ABSTRACT: Digital watermarking has been extensively employed to ensure copyrighting of images and authenticity, but in some cases, it becomes necessary to remove watermarks such as in restoration of old archives, in forensic investigations, or in the reuse of images in a research database. Conventional watermark removal methods tend to destroy the clarity and structure of the image resulting in conspicuous distortions or blurring. In order to rectify these shortcomings, this paper proposes an efficient watermark removal structure grounded on IMPRINTS (Image Processing and Restoration Techniques). The suggested solution uses superior inpainting algorithms to recreate the hidden image areas following watermark erase and retains the texture, color consistency, and edge continuity. The scheme involves steps of preprocessing, detection of watermarks, mask generation and repair to produce minimum visual artifacts. Experimental tests performed with various watermarked image datasets demonstrate that IMPRINTS outperforms the traditional interpolation and CNN-based watermarking in Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM). The findings depict that the IMPRINTS process is effective in removing watermarks of high quality in visual images and structural integrity, thus, it is a dependable process in image restoration procedures in the real world.
Keywords: Image Restoration, IMPRINTS, Watermark Removal, Image Inpainting, Edge Detection, PSNR, SSIM, Digital Image processing.

ABSTRACT: Cognitive ability plays a key role in a student’s growth, influencing how well they can think, remember, calculate, and understand information. This framework introduces a smart semantic chatbot designed to evaluate students' cognitive abilities in a more efficient and personalized way. It automatically gathers student responses through Google Forms, where the questions are thoughtfully grouped into four key areas: Logical Reasoning, Memory & Attention, Mathematical Skills, and Verbal Reasoning. Responses are stored in Google Sheets and fetched dynamically using an API for processing. The scoring system gives points for correct answers (10), medium answers (5), and incorrect ones (0), allowing for a detailed assessment in each domain. After assessing scores, the chatbot provides personalized feedback, identifying strengths and domains to improve, all without repeating itself. Unlike traditional tests, this system shows results in a clear table, helping students easily understand how they did. This proposed framework fills the gap in education by providing a simple, fast, and meaningful way to assess cognitive abilities, with an easy-to-use chatbot. A built-in recommender system further enhances the framework by analyzing students’ scores across the four domains and generating targeted suggestions. Based on performance patterns, it identifies areas where students are lacking, whether in logical reasoning, memory and attention, mathematical skills, or verbal reasoning, and provides specific recommendations for improvement in those weaker domains using association rule mining.
Keywords: Cognitive Ability Assessment, Semantic Chatbot, Student Performance Analysis, Feedback, Recommender System, Association Rule Mining.

ABSTRACT: This paper presents an innovative technology called the "smart shoe," designed to assist people with visual or hearing impairments in navigating their surroundings more easily. Equipped with advanced sensors and mini-computers, the shoe can detect uneven surfaces or obstacles in the wearer’s path. It provides feedback through vibrations or sound alerts to help guide them. Developed with ease of use and practicality in mind, this device aims to empower users, fostering greater confidence and independence. The goal of this technology is to create a more inclusive and secure environment for those with sensory challenges.
Keywords: Assistive Technology, Smart Wearables, Obstacle Detection, Accessibility & Inclusive Mobility.

ABSTRACT: The rapid growth of electric vehicles (EVs) necessitates the development of robust charging infrastructure. This paper focuses on reviewing advanced charging infrastructure, including converter topologies and quick-charging technologies to meet future demands. The burgeoning popularity of electric vehicles (EVs) necessitates a robust and efficient fast charging infrastructure. This review delves into the technological advancements and practical hurdles associated with high-power charging systems. We examine the latest charging technologies, their impact on EV battery life, and the critical factors influencing widespread deployment. By analyzing the interplay of technology, economics, and policy, this paper offers insights into overcoming stability challenges and optimizing fast charging networks to support the transition to sustainable transportation.
Keywords: Electric vehicle charging station, EV Charging Level, Charging facility, Real-world implementation.

