The increasing distribution of modern data ecosystems across geographically dispersed and isolated nodes demands machine learning techniques capable of functioning without centralized access to raw data. This paper proposes an advanced collaborative learning framework tailored for non-co-located datasets, addressing challenges such as privacy restrictions, communication bandwidth limitations, asynchronous client participation, and data heterogeneity. The proposed scheme integrates adaptive aggregation, gradient compression, asynchronous synchronization, and privacy-preserving mechanisms such as secure aggregation and differential privacy. Experimental evaluations on MNIST, CIFAR-10, and synthetic non-IID benchmarks demonstrate significant improvements in convergence speed, communication efficiency, and model robustness compared to classical federated learning methods. In particular, our approach achieves up to 9–12

Crop diseases play a significant role in food production globally; therefore, there is an urgent need to develop quick and accurate diagnostic techniques that are more effective than manual inspection methods. The proposed hybrid multimodal learning framework in this research provides a solution that integrates adaptive therapy suggestion, market price prediction, and image-based disease detection. This study also proposes a framework for pesticide recommendation and the treatment of plants. This study experiment on tomato and cotton crop leaf data for disease detection. Experimental results on a tomato crop disease detection dataset show that the proposed model shows high performance. EfficientNetB0 provides more stability and generalization capabilities in different scenarios compared to other models, such as YOLOv8, ResNet50, and a custom CNN model. The use of a knowledge-based decision support system provides sustainable pesticide recommendations based on environmental and symptom-specific parameters. Forecasting of pesticide prices through LSTM methods yields forecasts within 3.2% and 4.1% MAE, enabling improved decision-making by providing instant points of reference for potential price movements. Research uses SHAP and LIME to provide explainability to users, thus improving user buy-in through transparency. Overall, this modular system provides a data-driven decision-making model to improve the efficiency of managing crops.

Prasad Chaudhari1, Ritesh V. Patil2 Parikshit N. Mahalle3 1Department of Computer Engineering, Smt. Kashibai Navale College of Engineering Research Center, Savitribai Phule Pune University, Pune, Maharashtra, 411041, India 2Department of Computer Engineering, PDEA's College of Engineering, SPPU, Pune, Maharashtra, 411041, India 3Department of Artificial Intelligence and Data Science, Vishwakarma Institute of Technology, SPPU, Pune, Maharashtra, 411041, India

The integration of Federated Learning (FL) in Healthcare Internet of Things (H-IoT) has the potential to enhance data privacy, security, and model generalization while enabling collaborative learning across distributed healthcare nodes. FL allows multiple healthcare institutions and edge devices to train a shared model without sharing raw patient data, thus ensuring regulatory compliance and reducing privacy risks. However, implementing FL in H-IoT presents several challenges, including communication overhead, heterogeneous data distributions, and resource-constrained edge devices. This paper presents a comprehensive analysis of the state-of-the-art FL techniques applied in H-IoT, highlighting key challenges, methodologies, and future research directions. We explore different FL architectures such as centralized, decentralized, and hierarchical models, as well as communication efficiency techniques like model compression and adaptive client selection. Moreover, the privacy preserving mechanisms, namely differential privacy and secure multiparty computation are also studied. This study gives a comparative analysis of FL approaches and considering their respective performance, scalability and security features to serve as a valuable reference to researchers and practitioners in healthcare AI. The future directions of FL in H-IoT are to increase communication efficiency, model robustness, and make the deployment in real world feasible.

In recent years, there has been a boom in the Internet that we know today, and it forms a vital part of transferring data, data analysis, and sharing of resources. The sharing of resources becomes a serious concern when the data is highly valuable, especially in the case of Medical Cyber-Physical Systems. Since sensitive data is sent using some protocols that could be easily intruded into, with the help of emerging quantum systems in a matter of seconds, the security of the data becomes a major concern. This paper presents a model of ECC-based encryption/decryption using quantum cryptography. This model ensures the integrity of the data and alerts the system if the data has been intercepted or eavesdropped by an external entity by using the quantum properties of the quantum key and also the algorithm needed to implement the system.

Cotton cultivation is one of the pillar crops in the Indian agriculture; however, it is under high pressure from pest and insects attacks leading to yield losses of 24.4 % (ICAR, 2019). This paper proposes a ensemble deep learning-driven method for early detection of cotton leaf diseases and contextaware pesticide suggestion system. A literature review outlines available literature, efforts, and holes in cotton disease identification and pest management. We compared/presented fine-tuning techniques based on various pre-computed CNN architectures (VGG19, ResNet50, ResNet152V2, and InceptionV3) and evaluated their performance on a real-time, field-acquired indigenous cotton leaf image dataset. The hybrid CNN model obtained the best accuracy of 97.04 %, and showed the best performance in most measures. In addition, a pesticide recommendation module gives high priority to a symptom similarity and provide the top three pesticides suitable with JSON format so that it can be available through any mobile or web API. Our unique integrated solution has been developed to help Indian farmers make accurate, timely and actionable decisions which help to increase productivity while reducing crop loss.

As the world becomes more computerized, keeping data transfer safe has become very important. Quantum computers and other strong computers have the ability to break classical encryption methods, even though they are widely used. Quantum Key Distribution (QKD) is a new way to protect data. It uses quantum physics to make sure that the sharing of cryptographic keys is safe. The aim of this study is to look into how QKD can be used to improve the security of data transfer. The suggested method combines QKD with current encryption standards to create a safe and effective way to send data. The method includes setting up a key sharing system based on QKD and then using symmetric encryption algorithms to keep the data being sent safe. The system is tried in different communication settings to see how well it protects against spying and "man-in-the-middle" attacks. The study's results show that QKD greatly improves the safety of sending data by keeping the cryptography keys secret. The results show that QKD has the potential to stop people from getting in without permission and find any attempts to change the keys while they are being sent out. The table below shows performance numbers that show how well QKD works in real-world situations. It shows how well communication speeds and security measures compare.