VOLUME 3, ISSUE 5, MAY 2026

Abstract: The fusion of Generative AI and DevOps is reshaping the landscape of cloud computing by introducing dynamic intelligence into automated software delivery pipelines. Traditionally, DevOps has relied on deterministic scripts and manual configurations to manage infrastructure, CI/CD workflows, and system operations. However, as applications scale across hybrid and multi-cloud environments, these static approaches face limitations in flexibility, responsiveness, and resilience. Generative AI addresses these challenges by leveraging large language models (LLMs) and agentic architectures to understand context, generate code, interpret telemetry, and take proactive actions.Our research demonstrates that generative AI is not just a tool for automation but a catalyst for building self- optimizing, context-aware, and resilient DevOps systems. As organizations adopt these technologies, they will transition from reactive incident handling to predictive and autonomous operations, setting the stage for the next era of intelligent cloud engineering.

Keywords: Generative AI, DevOps Automation, Cloud Workflow Optimization, Large Language Models (LLMs)

Abstract: Human Activity Recognition (HAR) plays a vital role in modern healthcare systems by enabling continuous monitoring of individuals' physical activities and behaviors. Accurate recognition of daily activities such as walking, sitting, running, and sleeping can support early diagnosis, rehabilitation, elderly care, and fitness tracking. Traditional activity recognition systems often rely on a single data modality, such as wearable sensors or video data, which limits their accuracy and robustness. To overcome these limitations, this project proposes an intelligent multimodal human activity recognition system that integrates multiple data sources to improve performance and reliability. The proposed system utilizes multimodal data inputs such as wearable sensor data (accelerometer, gyroscope), visual data from cameras, and possibly audio signals. Data preprocessing techniques including normalization, noise removal, and feature extraction are applied to prepare the dataset. Deep learning models, particularly Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), are employed to capture spatial and temporal patterns in the data. Feature fusion techniques are used to combine information from different modalities, resulting in a more comprehensive representation of human activities. Experimental results demonstrate that the multimodal approach significantly improves recognition accuracy compared to single-modality systems. The system is capable of accurately identifying a wide range of human activities in real time, making it suitable for healthcare applications such as patient monitoring, fall detection, and fitness tracking. However, challenges such as data synchronization and computational complexity remain. Overall, the proposed system provides an efficient and scalable solution for intelligent activity recognition in personal healthcare environments.

Keywords: Human Activity Recognition (HAR), Multimodal Learning, Deep Learning, CNN, RNN, Feature Fusion, Wearable Sensors, Healthcare Monitoring, Activity Detection, Smart Healthcare.

Abstract: In today's modern world, "fake news" has been a major concern, spreading like wildfire through many platforms. This phenomenon not only undermines the credibility of information but also misleads society. Nowadays, social media is the greatest means by which fake news spreads all over the place. This can cause many problems such as defamation of people and spreading news in favor of specific individuals. Fake news often targets the most prominent, powerful, and influential people in society, aiming to tarnish their reputation. The escalating impact of fake news knows no bounds. Fake news is often biased, favoring a single person or a section of people in society for their personal benefits. To mitigate these challenges and promote transparency, there is a need to reduce the spread of fake news. Introducing a "Fake News Classifier using NLP" offers a promising solution to combat this issue. By using machine learning algorithms, this classifier can effectively identify misleading information as fake news, thereby contributing to awareness in society and reducing losses.

Keywords: Natural Processing Language, TF-IDF, Flask, Classification, Multi nomialNB, Accuracy.

Abstract: The system is designed to work with live video feeds from cameras installed in strategic locations. It employs object detection algorithms to identify and track vehicles in real-time, allowing for accurate traffic analysis. The system incorporates speed violation detection by defining speed limit lines and calculating the speed of vehicles passing through those lines. Violation instances are flagged, and images or videos of the violations are captured for further analysis or evidence purposes. The project also includes a user-friendly interface that provides real-time traffic statistics, including the total number of vehicles, traffic congestion levels, and detected violations. Additionally, the system offers configurable settings for road-specific parameters, such as speed limits and the number of allowed vehicles. The proposed system aims to enhance traffic management and improve road safety by providing timely and accurate information to authorities. It can aid in monitoring traffic patterns, identifying congested areas, and enforcing speed limits. The system has the potential to reduce accidents, enhance traffic flow, and contribute to efficient transportation management. Overall, the project showcases the effective utilization of computer vision and deep learning algorithms to develop a comprehensive traffic monitoring and violation detection system that can significantly impact road safety and traffic management.

Keywords: Traffic Management, Traffic Flow, Deep Learning Algorithms, Traffic Monitoring, Computer Vision.

Abstract: Machine learning techniques are being widely used to identify and respond to unusual events in industrial controls systems (ICS), where they play a vital role in preventing potential catastrophes. This paper reviews the various techniques that are used in anomaly detection in these systems. The paper discusses the definition of an anomaly detection process and provides a comprehensive review of the various techniques involved in this area. It also explores the applications of machine learning and statistical techniques in this domain. Some of the techniques that are commonly used in this area include clustering, decision trees and random forests, and control charts. The paper also covers the applications and challenges of anomaly detection in different industrial control systems such as water treatment plants, power grid systems, and chemical plants. Case studies are presented to demonstrate the effectiveness of learning-based techniques in identifying anomalies in these facilities. The paper also presents an evaluation of the performance of various machine learning techniques in performing anomaly detection. The evaluation metrics that are used in these experiments include false positive rate, accuracy, recall, area under receiver characteristic curve, and F1 score. The paper concludes by providing a summary of the findings of the review and the future directions of the investigation in anomaly detection for industrial control systems. The paper offers valuable insights into the latest state-of-art techniques in this area, and it can help practitioners and researchers make informed decisions when it comes to choosing the appropriate ones for their specific projects.

Keywords: ICS, Cyber-attack, ML, Supervised learning, Un-supervised learning

Abstract: The rapid advancement of deep learning-based manipulation techniques has led to highly realistic object-based video forgeries, posing significant threats to digital security, media authenticity, and public trust. Existing studies reveal that modern DeepFake and tampering methods leave subtle visual, temporal, and semantic inconsistencies that can be effectively analyzed using deep neural networks [1], [2], [3]. Recent works in video forensics demonstrate the importance of detecting warping artifacts, head-pose inconsistencies, and spatio-temporal distortions to expose manipulated regions [1] Convolutional Neural Networks (CNNs), two-stream architectures, and recurrent networks have proven highly effective for identifying object removal, splicing, and GAN-generated forgeries by learning deep visual representations and motion cues [6]. Large datasets such as the DeepFake Detection Challenge (DFDC) further support robust model training and generalization across diverse manipulation styles. Object-level manipulation detection is strengthened by combining spatial CNN features with temporal modeling, enabling the system to capture fine-grained tampering artifacts across consecutive frames. Despite adversarial attacks and evolving forgery methods that challenge CNN robustness deep convolutional strategies-enhanced with recurrent layers and multi-stream analysis-remain the most reliable solution for advanced video forgery detection. This research builds upon these findings to develop a robust, deep convolutional neural network tailored for detecting complex object-based video forgeries, ensuring high accuracy, temporal stability, and resilience against modern manipulation techniques.

Keywords: Deep Convolutional Neural Networks, Video Forgery Detection, Object-Based Manipulation, DeepFake Detection, Spatio-Temporal Analysis, Video Forensics, CNN Features, Temporal Inconsistency, GAN-Based Forgeries, Tampering Localization, Motion Cues, Multimedia Security, Digital Forensics, Adversarial Robustness, Forgery Classification.