8333 research outputs found
Sort by
Block chain Enhanced Personal Health Records Sharing with Integrity Verification
No full textMainly, the need to share the personal health records (PHRs) is highly important in the current society due to the raging use of technology. This project specifically attempts to solve the problem of efficient PHR management through the incorporation of block chain technology. The system involves three primary stakeholders: administrators, patients who are the data owners and data users that are hospital staff, insurance companies, and other medical personnel. The existential question in such research includes is the data stored on the block chain tampered in any way. Proposed system effectively authenticates the data integrity with the help of SHA hashing to prevent unauthorized modifications in data. It is a hybrid cloud where the application part of it runs on a private cloud for security reasons while data is on the public cloud for efficiency. Attribute Based encryption guarantees that the health records data is only accessible by the right personnel while making the data searchable, improving performance, and protecting the information from the alteration
Enhancing security efficiency through the integration of Elliptic Curve Cryptography with Audio Steganography
No full textThis research combines Audio Steganography with Elliptic Curve Cryptography (ECC), focusing on the efficiency part of security to handle the performance problems in standard RSA encryption. ECC is a possible substitute because it has smaller key sizes and requires less operations. The study hopes to create a system that uses ECC's benefits in a many-layered security structure through audio steganography.
The objective of this research is to design a security app combining ECC with audio steganography which will offer faster processing and low computational power. The method includes implementing ECC algorithms in Python, using multilayer Least Significant Bit embedding for data covering up in audio files. Expected outcomes will include improved encryption efficiency, better data security measures, for secure data transmission and storage
Protecting User Privacy: Advanced Techniques for detecting and preventing Keyloggers Using Machine Learning
No full textA keylogger is a type of malware that silently monitors and logs all keystrokes that the user types on the keyboard, and it presents a grave threat to the privacy and integrity of the data of user. Conventional Anti-virus systems have limitations in detecting unknown or well-concealed keyloggers and therefore require more sophisticated detection techniques. The general goal of this study is to build novel approaches for detecting keyloggers utilizing machine learning to protect the users’ privacy. The focus goals are to establish viable and specific models for the identification of keyloggers and compare their effectiveness with prior methods; also, the study examines the implementation possibilities. The methodology used entails the data acquisition and preparation step where the different keyloggers and benign applications are obtained and preprocessed. The next step is featuring extraction where measures of the anomalous behavior of the keylogger are extracted. Different machine learning approaches, particularly reinforcement learning techniques, such as Q-learning algorithms, are applied to train models to detect keyloggers using system monitoring and API call analysis techniques. As a result, the models undergo a strict evaluation process and involve the use of measures such as accuracy, precision, and the rate of false positives to evaluate and enhance the model’s performance. The current study has shown that machine learning and reinforcement learning, especially reinforcement learning, holds massive possibilities in detecting keyloggers with relatively low false positives
Enhancing SDN Access Control with Private Ethereum Blockchain
No full textSoftware Defined Networks (SDN) have been implemented in various environments requiring agility, rapid and automated response to changing network circumstances and centralised management control of all network components. This is achieved by decoupling the data plane from the control plane in the SDN paradigm. Cloud Computing and Internet of Things are the most common environments where SDNs can handle large-scale network traffic loads efficiently and dynamically. However, the central management control of the entire network has also its drawbacks. Blockchain (BC) technologies come with numerous security-enhancing benefits and BCs offer decentralised design which could effectively complement centrally controlled SDNs. This paper highlights various aspects of combining BC with SDN to address inherent security concerns related to SDNs. BCs address all three aspects of the CIA Triad and this paper discusses their main benefits in the context of SDNs, which are decentralised design, data immutability and integrity, enhanced authentication and non-repudiation. Ethereum BC networks and the Proof of Authority (PoA) consensus mechanism are of particular interest in this paper. The PoA consensus is best suited for private Ethereum BC implementations
Enhancing E-commerce Supply Chain and Shipping Efficiency with Machine Learning and Deep Learning Models
No full textE-commerce businesses are increasingly prioritizing with the efficient supply chain and shipping processes to meet rapidly growing customer demands while also mitigating the operational costs. Despite advancements in machine learning (ML) and deep learning (DL) techniques, existing works predominantly focuses on predicting mainly late delivery risk rather than estimating scheduled and actual shipping durations which is very critical aspect for enhancing supply chain resilience. To address the gap, this study aims to predict both scheduled and actual shipping durations using the DataCo SMART SUPPLY CHAIN dataset. This research aims to make predictions as a multi-output regression problem while also employing extensive preprocessing, feature selection, and hyperparameter optimization to mitigate the prediction error. Results shown that tree-based ML models outperform DL models in capturing complex, nonlinear relationships within tabular data, with XGBoost achieving the highest accuracy. These findings highlight the potential of advanced predictive analytics to optimize logistics, minimize delays, and improve decision-making in e-commerce supply chain operations
Enhancing Customer Retention in Online Games Using Customer Lifetime Value
