing of low-resolution images. A dual upscaling methodology based on bicubic interpolation and an encoder–bank–decoder configuration is used for anomaly classification. The two-stream architecture combines the power of a pre-trained Convolutional Neural Network (CNN) for spatial feature extraction from RGB imagery in the spatial stream, while the temporal stream focuses on learning short-term temporal characteristics, reducing the computational burden of optical flow. To analyze long-term temporal patterns, the extracted features from both streams are combined and routed through a Gated Recurrent Unit (GRU) layer. The proposed framework (TempoFuseNet) outperforms the encoder–bank–decoder model in terms of performance metrics, achieving a multiclass macro average accuracy of 92.28%, an F1-score of 69.29%, and a false positive rate of 4.41%. This study presents a significant advancement in the field of video anomaly recognition and provides a comprehensive solution to the complex challenges posed by real-world surveillance scenarios in the context of 5G and IoT.

The growing adoption of digital multimedia devices and the greater reliance on compression and wireless channels for data transmission has brought renewed focus to the traditional challenge of evaluating image quality. Image Quality Assessment (IQA) is needed to optimize bit rate, compression, or processing and communication strategies for these multimedia technologies. Visual representation learning enables the model to undertake upstream training on large-scale data and then fine-tune the model on downstream data using fewer training samples. Data annotation for IQA is expensive due to the difficulty of grading a picture’s quality, the need to gather quality labels from numerous observers, and the diversity of perceptual quality and content of the images. This challenge has limited the amount of the labeled training dataset for IQA to a few thousand. In this study, a deep Convolutional Neural Network is trained on a large-scale image dataset produced by simulating 165 distortion scenarios on 150,000 images resulting in 24.75 million distorted images. These images are labeled via an ensemble of full-reference quality assessment models which assign the quality rating to each distorted image by using its reference image. This trained model is fine-tuned on two datasets TID2013 and Kadid-10K datasets containing simulated distortions and two datasets KonIQ-10K and BIQ2021 containing authentic distortions. The fine-tuning performance has resulted in state-of-the-art IQA performance and yielded a Spearman’s correlation coefficient of 0.921, 0.893, 0.884, and 0.793, respectively. Moreover, comparison with the ImageNet pre-trained model revealed that the proposed VRL-IQA model provides higher performance in terms of Pearson and Spearman’s correlations and achieves the validation criteria with fewer epochs than the ImageNet pre-trained model. These findings contribute to the advancement of IQA, offering a promising approach for robust and accurate quality prediction in various applications.

Smart surveillance is a difficult task that is gaining popularity due to its direct link to human safety. Today, many indoor and outdoor surveillance systems are in use at public places and smart cities. Because these systems are expensive to deploy, these are out of reach for the vast majority of the public and private sectors. Due to the lack of a precise definition of an anomaly, automated surveillance is a challenging task, especially when large amounts of data, such as 24/7 CCTV footage, must be processed. When implementing such systems in real-time environments, the high computational resource requirements for automated surveillance becomes a major bottleneck. Another challenge is to recognize anomalies accurately as achieving high accuracy while reducing computational cost is more challenging. To address these challenge, this research is based on the developing a system that is both efficient and cost effective. Although 3D convolutional neural networks have proven to be accurate, they are prohibitively expensive for practical use, particularly in real-time surveillance. In this article, we present two contributions: a resource-efficient framework for anomaly recognition problems and two-class and multi-class anomaly recognition on spatially augmented surveillance videos. This research aims to address the problem of computation overhead while maintaining recognition accuracy. The proposed Temporal based Anomaly Recognizer (TAR) framework combines a partial shift strategy with a 2D convolutional architecture-based model, namely MobileNetV2. Extensive experiments were carried out to evaluate the model’s performance on the UCF Crime dataset, with MobileNetV2 as the baseline architecture; it achieved an accuracy of 88% which is 2.47% increased performance than available state-of-the-art. The proposed framework achieves 52.7% accuracy for multiclass anomaly recognition on the UCF Crime2Local dataset. The proposed model has been tested in real-time camera stream settings and can handle six streams simultaneously without the need for additional resources.

Due to the increasing volume of multimedia content and the need for automated processing, anomaly identification in surveillance videos is crucial. Deep learning has been demonstrated to be an effective solution for numerous visual classification tasks; consequently, it has been adopted for learning anomaly identification. This study pre-processes video frames following temporal segmentation of anomalous video segments. These frames are fed into a two-stream architecture that uses spatiotemporal learning and late temporal fusion to classify anomalous videos. In this paper, surveillance system is proposed which is named as ”Surveilia” and focused at detecting anomalies within surveillance videos. The spatial stream of the proposed model (Surveilia) uses a ResNet-101 model and every eighth frame of the video stream for predictive modelling. In contrast, the temporal stream generates SGF, which consists of eight grayscale video frames. Additionally, this stream also employs a ResNet-101 model that reduces learned filters by a factor of up to eight. After each of ResNet’s four bottleneck blocks, information is passed from the temporal stream to the spatial stream. In lieu of a global average pooling mechanism, the results of both streams are combined with an attention-based mechanism. The loss of information obtained from temporal stream is caused by the loss of substantial temporal information as a result of global average pooling. The model is trained with 80% of the data and validated with the remaining 20%. Moreover, data augmentation is applied to training data in order to increase the effective dataset size and model generalization. The proposed model (Surveilia) outperforms state-of-the-art in multi-class anomaly identification, with a validation accuracy of 56.4% on the