Research & Review: Machine Learning and Cloud Computing (e-ISSN: 2583-4835)

Balancing Performance and Privacy in Machine Learning: Issues and Practical Solutions

Ritu Kaushik — 2026-06-24

Finding the correct balance between model accuracy and data privacy has emerged as one of the current machine learning's most significant concerns. On the one hand, companies seek extremely precise models that produce solid predictions. On the other hand, they must safeguard the sensitive information needed to train those models, especially as concerns about data misuse and security breaches continue to grow. This balance necessitates a thoughtful combination of technology techniques and acceptable data practices. Differential privacy, federated learning, and secure multiparty computation are examples of privacy-preserving techniques that can minimize raw data exposure while still enabling models to identify significant trends. However, the risk of privacy leakage can be reduced without appreciably impairing performance by using strategies like regularization, adversarial training, and controls that prohibit models from memorizing certain data points. The entire machine-learning pipeline is further protected by robust security measures, such as encryption, access controls, and ongoing monitoring. However, these safeguards might create noise or limit the amount of information available to the model, thereby leading to decreased accuracy if not effectively controlled. As a result, teams must constantly test and change their techniques, considering the trade-offs based on the project's objectives, risk level, and regulatory requirements. Finally, establishing a balance between accuracy and privacy is a continuous process that involves coordination among data scientists, security professionals, and policy experts.

Predictive Health Monitoring: Leveraging IoT and Cloud Computing

Dipali Tupe — 2026-05-30

The Internet of Things (IoT) has developed significantly in recent years. Cloud computing plays an important role in data storage and processing. During COVID-19 many people lost their lives due to a lack of monitoring of vitals. Keeping track of that many vitals was the challenge. This Smart Health monitoring system (SHMS) effectively monitors patients’ health status and saves lives on time. SHMS helps in monitoring many vital signs like body Temperature (BT), heart Rate (HR), toileting habits, blood pressure, as well as sleep patterns. Think of this system as a smart health companion that quietly looks after the user. It uses small sensors, a data collection unit, and a microcontroller, like Arduino, to keep track of the health in real time. All this information is instantly sent to a secure cloud platform, where intelligent tools continuously check for anything unusual and quickly alert a doctor if something seems off.

It doesn’t just stop there; the system can also connect with a decision support system (DSS) to create clear and helpful medical reports, making it easier to understand health over time. With an accuracy of 91.68%, it offers a dependable and efficient way to stay on top of well-being, especially when regular hospital visits aren’t possible. What makes it even more valuable is its ability to adapt during health emergencies. It can monitor additional vital signs like breathing rate and oxygen levels, which are especially important during illnesses such as COVID-19 or the flu. Overall, this system makes healthcare more accessible, helps detect problems early, supports timely treatment, and reduces the burden on hospitals—acting like a constant, caring presence for health.

Phishing Attack Identification using Hybrid ML Models

Ankita Jangir — 2026-06-30

A phishing attack is a type of threat where attackers steal peoples personal and financial information for bad purposes. Many strategies have been used to stop cyber threats and find patterns in data. Machine Learning and Deep Learning are very important in this area. Some models like Long Short-Term Memory (LSTM), Support Vector Machine (SVM) and Convolutional Neural Networks (CNN) are often used to detect and prevent phishing attacks and other cybersecurity threats. This research provides a method for identifying phishing emails using several machine learning algorithms, such as Support Vector Machine and Logistic Regression (LR). Two different datasets were used in this research. The important features were identified, and they included a mix of content-based, URL lexical-based and domain-based attributes. Then various machine learning models were. Their performance was compared. The results showed that feature selection is crucial in improving the performance of the models. The study also tried to find the important features that help detect phishing websites accurately. The experimental results show that optimized preprocessing and feature selection greatly improve the effectiveness of phishing email detection systems. Phishing attacks are a problem, and using machine learning algorithms and feature selection can help prevent them. Machine Learning and Deep Learning are key to stopping these attacks. Phishing emails and websites can be detected using these methods. The findings of this research can help improve cybersecurity.

A Hybrid LLM and NLP Framework for Aspect-Based Sentiment Analysis of Mobile Application Reviews

Ramtilak Nadar — 2026-05-30

Feedback from mobile app reviews helps uncover how applications perform, feel, and function in real-world usage. However, interpreting large volumes of unstructured textual data remains a significant challenge. A hybrid approach combining traditional Natural Language Processing techniques with Large Language Models forms the foundation of this methodology. The process begins with automated data collection followed by text preprocessing to improve data quality. Aspect extraction is primarily performed using LLM-based semantic understanding, with fallback statistical methods ensuring robustness and completeness. Sentiment classification relies on lexicon-based approaches to determine emotional polarity efficiently. Summarization is achieved using transformer-based models capable of generating concise and meaningful representations of user feedback. Visualization techniques, such as word clouds and sentiment distribution charts, highlight frequently discussed topics and overall sentiment trends. By integrating statistical and semantic techniques, the system improves resilience and contextual accuracy compared to standalone approaches. Experimental results demonstrate effective identification of key aspects such as performance issues and functional defects, along with their associated sentiment polarity. These insights support developers in aligning application improvements with actual user feedback.