Daily Archives: June 14, 2025

Authors: Yash Malsuare, Aryan Purohit, Isha syed, Dr. Maheshwari Birada

Abstract: This paper presents SuperAgent, a novel multi-agent AI framework designed to autonomously handle complex, real-world tasks through intelligent collaboration among dynamic language agents. As the capabilities of large language models (LLMs) continue to advance, there remains a gap in practical deployment frameworks that can translate user intentions into real-world actions with minimal supervision, explainable reasoning, and reliable execution. SuperAgent+ bridges this gap by combining prompt-driven agent generation, transparent multi-step task planning, and API-integrated tool use in a modular architecture that supports human oversight and customization. At the core of SuperAgent+ lies a flexible orchestration engine that dynamically instantiates and manages specialized agents for subtasks such as information retrieval, summarization, decision-making, scheduling, verification, and real-world communication. Users can design and visualize workflows using a drag-and-drop interface, enabling domain experts and non-technical users alike to create autonomous workflows without writing code. The system further integrates a memory layer for context retention, a reasoning logger for traceability, and real-world tool access (e.g., calendars, calls, databases) for execution beyond the digital domain. We evaluate SuperAgent across a variety of tasks such as academic research assistance, enterprise automation, personal productivity planning, and multi-modal content generation. Our results demonstrate improvements in task completion rates, reasoning transparency, and adaptability compared to baseline single-agent and static pipeline systems. This research lays the foundation for future work on fully autonomous AI ecosystems capable of safe, reliable, and cooperative task execution across domains. Furthermore, this research integrates a modular plug-and-play architecture, enabling extensibility for future agents, tools, or models (e.g., vision, audio, or robotic modules). Experimental evaluations indicate substantial gains in task efficiency, traceability, scalability, and user satisfaction, especially in domains such as software development, research summarization, data analysis, and automated reporting. (Stein, Helge Sören and J. Gregoire).

DOI: http://doi.org/

 

 

Published by: vikaspatanker

Authors: Nikhil Chavan, Dr. Rachna Chavan,, Dr. A. A. Khan, Dr. R. S. Deshpande

Abstract: – Water scarcity is a growing global challenge, making it essential to monitor and optimize water consumption effec- tively. This paper presents an AI-based Smart Water Consump- tion Monitoring System that leverages machine learning and IoT technologies to enhance water management. The system utilizes real-time sensor data, applies predictive algorithms, and generates insights to optimize water usage, reduce waste, and detect anomalies. The proposed system aims to encourage sustainable water consumption practices and prevent potential water-related crises.

Published by: vikaspatanker

Authors: Isha Vyas, Professor Amit Thakur

Abstract: Advances in wireless sensor network (WSN) technology has provided the availability of small and low-cost sensor nodes with capability of sensing various types of physical and environmental conditions, data processing, and wireless communication. Variety of sensing capabilities results in profusion of application areas. However, the characteristics of wireless sensor networks require more effective methods for data forwarding and processing. In WSN, the sensor nodes have a limited transmission range, and their processing and storage capabilities as well as their energy resources are also limited. Routing protocols for wireless sensor networks are responsible for maintaining the routes in the network and have to ensure reliable multi-hop communication under these conditions. In this Review work, we give a survey of routing protocols for Wireless Sensor Network and compare their strengths and limitations.

 

 

Published by: vikaspatanker

Authors: Deepti Tripathi, Professor Amit Thakur

Abstract: The Wireless sensor networks (WSN) are becoming popular as an emergent requirement for manhood. Although, these networks are developing vary rapidly but, they can be used in approximately all aspects of the life. A thorough analysis of existing protocols was conducted to understand problems of WSN and few evaluation tables have been provided for the review summary of the performance of the protocols according to parameters such that latency, scalability, transmission type, network traffic and energy perception. Components of the WSN have been discussed in detail. Issues and challenges of WSN were discussed. Different energy harvesting resources and technologies have been analyzed.

 

 

Published by: vikaspatanker

Authors: Hari Om, Steven David

 

Abstract: Social media and review platforms have become key spaces for individuals to voice opinions, shaping trends across sectors like entertainment and business. This study introduces a comprehensive sentiment analysis framework that combines lexicon-based methods, machine learning models, and privacy-conscious data collection practices. Drawing on insights from three notable research works, the proposed approach effectively categorizes movie reviews and general online comments into positive, negative, or neutral sentiments. Emphasis is placed on thorough data preprocessing, accurate classification, and ethical data management, resulting in a practical and adaptable solution for sentiment analysis in real-world applications.

