8333 research outputs found
Sort by
A Deep Learning-Based Plant Disease Detection and Classification for Arabica Coffee Leaves
No full textCoffee leaf disease is a growing concern to coffee agroforestry, predominantly caused by pathogenic fungi and, to a lesser extent, bacteria and viruses reducing the yield and adversely affecting the quality of the coffee. Detecting and controlling these diseases in their early stages represent formidable challenges, since traditional methods rely on visual observation by experts and often fail in accurate diagnosis. Machine learning (ML) techniques are alternative solutions for automating the classification of plant diseases.
This research study aims to provide a comprehensive understanding of the strengths and weaknesses of various deep learning models and transfer learning approaches such as EfficientNetB0, MobileNetV2, CNN and VGG16, shedding light on their effectiveness in addressing the complexities of the multi-class label problem in the context of leaf disease detection in Arabica coffee leaves. Utilizing the “JMuBEN” dataset with 58,405 arabica coffee images across five classes consisting of four diseases such as Phoma, Cercospora, Leaf Rust, Miner and one class of healthy leaf images, the research aims to comprehensively assess each model's efficacy. The test accuracy of the investigated models ranged from 32.49% to 99.7%. EfficientNetB0 outperformed all the models with a test accuracy of 99.72% and an overall F1 score, Recall and Precision of 99.7%. Beyond Arabica coffee, the findings may extend to Robusta coffee and have broader applications in crop disease detection
Deep Learning Framework for Analysing Land Use and Land Cover Classification and Change Detection
No full text
Achieving Green Data Centres for Sustainable Cloud Computing
No full textThe growth of cloud computing comes with a cost, data centres, extremely power hungry and large CO2 producers, are becoming increased sustainability problems. Their electricity consumption might emerge at 3000 TWh annually by 2030 according to estimates, or 8000 TWh in worst-case scenario. This research explores how data centre energy consumption can be minimized to extend cloud reliability and efficiency.
Using Google Cluster Workload Traces, derived from real world dynamic workloads in Google’s Borg clusters, the research examines computational resource utilization and energy profiling. Machine Learning models - GBDT, DQN, and HEPGA are used to estimate energy consumption in a cloud computing data centre to optimize on resource utilization and optimal resource scheduling. Simulations of energy consumption, conducted using Python’s SimPy framework, demonstrate energy-saving potential under various workload conditions.
The results highlight the power of predictive modelling with low MSE in forecasting energy demands and the impact of dynamic scheduling algorithms on reducing consumption. This approach demonstrates the path towards green and high-performance data centres for the next generation meeting the ecological issues of cloud computing
Impact of containerization in reducing vendor lock-in
No full textThe adoption of cloud is on an all time rise due to the numerous advantages it provides. These advantages include pay-as-you-go pricing, flexibility, on demand resources and scalability. One of the major advantages of cloud is that it relieves users of managing the hardware infrastructure. Most companies when deploying their applications use one of the famous providers like AWS, Azure or Google. By utilizing all services from the same cloud provider and using their easy to use deploy UI services, they become dependent on that one cloud provider for all their deployments. ”Vendor lock-in” is the term commonly used for these scenarios, and with the increase of use in cloud without the proper knowledge and tactics used by providers to enforce their services increase, this is becoming a challenge for organization. A potential solution to this issue is the approach of using a multi-cloud or hybrid cloud strategy to deploy applications but it is not very complex to implement and not very cost efficient to maintain. This research will explore the use of Containerization to tackle this challenge by making flexible and portable cloud deployments by using technologies like Docker. This study will highlight the performance and cost benefits of deployment containerized applications to cloud as compared to non-containerized ones and will compare the migration possibilities across multiple clouds. This will offer many guidelines on how to mitigate vendor lock-in and optimize deployments to the cloud
Enhancing Performance and Security in Mobile Cloud Computing Through Machine Learning-Based Edge Computing Integration
No full textThe popularity of Mobile Cloud Computing (MCC) is facilitated by innovative implementation, and yet it suffers from persistent challenges in performance optimization as well as security. For these critical challenges in MCC environments, this research investigates how machine learning can be integrated with edge computing. We propose new approaches for resource allocation with deep reinforcement learning and advanced neural networks for security threat detection at the edge layer. This research shows through comprehensive experimentation and analysis that machine learning models can tremendously help reduce latency, improve resource utilization and detect MSS threats in MCC environments. The results provide important steps towards designing intelligent edge computing solutions alongside a framework for future mobile cloud computing optimization
Image Security in Cloud using hybrid Compression and Encryption Technique
No full textThe exponential growth of cloud-based image storage demands efficient, secure, and scalable solutions for managing large volumes of image data. This research introduces the Cloud Hybrid Compatible Algorithm (CHCA), which is an advanced framework for image compression, encryption, and meta-embedding to improve storage, security, and data integrity. Using 2D-Discrete Wavelet Transform (2D-DWT) for compression and a SHABlowfish encryption model, CHCA has made developments in the aspects of compression ratio, security, and scalability. Meta-embedding using SHA-256 means that image traceability, tampering, and self-checking are possible.
Evaluation reveals that the proposed CHCA framework is more effective than current strategies in critical success factors. For JPG images, the proposed method provides an average compression ratio of about 1.43, PSNR of 40.96 dB and up to 30% overhead reduction in encryption. In the case of PNG images, the proposed CHCA achieves the average compression ratio of 0.88, PSNR of 44.92 dB, and the encryption overhead is reduced up to 6%. Unlike other approaches, CHCA provides real-time, serverless processing through AWS Lambda, which is flexible and inexpensive. The system retains high-quality image and provides adequate security against the unauthorized access and data theft. More will be done in the future to enhance the decryption process as well as to identify future uses of CHCA in complex, multiple cloud environments
IaC for Secure Serverless: A Comprehensive Approach to Deployment and Configuration Management
No full textServerless computing has started to be adopted by more and more organizations in the last decade, but this widespread adoption has also introduced a slew of security challenges. Human error, a common cause of security vulnerabilities, can be significantly reduced by using Infrastructure as Code tools like Terraform to provision and manage infrastructure. This study evaluates the impact of using such tools to deploy a simple serverless application built on AWS Lambda. Potential threats will be identified using threat modeling frameworks and the Iac tools will be appraised on their ability to effectively minimize these threats. The experimental results show that security tools embedded in a pipeline identified 100% of intentional vulnerabilities, significantly outperforming manual inspection. The experiments also found that IaC tools reduced deployment time by over 80%. The research aims to establish a framework for secure IaC deployments in cloud-native environments, and also establish practical guidelines for developers to enhance security in their serverless applications
Hybrid Detection of Cross-Site Scripting (XSS) Vulnerabilities in Web Applications
No full textCross-Site Scripting (XSS) is a vulnerability for web applications and their security, by which attackers can inject malicious JavaScript into the system, leading to account takeover, session hijacking, or cookie stealing, among other issues. To address this, the research introduces a novel hybrid XSS detection approach and its implementation in the form of a tool named XSSFind. This tool integrates Static Application Security Testing with Dynamic Application Security Testing methodologies through white-box and black-box testing techniques to improve the discovery rate of cross-site scripting vulnerabilities. This work was motivated by existing tools that rely either on static or dynamic analysis but cannot provide comprehensive coverage. XSSFind offers the strength of both static and dynamic approaches, detecting vulnerabilities at the code level and runtime level-wise, thereby enabling a comprehensive security assessment. Results show that these combined approaches have given required results in terms of finding actual XSS vulnerabilities, making them a promising addition to the web security community. The findings further suggest directions for future improvement, such as extending payload libraries and detection models based on machine learning