University of Ibadan Journal of Science and Logics in ICT Research

IoT-Based Gas and Smoke Detection System using Blynk application with Automatic SMS and Alarm Notifications

2024-06-12T06:57:37+00:00

The Internet of Things (IoT) has revolutionized safety and security by providing innovative solutions to critical challenges. Gas leakage, a dangerous chemical from petroleum, can cause health issues and disrupt workspaces. To prevent such accidents and maintain a clean air environment, a monitoring gas leakage detector system is proposed. The system uses a NodeMCU ESP8266 Wi-Fi microcontroller and a combustible gas sensor (MQ-2) to detect the presence of propane, butane, and Liquefied Petroleum Gas (LPG). The sensor's voltage output determines gas concentration, and the ESP8266 sends data to the blynk application. This system aims to maintain a clean and safe workspace. The IoT-based gas and smoke detection system uses sensors and a mobile app to monitor premises remotely. It sends SMS alerts to users and authorities in case of incidents and activates audible alarms to the users. This system enhances security by reducing risks and reducing the need for manual monitoring, making it a significant step towards safer environments.

Machine Learning in Cyber Security Operations

2024-06-12T06:19:29+00:00

The defense of computational devices as well as computer networks against information leaks, theft, and damage to their electronic data, software, hardware, or other components, as well as against interruption or misrepresenting the services they offer, is defined as cyber security by securitystudio.com. In recent years, there has been an unparalleled increase in public interest in machine learning (ML) research. People's learning and working styles are changing as the Internet and social life become more intertwined, yet this also exposes them to major security risks. Protecting confidential data, networks, and computer-connected systems against illegal
cyberattacks is a difficult challenge. Effective cyber security is crucial for this. To solve this issue, recent technologies like machine learning and deep learning are combined with cyberattacks. The write-up covers machine learning technology in cyber security, explores the benefits and limitations of employing them, and offers recommendations for future research. The world of today is highly network-interconnected due to the prevalence of both small personal devices (like smartphones) and large computing devices or services (like cloud computing or online banking). As a result, millions of data bytes are generated, processed, exchanged,
shared, and used every minute to produce results in specific applications. As a result, protecting user privacy, machine (device) security, and data in cyberspace has become a top priority for private citizens, corporate entities, and national governments. Machine learning (ML) has often been used in cybersecurity in recent years, including for biometric-based user authentication and intrusion or virus detection. But ML algorithms are vulnerable to intrusions during both the training and testing phases, which often lead to noticeable performance decreases and security vulnerabilities. Comparatively little studies have been conducted to ascertain the type,
extent, and defense mechanisms of ML methods' vulnerabilities against security threats. Systematizing recent cybersecurity-related initiatives leveraging ML is vital to garner the interest of researchers, scientists, and engineers

A Review of Open-Source Fully Homomorphic Encryption Libraries: Zama.ai Concrete Compiler, Applications and Vulnerability

2024-06-11T15:35:21+00:00

Fully Homomorphic Encryption (FHE) is an advanced cryptographic technique that enables computational operations to be
performed on encrypted data without the need for decryption. In other words, FHE allows operations to be conducted
directly on ciphertexts, producing encrypted results that, when decrypted, correspond to the output of the operations
performed on the plaintext data. This revolutionary capability ensures data privacy and security throughout the entire computation process, as the data remains encrypted at all times, even during computation. FHE schemes typically involve
complex mathematical operations and algorithms, often based on lattice-based cryptography or other mathematical structures, to enable secure and efficient computation on encrypted data. Substantial progress has been achieved in the realm of FHE and its application since 2015, yielding enhanced efficacy, heightened security, and augmented feasibility. This review paper discusses and reviews diverse FHE schemes/libraries, and the extent of progress attained hitherto and how the possibilities of adoption of the scheme in industry is being propagated, using research questions as a guide, we endeavor to utilize searches across various academic databases and industry repositories for peer-reviewed papers, articles, and books. While some of the examined papers suggested new techniques to improve the security of transferred data, several of the publications provided novel schemes for FHE to maximize efficiency and minimize noise. Special emphasis is placed on the open-source tools and libraries implementing FHE scheme, notably Concrete (developed using TFHE Scheme), an innovation by Zama.ai, a preeminent research establishment specializing in FHE research and development. Since writing FHE programs can be difficult, Concrete, based on LLVM, makes this process easier for developers with the ability to compile Python functions (that may include NumPy) to their FHE equivalents, to operate on encrypted data. The applications of the library are examined, encompassing accomplishments, limitations, and vulnerabilities. Conclusively, prospective avenues for advancement are underscored, deliberated upon, and illuminated.

