Volume 13 | Issue 7 |

 Abstract

Cyber deception threats are becoming increasingly sophisticated, often evading traditional security measures such as firewalls and intrusion detection systems. Attackers can exploit unpatched systems or use advanced techniques to infiltrate networks. This paper introduces an AI-powered IP-based cyber deception system designed to confuse and deceive attackers using intelligent honeypots and anomaly detection. Our approach enhances threat intelligence and adapts dynamically to evolving threats.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Component, formatting, style, styling, insert.

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

The rapid advancement of deepfake generation techniques has created significant challenges in preserving the authenticity of digital media. This comprehensive survey examines the state-of-the- art in deepfake video detection, with a particular focus on hybrid Long Short-Term Memory (LSTM) models that combine spatial and temporal analysis capabilities. We analyze over 50 recent studies (2019-2024) to evaluate the effectiveness of various architectural approaches, including Convolutional Neural Network- Long Short-Term Memory (CNN-LSTM), Three Dimensional Convolutional Neural Network- Long Short-Term Memory (3DCNN-LSTM), and attention-enhanced variants. The paper provides a detailed comparison of model performance across benchmark datasets such as FaceForensics++ and Celeb-DF, while discussing key evaluation metrics like AUC-ROC and F1-score that are critical for assessing detection reliability. We systematically identify current limitations in generalization capability, computational efficiency, and adversarial robustness that hinder real-world deployment. The survey concludes by outlining promising research directions, including multimodal fusion techniques, lightweight model architectures for edge deployment, and explainable AI approaches to enhance forensic credibility.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Hybrid Long Short-Term Memory (LSTM), Convolutional Neural Network- Long Short-Term Memory (CNN-LSTM), Three Dimensional Convolutional Neural Network- Long Short-Term Memory (3DCNN-LSTM), multimodal fusion techniques.

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

Deepfake technology, driven by artificial intelligence, has developed rapidly over the past few years, raising issues of misinformation, privacy violations, and online security threats. This project is centered around creating a robust Deepfake Detection System based on machine learning methods to distinguish real media from the manipulated one. The system has a user authentication module for secure access via a login system. In addition, it incorporates an advanced deepfake detection algorithm that can scan images and videos to verify whether they are authentic. The detection model generates a fake accuracy percentage, reflecting how much media are likely manipulated. This measure adds transparency and gives users quantifiable feedback into possible deepfake risks. The system utilizes convolutional neural networks (CNNs) and deep learning to make high-precision identification of synthetic content. The technology can be applied to real-world scenarios such as media authentication, law enforcement, and social media surveillance, helping in the mitigation against misinformation. To make it scalable and efficient, the platform will be developed with an easy-to-use interface where individuals and organizations can upload and examine media easily.Through the creation of a correct and accessible detection system, we are moving closer to maintaining trust in digital content and preventing the risks involved in synthetic media manipulation.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Deepfake Detection, Convolutional Neural Networks (CNNs), deep learning techniques, AI-driven cybersecurity

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

This paper presents the design and implementation of a comprehensive web-based College ERP system aimed at enhancing the efficiency of academic and administrative operations in educational institutions. The system automates critical functions such as student enrolment, faculty management, attendance tracking, and examination processing, replacing traditional manual methods that often lead to inefficiencies and errors. Developed using modern web technologies, the solution ensures scalability, robust data security, and user-friendly access across various roles within the institution. It incorporates features such as role-based access control, a modular architecture, and real-time analytics to support data-driven decision-making and institutional transparency. By streamlining operations, reducing administrative workload, and improving communication among stakeholders, the system fosters a more organized and technology-driven educational environment. Furthermore, it is designed with future extensibility in mind, supporting cloud deployment and integration with advanced tools such as AI analytics and Learning Management Systems (LMS). This ERP system not only provides a practical approach to managing college operations efficiently but also serves as a foundational step toward ongoing innovation in educational technology.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

College ERP, Student Information System (SIS), Role-Based Access Control (RBAC), Attendance Management, Examination Management, Web-Based ERP System, Database Management, Cloud-Based Deployment & Data Security.

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

In the rapidly evolving landscape of e-commerce and retail, the integration of predictive modelling, real-time data analytics, and artificial intelligence (AI) has significantly transformed pricing strategies, customer engagement, and operational efficiencies. This research investigates the implementation of predictive analytics techniques for new product pricing, the role of real-time data processing in enhancing business agility, and the transformative impact of AI in delivering personalized consumer experiences. Predictive modelling techniques leverage historical data, market trends, and consumer behavior to optimize pricing decisions, while real-time analytics architectures utilizing technologies like Apache Kafka and Apache Flink facilitate immediate insights into inventory management, customer preferences, and dynamic pricing.This paper presents a comprehensive analysis of how these technologies collectively empower businesses to achieve operational excellence, enhance customer satisfaction, and sustain competitiveness in the digital marketplace. Ethical considerations regarding data privacy and algorithmic fairness are also highlighted, ensuring responsible deployment of AI- drivennsolutions. The study ultimately emphasizes the critical role of data-driven, real-time, and AI- augmented approaches in shaping the future of e-commerce and retail industries.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Predictive Modelling, Real-Time Data Analytics, Artificial Intelligence, Dynamic Pricing, Customer Personalization, E-Commerce.

