Rajeev Gandhi Memorial College of Engineering and Technology

This study examines the magnetohydrodynamic flow and thermal characteristics of a Cu - ZrO_2 / EG - H_2 O hybrid nanofluid across two configurations: a wedge surface and a flat plate. The analysis includes the impacts of thermal radiation, porous medium characteristics, internal heat generation, and the Weissenberg number. The main objective is to assess how these physical parameters alter the velocity profiles, temperature distributions, entropy generation rates, skin-friction coefficients, and Nusselt numbers for both geometric configurations. To verify the precision of the solutions, the nonlinear boundary-layer equations are transformed into similarity forms and resolved utilizing the Runge–Kutta shooting method and the Homotopy Perturbation Method (HPM). Quantitative findings demonstrate that thermal radiation and heat-source parameters raise the temperature by 14–18

As organizations increasingly transition to cloud infrastructures, the risk surface for insider threats has broadened, with privilege escalation emerging as a critical cybersecurity concern. Traditional security mechanisms—such as static rule-based systems and conventional access controls—often fail to detect subtle behavioral anomalies, particularly when insiders misuse legitimate credentials. To address this gap, this research presents a behavior-centric ensemble learning framework for detecting insider threats and preventing privilege escalation in cloud environments. The system incorporates user behavior analytics (UBA) alongside supervised machine learning models—Random Forest, XGBoost, and LightGBM—aggregated through majority voting to enhance detection robustness. Behavioral indicators, including session duration, file access frequency, and protocol usage, are derived from the NSL-KDD dataset and processed using one-hot encoding and z-score normalization to form feature vectors. Integrated within a modular architecture, the model operates in real-time and employs RSA-based cryptographic mechanisms for secure data handling. Experimental results reveal that LightGBM achieves superior performance, with an accuracy of 97.00

Human–robot interaction is an important area in modern robotics, where intuitive and efficient control methods are required for operating robotic systems. Traditional robot control approaches usually depend on keyboards, controllers, or manual programming, which can make interaction less flexible and more complex for users. To overcome these limitations, this project presents a vision-based hand gesture control system for the Dobot Magician robotic arm. The system allows users to control the robot using natural hand gestures captured through a camera, enabling a more interactive and contactless control mechanism. By using computer vision techniques, the system detects and interprets hand gestures in real time, allowing users to perform robot movements without the need for physical control devices The proposed system integrates computer vision and robotic control to enable real-time gesture-based interaction with the Dobot Magician robotic arm. Hand gestures are detected using OpenCV and MediaPipe, which track hand landmarks from the live video stream. The recognized gestures are converted into robot commands through a command mapping module for controlling arm movements. The system uses the ROS 2 communication framework to transmit commands between modules efficiently This approach enhances the intuitiveness and accessibility of robotic systems and can be applied in areas such as robotics education, industrial automation, human–robot interaction, and assistive technologies. The developed system demonstrates the potential of vision-based gesture interfaces as an efficient and user-friendly method for robotic control

With the rapid expansion of Android apps, the threat posed by sophisticated malware has grown enormously because of improved evasion and obfuscation techniques. Conventional detection mechanisms that primarily employ static analysis, Convolutional Neural Networks (CNNs), and Discrete Fourier Transform (DFT) usually fail to detect sophisticated or camouflaged malware. The primary objective of this project is to enhance malware detection strength and accuracy with a multimodal system utilizing spatial, frequency, and semantic features. This is motivated by the increased demand for a more intelligent system able to counter prevalent yet newer malware attacks that traditional models are not good at dealing with. The suggested system introduces a number of enhancements. It substitutes Wavelet Transform for DFT to better extract time frequency features from malware signatures. Grayscale images of APK bytecode are handled by CNNs to extract spatial features whereas semantic information such as permissions and API calls are represented by Transformer-based models. These various features are mixed up using a cross-modal attention mechanism to dynamically combine multiple modalities. To enhance resistance to adversarial malware, the system also incorporates contrastive learning and adversarial training techniques. This system is highly effective in mobile security contexts, including app stores, company mobile protection systems, and cybersecurity research and test platforms. It markedly outperforms conventional CNN based systems in detection accuracy, flexibility, and resistance to concealed malware. By addressing the vulnerabilities of current systems and proposing novel solutions, the project presents a more Intelligent and future-resistant solution to contemporary Android malware detection.

Cardiovascular disorders require therapeutic interventions beyond the limitations imposed by traditional drug delivery. Motivated by the need to understand heat and momentum transport in diseased arterial environments, the present study investigates the flow and thermal characteristics of a Casson-Maxwell ternary nanofluid over a Local Wall-Bounded Stretching Surface (LWBSS) representing a stenotic arterial segment. The analysis incorporates magnetic field effects, thermal radiation, and heat generation/absorption, while the governing equations are solved numerically using the MATLAB bvp4c solver. The study reveals that the Casson-Maxwell model produces a lower velocity field than the Casson fluid model, indicating enhanced flow resistance due to viscoelastic relaxation effects. An increase in the Maxwell parameter significantly suppresses fluid motion, whereas thermal radiation enhance the temperature distribution within the thermal boundary layer. Among the nanoparticles considered, copper and aluminium oxide nanoparticles increase the heat transfer rate, whereas silver nanoparticles reduce it. The skin-friction coefficient decreases by approximately 213.8