Deep Learning Algorithms for Multiple Cyberattacks Detection

2024-06-12T06:36:21+00:00

Data is pervasive and accessible through the internet. The proliferation of smart devices worldwide, such as computers and mobile phones, has led to a significant increase in internet usage. Consequently, this surge has also given rise to a corresponding increase in cyberattacks, which are a prevalent issue faced by internet users. To address this problem, it is crucial to have an effective cyberattack detection mechanism in place to safeguard computer networks, systems, and data. While intrusion detection systems (IDS) play a significant role in this regard, they do have their limitations. Therefore, in this research, two deep learning algorithms, namely Multilayer Perceptrons (MLPs) and Recurrent Neural Networks (RNN), have been proposed. The NSL-KDD
and CIC-IDS-2017 datasets were utilized for this project. When using the NSL-KDD dataset, the MLP algorithm achieved an accuracy of 99.44% with a false positive rate of 0.52%, whereas the RNN algorithm achieved an accuracy of 98.02% with a false positive rate of 2.21%. On the other hand, when employing the CIC-IDS-2017 dataset, the MLP algorithm achieved an accuracy of 99.98% with a false positive rate of 2.06%, while the RNN algorithm achieved an accuracy of 99.09% with a false positive rate of 39.65%. Furthermore, various metrics such as precision, recall, F1-score, error rate, and others were calculated and compared for both models. The obtained results clearly indicate that the MLP algorithm outperformed the RNN algorithm in terms
of performance when applied to both datasets

Crop Recommendation Analysis and Validation in Nigeria Using Machine Learning Algorithms

2024-06-11T15:49:55+00:00

Crop recommendation systems are crucial for optimizing agriculture by suggesting crops based on environmental and soil conditions. Failure in selecting suitable crops can result in low yields and resource wastage. This study builds an improved recommendation system for Nigerian farmers. Data from various sources, including the Nigeria Metrological Agency, the Agronomy Department University of Ibadan, Ahmadu Bello University Zaria, and Federal University Wukari, were preprocessed using numpy and pandas. The climate parameters used were Rainfall, Temperature and Humidity while the soil parameters were Nitrogen (N), Phosphorus (P), Potassium (K), Calcium (C) and Magnesium (Mg). The pH was used to measure the soil acidity or alkalinity. The 18 crops considered were Bambara Nut, Cassava, Cocoyam, Tomato, Yam, Acha, Cocoa, Beans, Groundnut, Beniseed, Maiza, Rice, Oil Palm, Cashew, Sugar cane, Sweet Potato, Pepper and Coconut. After preprocessing, the dataset was partitioned into training, validation, and testing sets in the ratio 80:10:10. Four Machine learning algorithms which are Random Forest, Naïve Bayes, K Nearest Neighbor, and Support Vector Machine (SVM) were employed, with Random Forest outperforming others in accuracy, precision, recall, and F1 score. Naïve Bayes ranked second, followed by K Nearest Neighbor, and Support Vector Machine performed as the poorest. The models effectively recommended crops for specific climates and soils, with SVM being the least effective. Hence, this study demonstrates the importance of accurate crop recommendations in maximizing agricultural productivity

FUTACOVNET: A Deep CNN Network for Detection of Corona Virus (Covid-19) using Chest X-ray Images

2024-06-11T13:00:56+00:00

In December 2019, WHO declared COVID-19 as morbidity and mortality rates continue to soar high with a global cumulative case of 460,280,168 and cumulative mortality of 6,050,018. The standard clinical golden tool mostly used for the diagnosis of COVID-19 is the Reverse transcription polymerase chain reaction (RT-PCR). It is adjudged to be very expensive, less-sensitive, not readily available in hospitals and most significantly, requires the services of a specialized medical expert. X-ray imaging is an easily accessible tool that can be an excellent alternative tool in COVID-19 diagnosis. This paper proposed a technique to automatically predict the presence of COVID-19 pneumonia from digital chest X-ray images using deep learning. Any technological tool that can help in the effective screening of the COVID-19 infection with high level of accuracy is highly required. In this research, the use of transfer learning approach in the rapid and accurate diagnosis of COVID-19 from chest X-ray images is carried out. A new CNN architecture that is trainable optimally while maximizing the detection accuracy is developed. A database was created by combining several public databases and also by collecting images from National Hospital, Abuja. The database contains a mixture of 3616 COVID-19 and 10,192 normal chest X-ray images. The X-ray images were used to train and validate the deep Convolutional
Neural Network (CNN) model. The trained network was then used to classify the normal and COVID-19 patients. The proposed CNN classification accuracy, precision, recall and F1-Score of the model are 96.5%, 96%, 96% and 96% respectively. The model was then compared with the state-of-the-art CNN models and it outperformed all of them. The high accuracy of this model can significantly improve the speed and accuracy of COVID-19 diagnosis in our local hospitals. This would be extremely valuable during an outbreak of pandemicrelated diseases when there are limited facilities and human resources for early diagnosis and management.