DOI: http://doi.org/

 

Published by: vikaspatanker

Authors: Professor Dr. Rachna Chavan, Ashish Singh

Abstract: Individuals with hearing and speech impairments rely on Indian Sign Language (ISL) for communication. Despite its importance, ISL lacks broad technological integration, limiting accessibility. This paper presents a vision-based recognition model built using Convolutional Neural Networks (CNNs) to classify static ISL gestures. The system undergoes preprocessing, augmentation, training, and real-time classification. A custom dataset was collected to ensure diversity in hand gestures and backgrounds. Our trained model achieved a classification accuracy of 96.4%, showing its capability to assist in inclusive communication tools.

Published by: vikaspatanker

Authors: Aarti Dengale, Dr. Nagsen Bansod, Dr. A. A. Khan, Dr. R. S. Deshpande

Abstract: Data leakage in cloud environments poses serious threats to data confidentiality, integrity, and availability. This paper proposes a robust system combining Role-Based Access Control (RBAC), watermarking, and anomaly detection to prevent and detect data leakage in real-time. Simulations demon- strateim proved performance in detecting unauthorized access attempts. We present the architecture, algorithms, requirements, and security mechanisms, along with a comprehensive literature survey and experimental results. Additionally, the paper discusses the integration of emerging concepts such as Zero Trust Architecture, Attribute-Based Access Control, and forensic auditing to enhance cloud data security.

 

 

Published by: vikaspatanker

Authors: Divyarth Rai, Tasneem Jahan

 

Abstract: Traditional image retrieval methods which use plain images suffer security risks in fields like medicine, military, space exploration, stocks and finance. Image classification using adversarial machine learning models are vital for enhancing security and detecting intrusion. This paper attempts to present a comparative study and highlights the potential of most promising models for efficient and effective retrieval with feature learning in image classification tasks. The best approaches can eventually strengthen its impact on the field for further implementation. The various machine learning models which could intercept adversarial attacks are classified with their results and advantages. Across social media websites and recommender systems, malicious advertisements are increasingly popular. The approaches discussed here are robust to classify the advertisement images as malicious or benign. This is a good strategy for ensuring smooth user experience and maintaining user security.

DOI: http://doi.org/10.61137/ijsret.vol.11.issue3.153

 

Published by: vikaspatanker

Authors: Assistant Professor Yamini Warke, Sairam Misal, Tushar Karwar, Nikita Wagh, Aishwarya Sawant

Abstract: Agriculture is vital to India's global economy and significantly contributes to GDP. As the human population grows, the nation's agricultural output is crucial in ensuring food security. Climate factors such as temperature, precipitation, soil quality, and fertilizers primarily influence a crop's yield. The variability of these elements adversely impacts productivity, posing a significant challenge for the agriculture industry to accurately estimate crop yields under fluctuating climatic circumstances. Recently, researchers have used machine learning algorithms to forecast crop yields before actual planting. This research study has introduced a machine learning technique, namely linear regression and multilayer perception, to forecast crop production based on characteristics such as state, district, area, seasons, NPK, pH values, rainfall, temperature, and area. To improve yield, this research study recommends a fertilizer tailored to soil conditions, including NPK levels, soil type, pH, humidity, and moisture. Fuzzy algorithms primarily guide the recommendation of fertilizers.

DOI: http://doi.org/

 

 

Published by: vikaspatanker

Authors: P.Selvaraj, A Joshua Issac, Dr.S.Shanmuga, M.Bharathi

 

 

Abstract: The increasing advancement of generation of deepfake techniques – especially manipulations involving face-swapping has brought up major concerns related to integrity of online media, data privacy and societal trust. The computer generated videos, created using advanced models can easily replace an individual face with another often fool regular detection tools because changes in lighting, skin tone, facial expressions are so small and hard to notice. Although many AI-based methods have been developed to spot deep fake, most current models still struggle because they only look at single images , don't consider changes over time or require too much computing power. This research proposes a hybrid deepfake detection framework that leverages the strengths of Convolutional Neural Networks (CNNs) for robust spatial feature extraction and Vision Transformers (ViTs) for capturing temporal and contextual relationships across video frames. The CNN part looks for small changes and edits in the face, while the Vision Transformer looks at a series of frames to catch unusual expressions , movements and facial tone. Together, this combination aims to overcome the challenges posed by diverse and highly realistic face-swap techniques. The system is trained and tested on known datasets like FaceForensics++ and DFDC-Preview, providing a complete way to detect face-swap deep fake. By improving on current methods and looking at both the details in each frame and changes over time, this study helps create a stronger and more flexible deepfake detection system that can handle new and growing threats in visual content.

DOI: http://doi.org/

 

 

Published by: vikaspatanker