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

The calculation of the Air Quality Index (AQI) in India differs greatly from global norms due to regional characteristics such as geographical diversity, seasonal fluctuations, and pollution sources. While most countries use consistent techniques that emphasize pollutants such as PM 2.5, NO2, and O3, India's methodology favors PM 10 and PM 2.5 due to high dust levels, industrial emissions, and biomass combustion. The AQI calculation for India includes adaptive seasonal modifiers to account for crop burning, festivities like Diwali, and climatic conditions such as monsoons and winter inversion. Additionally, regional weightage variables are added depending on local pollution sources, which improves accuracy. Unlike worldwide models, which rely mainly on static pollution criteria, India's model makes dynamic modifications to account for real-time environmental and demographic conditions. This approach provides a more relevant and accurate representation of air quality, catering to India's unique climatic, industrial, and cultural conditions. In addition, we present a detailed investigation of chemical processes and how their various quantities influence the toxicity of the compounds produced. We investigate the significance of five key gases. We assess the adverse effects of the produced items utilizing data from internet sources and a variety of calculation and visualization methodologies. The evaluation is based on established threshold values for all gases involved.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Air pollution, Pollution Control Board, Pollutant data analysis, Predictive modelling, Random Forest ML algorithm, User Friendly website, Data visualization.

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

The AI-Powered Afforestation Planner project aims to address the growing issue of air pollution through strategic afforestation. By leveraging advanced remote sensing and machine learning techniques, the project identifies barren land areas suitable for tree planting to improve air quality. The study focuses on the Kanakapura Taluk in Ramanagara District, where land classification is performed using Google Earth Engine (GEE) with manually provided training samples. These samples were used to classify the region into urban areas, water bodies, vegetation, and barren lands using the Random Forest algorithm. The project fetches real-time Air Quality Index (AQI) data to assess pollution levels and recommends the optimal number and species of trees for planting. The final output is a web application that provides users with land classification results, barren land area calculations, and tree species recommendations tailored to improving air quality based on AQI levels. The web-based approach ensures accessibility for end users, offering an interactive tool for better environmental decision-making.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Afforestation, Land Classification, Google Earth Engine (GEE), Random Forest, Air Quality Index (AQI)

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

As virtual statistics turns into an increasing number of popular ensuring strong safety for sensitive files is critical this studies introduces a complicated protection framework that mixes biometric watermarking with device mastering to establish a tamper-resistant and adaptive protection system by way of encoding intricate iris and fingerprint patterns the usage of a custom designed rubiks cube encryption algorithm the method creates a comfy embedded watermark that is tremendously proof against manipulation in parallel convolutional neural networks CNNs examine and authenticate biometric statistics permitting real-time detection of spoofing tries and unauthorized changes the adaptive gaining knowledge of functionality of CNNs lets in the system to refine its detection accuracy through the years strengthening its resilience against rising threats this precise integration of encryption and shrewd pattern recognition gives extensive improvements in file security with ability packages in sectors which include healthcare finance and authorities wherein records integrity and authentication are paramount.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Biometric Watermarking, Document Security, Rubik Encryption, Convolutional Neural Networks (CNN), Machine Learning, Iris and Fingerprint Fusion, Zero-bit Watermarking, Authentication, Spoofing Detection, Fraud Detection.

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

The AI Based Crop Recommendation System in Karnataka is an online platform designed to assist farmers in making optimal decisions regarding crop cultivation. It incorporates features such as analysis of soil health, weather forecasting, and market price predictions. Farmers can also access information on suitable crop varieties, irrigation management, and pest control methods through the system. By using this AI-driven system, farmers can improve crop yields, reduce input costs, and enhance overall agricultural productivity. This digital solution streamlines agricultural decision-making and supports sustainable farming practices in Karnataka by providing farmers with essential information and recommendations.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

artificial intelligence, crop recommendation, agriculture, Karnataka, precision farming, sustainable agriculture.

  License

Creative Commons Attribution 4.0 and The Open Definition

 Abstract

The threat of ransomware is considerable in cybersecurity risk and often goes undetected by traditional signature-based detection approaches. In this paper, we present a deep learning-based behavioral analysis framework supporting pro-active detection and disruption of ransomware. Rather than depending on signatures, the framework analyzes system-level activities, such as file encryption, abnormal access, and process relations. The framework utilizes Long Short- Term Memory (LSTM) networks to analyze temporal activities and Recurrent Neural Networks (RNNs) to extract features, enabling real-time identification of ransomware. Our system detects anomalies present in suspicious behavioral patterns, it provides warnings to the administrators, and automatically either quarantines files or isolates from the network. By using deep learning, our framework detects better and has fewer false positives compared to traditional methods. This study demonstrates the potential for deep learning for analyzing behavior for ransomware protection purposes, giving us a strong and adaptive means of defending against evolving cybersecurity threats.

Licence: creative commons attribution 4.0

  License

Creative Commons Attribution 4.0 and The Open Definition

 Keywords

Ransomware Detection, Deep Learning, Behavioral Analysis, Long Short-Term Memory (LSTM), Recurrent Neural Networks (RNNs).

  License

Creative Commons Attribution 4.0 and The Open Definition