Deep Convolutional Neural Networks Architecture with Pre-Filtered and Segmented Dermoscopic Images

2024-06-12T06:48:23+00:00

Deep Convolutional Neural Networks (DCNN) involve alternating convolutional layers, non-linearity layers and pooling layers for identifying patterns in input. The pooling retains important information while down sampling the dimensionality of the feature map on dermoscopic images used for early cancer diagnosis. Existing DCNN for dermoscopic image analysis employs Max Pooling (MP) and Average Pooling (AP) due to their efficiency. The MP works best on images of dark background with lighter object, while AP works better on images of lighter background with darker object. An online International Skin Imaging Collaboration (ISIC) dermoscopic image dataset obtained from 2016 - 2019 was used for the research. A novel DCNN, IP-DCNN developed and configured with rectified linear unit activation function, multiclass cross entropy loss and Softmax functions.
Evaluation of the IP-DCNN with filtered-segmented images was done by comparing its performance with existing current studies which used DCNN architectures. The developed interpool deep convolutional neural networks provided an improved performance over the pure deep convolutional neural networks and its existing variants.

Food Components Recognition from Still Images Using Multi-Label Learning

2024-06-12T06:06:33+00:00

Food recognition, a recent research area in image processing, helps identify food items to keep track of the food
consumed, thereby maintaining a healthy diet. However, the task of food recognition is challenging due to the deformable nature of food items. Usually, there are more than one food item in a meal making the task more challenging. Therefore, the aim of this work is to develop a deep learning model to detect and enumerate visual food components present in a meal. In the multi-label learning approach, food images were collected to build a food image dataset, which comprised 2150 images. The images were pre-processed. Contrast Limited Adaptive Histogram Equalization was then applied followed by scaling to fit as input into the model for training/testing. Thereafter, Deep (VGG-16) and Dense (DenseNet50) models were used to extract deep features. The final layer of the model was applied with a multi-label technique to train on the selected features. The multi-label model was tested using appropriate metrics in which VGG-16 performed better than DenseNet50 with an accuracy of 91.90%, hamming loss of 8.10%, loss of 0.26%, precision of 73.49%. An independent test set was used on the model which showed impressive results. It was observed from this study that the proposed approach performed excellently well in predicting Nigerian Food components. It is recommended that this work be applied in real world this work in real world scenario such as dietary tracking to monitor food intake. Human-Computer Interaction with automatic purchasing systems at restaurants can be used to speed up services.

Comparative Performance Evaluation of Random Forest on Web-based Attacks

2024-06-11T13:22:50+00:00

The majority of typical online attack methods are thoroughly researched and documented. Countries,
corporations, people, and vital infrastructures that depend on information technology for daily operations have
suffered financial losses, the loss of personal information, and economic harm as a result of web-based intrusion. However, foreseeing an attack before it happens can aid in its prevention. This research proposes a predictive model for web-based attacks and a performance comparison of random forest with and without feature selection to secure the availability, integrity, and secrecy of networks, computer systems, and their data. The CIC-Bell-IDS2017 dataset, which includes typical and contemporary intrusion attacks, served as the raw data source for the proposed model. A python-based programming environment and interface for Anaconda Navigator, Jupyter Notebook, was used to create the predictive models. Performance evaluation and
comparative analysis were conducted, and the results demonstrate that, once big data analytics (feature scaling and feature selection) were applied to the dataset, the models' prediction accuracies improved, creating a potential intrusion detection system. The outcome yielded excellent accuracy and model development times in both cases, with 97% and 98% precision for both sets and model development times of 35 seconds for the raw set and 15 seconds for the reduced set, which is an important factor when deploying machine learning models in a real-time setting. Random Forest is more computationally expensive than Correlation feature Selection-based classifiers, but having higher predictive accuracy, according to a comparison. Both of these methods work well and each has advantages and disadvantages. The use of big data analytics (PySpark) was found to help machine learning models perform better, resulting in better intrusion detection system.