Digital multimedia content experiences a variety of degradations during the process of acquisition, processing, and transmission. Automatic approaches are required to evaluate the performance of any of these processes regarding perceptual image quality assessment. Several human observers can judge an image based on its perceptual quality, and the results may then be averaged to obtain an image quality score, making it the most accurate way of assessing image quality. Because this approach requires the engagement of human resources, an automatic method based on machine learning modelling of the human visual system is necessary. Machine learning techniques may be trained in a supervised manner by providing a set of images along with their subjective quality score as ground truth. This subjective score is sometimes expressed as the Mean Opinion Score, which is the average of numerous human judgments (MOS). In this study, we have trained a deep Convolutional Neural Network (CNN) using labelled images. The study focuses on images that are naturally distorted during acquisition, processing, or transmission rather than on synthetically distorted images with discrete distortion levels. Using naturally distorted images will result in a more general-purpose image quality evaluation model can be obtained. Furthermore, collecting a significant number of labelled images for image quality is an expensive task, and no dataset with a sufficient number of images is available, thus several ways for training the CNN will be investigated. The trained model was evaluated by determining the correlation between the ground truth (MOS) and predicted quality scores, resulting in PLCC and SROCC values of 0.7729 and 0.7489, respectively of 0.7729 & 0.7489, respectively. However, when tested on the BIQ2021 dataset, the PLCC and SROCC values were 0.7845 and 0.7588, respectively.
Nisar Ahmed, Muhammad Usman Younus, Kalsoom Safdar· International Journal of Inn...· 0 citations
Phishing and social engineering rank among the most significant facilitators of cyber-crime, which encompasses data breaches, cyber-attacks, ransomware schemes, and denial of service incidents. A robust threat detection model is essential for reducing the risk of phishing attacks, as these types of attacks are frequently discussed in dark web forums and are thus widely adopted. It has been noted that these attacks can have severe direct or indirect effects on our assets, with the resulting financial damage closely linked to their occurrence. In this research, two methodologies are introduced: the first is an explainable AI (XAI) model specifically designed to assess cyber-risks associated with correlated phishing threats. The second is a hybrid approach referred to as a classifier ensemble, which employs a combination of three top-performing machine learning models through ensemble learning and categorizes them based on several high-impact parameters related to model performance. Ultimately, both proposed models have been compared and evaluated across multiple factors. The proposed model in addition carried out a precision of 97.5%, consider of 98.7%, and F1-score of 98.1%, with an average development of about 2–3% in comparison with baseline models. SHAP evaluation recognized Page Rank, URL Length, Number of Hyperlinks, Domain Age, Google Index, and Phishing Hints because the maximum influential capabilities affecting phishing prediction, while LIME supplied instance-stage reasons to enhance transparency and analyst confidence. The experimental findings display that integrating explainable artificial intelligence with ensemble getting to know appreciably improves phishing detection accuracy whilst simultaneously improving version interpretability, making the proposed framework appropriate for deployment in realistic cybersecurity environments.
Kalsoom Safdar, Huraira Irfan· International Journal of Inn...· 0 citations