ABSTRACT: In windows environment PowerShell becomes an integral part who’s responsible for system automation, management and configuration tasks. It is very flexible and helps in performing all the operations, which makes it more popular among developers and users. Yet, this flexibility is also the reason for being targeted for criminal activities by implanting PowerShell scripts in the form of File less malware as it doesn’t leave any trace in memory. So detecting these scripts before execution is the best way to counter the filmless attacks. For that, an authentic and reliable dataset with advanced Machine Learning (ML) and Deep Learning (DL) models was required with the most useful feature sets for better detection. This paper implemented a generalized workflow to create a reliable dataset by collecting PowerShell scripts from different repositories and cleaning the obfuscated script by implementing a deobfuscation technique. Then evaluated them with a well-known antivirus (AV) platform, Virus Total, to label those scripts. The extracted meaningful features from those clean and labelled scripts are passed as inputs to ML models for evaluation of those models. Among defined models, Trigonometric Functional Link Artificial Neural Network (TFLANN) performed outstandingly with an accuracy of 97.68%. The proposed reliable dataset offers effective support to future research in PowerShell-based file less malware detection and can be used to create more effective machine learning-based cybersecurity mechanisms.
Keywords: PowerShell, Fileless Malware, Deobfuscation, RBFNN, Machine Learning.

ABSTRACT: This study proposes a new configuration called Hetero-stacked Dual Metal Gate Tunnel Field-Effect Transistor (HS-DMG-TFET). The suggested device showcases enhanced ON-state current (ION), a heightened ION/IOFF current ratio, and decreased sub-threshold swing, representing a substantial advancement in both DC and switching performance compared to existing TFET topologies. The proposed design includes a dual-metal gate featuring Molybdenum and Aluminium, which overlaps with the Si-Ge pocket. The resulting modifications in electron flow and tunneling behavior within the device lead to improvements in various performance metrics. Significantly, an ION/IOFF ratio of ~1014 is attained, with an ON-state current of 1.41 × 10-3 A/μm and an OFF-state current of 1.17 × 10-17 A/μm. The device also exhibits a significantly reduced sub-threshold slope, measuring 9.52 mV/dec, indicating superior transient performance. These DC characteristics underscore the potential of utilizing composite material gates for low-power applications, showcasing their advantages over traditional single-metal counterparts. Besides, enhanced transconductance and other RF performance metrics have been obtained for suggested device which confirms the device's suitability for a variety of analog and RF applications.
Keywords: Tunnel-Field-Effect-Transistor, Subthreshold Slope, Composite-Gate, Molybdenum, Low Power, High-Frequency Applications.

ABSTRACT: Liver Disease (LD) is one of the most dangerous risks to human health worldwide. Early detection is essential for timely intervention and improved patient survival. The traditional diagnosis relies on manual interpretation by medical experts, a process that can be time-consuming and prone to human error. Machine Learning (ML) offers a promising alternative, delivering faster and more consistent analysis than manual methods. However, relying on a single ML algorithm is limited by its architectural constraints, and the black-box nature of Artificial Intelligence (AI) restricts its real-life adoption. To address these limitations, this work introduces a strategic, explainable AI-based ensemble approach for efficient detection of LD disorders. A two-level learning approach is employed. The original binary dataset is first converted into a multiclass severity structure using recognized medical biomarkers, giving a more clinically useful label than a simple yes/no output. SMOTE is applied to balance the classes before scaling the features. Twelve classifiers are then evaluated individually, and the three strongest models: Random Forest, XGBoost, and CatBoost are combined into a stacking ensemble through an interpretable logistic regression meta-learner. SHAP is applied throughout to explain feature contributions and confirm the meta-learner’s reasoning remains transparent. In a second stage, only Moderate and Severe LD cases are considered, excluding healthy individuals, to give clinicians a sharper basis for treatment decisions. The proposed ensemble achieved an F1-Macro score of 0.8600 (± 0.0158, 95% CI: 0.8461–0.8739), with a Wilcoxon signed-rank test confirming a statistically significant improvement over the best individual classifier (p = 0.0312). This presents a reliable nature of the approach. Its lightweight and interpretable nature enables ease of deployment and adoption in other domains.
Keywords: Liver Disease, Machine Learning, ArtificialIntelligence, Ensemble Learning, LR, SMOTE, SHAP.