No full textCustomer Lifetime Value (CLV) is a crucial metric in the online gaming industry, and proper pathway strategies to enhance user engagement, monetisation, and retention. The research dug deeper to find the key factors influencing the CLV in freemium gaming environments and the application of advanced machine learning techniques to predict the CLV accurately. The analysis of the player data, including gameplay behaviour, in-app purchases and churn likelihood, aims to identify the patterns that categorise high-value players from casual users. Predictive modelling is a hybrid approach of convolutional neural network (CNN) and recurrent neural network (RNN) where the output recommends optimising the game design, user retention strategies and personalised recommendations to retain users. The model achieved a testing accuracy of 90 per cent and a minimal loss of 0.36. Personalised interventions based on the CLV predictions lead to enhanced retention rates and faster revenue growth, with the practical applicability of the frameworks. This research bridges the gap between the theoretical CLV research and its application in dynamic ecosystems, offering insights for the developers and stakeholders in the online gaming industry, with contributions to the ever-evolving discourse on the customer centric design to provide a data-driven road map to improve player experience and long term value
Hypertension Risk Prediction using Machine Learning Models and Ensemble Techniques
No full textHypertension has generally been considered the "silent killer" and one of the main health issues in the world because it results in serious complications such as cardiovascular diseases and stroke. The present study investigates the ability of machine learning models to predict hypertension effectively by using combined lifestyle and physiological factors. Thus, this paper uses a dataset based on features comprising age, BMI, cholesterol, glucose, blood pressure, and Smoking status, with preprocessing and certain robust machine learning methodologies in order to derive these predictions. Up to nine different algorithms were compared for their F1-scores, recall, and AUC-ROC-comprised Random Forests, Gradient Boost, CatBoost, and XGBoost. The ensemble method, which combined the strengths of the top performing models, had a very strong predictive power with a F1-score of 0.8528 and recall of 0.8933. Feature importance analysis showed that systolic blood pressure, diastolic blood pressure, age, and BMI were the most influential factors contributing to hypertension risk. The results confirm the effectiveness of machine learning in the detection of hypertension risk. These findings represent a scalable, interpretable solution to improve clinical decision-making. This research therefore calls for an immediate need for data-driven approaches in healthcare to improve patient outcomes
Advanced Threat Detection in IoT Networks using Hyperparameter-Tuned Machine Learning Models
No full textThe use of IoT (Internet of Things) devices has brought several new forms of security risks which cause growth of malware that targets IoT networks. Current techniques for intrusion detection involve a IoT of interaction and may thus not be very efficient in handling new kind of threats.
The purpose of the research is to propose the IoT-2023 based efficient machine learning models for improving the identification of the illicit actions in the IoT networks. The proposed method combines feature selection and classification techniques which are useful for identifying the features that best characterize the malicious and benign network traffics. To get the best from the machine learning models developed hyperparameter tuning is utilized.
Machine learning classifiers have been developed such as Logistic regression, Decision tree, Decision stump, Random Forest, Naïve Bayes, K Nearest neighbours (KNN), Support Vector Machines (SVM), Multi Layer Perception (MLP), Gradient Boosting (GB) and Extreme Gradient Boosting (XGB). All these models work for have been made to work with binary and multi class classification tasks. In addition hyperparameter tuning is done for all the models. There is a significant improvement in the accuracy of the model.
Boosting algorithms like Gradient Boosting (GB) and Extreme gradient boosting (XGB) are working greatly with binary classification of the attack types with 95% accuracy and 95% F1 score each. Decision tree and random forest algorithm is capable to handle multi class classification in a better way with 84% accuracy each. Although hyperparameter tuning is responsible for model improvement, it has no compatibility with models like Naïve Bayes and Decision Stump models. With this accuracy of the model receded with complex classification tasks like 33% accuracy for decision stump and 67% accuracy for Naïve Bayes. With these studies we were able to reach the research objectives
Skin Cancer Detection: Image Classification Using CNN Architectures with CLAHE
No full textSkin cancer is a significant health problem that is common and can be improved if recognized early and avoided. In this paper, the authors examine Contrast Limited Adaptive Histogram Equalization (CLAHE), which is applicable for changing the images to enhance their contrast, to improve the efficiency of various Convolutional Neural Networks (CNNs) in detecting various skin lesions. They used the HAM10000 dataset with 10,015 skin images for testing the three implemented CNNs ResNet, DenseNet, and EfficientNet with raw and CLAHE applied images. The first case results showed that CLAHE works well with models in detecting skin features and increases classification accuracy. The DenseNet turned out to be the one with the highest performance of 77% accuracy and recall when given CLAHE-enhanced images, whereas ResNet came up to 75% accuracy; however, the precision revealed a poorer performance compared to the others for EfficientNet. The study highlights the good use of CLAHE as the best means of improving the detection of skin cancer by means of CNN models, namely DenseNet for most effective results. New image enhancement methods and larger datasets will be discussed in future work
Addressing Cloud Security Challenges using AI-Driven IoT Intrusion Detection Systems with UQ-IDS Dataset
No full textCloud computing and the Internet of Things (IoT) have expanded the attack surface and led to the introduction of new cybersecurity challenges, necessitating improved security measures to protect against the increased risk. However, traditional Intrusion Detection Systems (IDS) have been unable to keep up with use of evolving threats. Issues with traditional IDSs include high false-positive rates and poor ability to detect new types of attacks by utilizing sophisticated techniques. To overcome these challenges in this research, we propose a Conv-LSTM hybrid model that takes the strengths of Convolutional Neural Networks (CNN) in identifying patterns and the strengths of Long Short-Term Memory (LSTM) networks in processing sequential data. By using a hybrid architectural approach to the problem, we can improve both the accuracy of detection and reduce false positives. This research also includes an implementation of a web application that can be used in real time to present alerts to administrators through a friendly web interface. The novelty aspect of this research is the implementation of hybrid model ConvLSTM, which is the most accurate model for detecting the anomalies from the system as compared to CNN and LSTM. Also, the implementation of web application in the cloud environment, offers a realistic and industry valid framework for a scalable and efficient cybersecurity solution for the modern network infrastructure