Kabombo Katutwa, Dani E. Banda, Julien Shabani

Adware, or advertising-supported malware, is a term used to describe unwanted software that displays advertisements. Adware can infect and root-infect a device, forcing it to download specific adware types and allowing attackers to steal personal information therefore making it a major threat category for consumers. Little research has been conducted on android adware analysis. In this research, the goal is to leverage the use of visualization and transfer learning to detect android adware, to show that a pre-trained neural network can be used to detect android adware from benign application package (APK) files and to compare the performance of selected pre-trained neural networks in accomplishing this task. Using the CICMalDroid2020 dataset, the Dalvik Executable (DEX) files were extracted from APK files and then converted to grayscale images. These images were then used as input to pre-trained deep neural networks that have been trained on the ImageNet dataset. The pre-trained models used are VGG16, ResNet 50, Inception v3, EfficientNet v2 and MobileNet v2. The results show a maximum performance measurement 92% on the F1 – score metric which was achieved by the MobileNet v2 model.

Lukumba Phiri, Simon Tembo

The use of hydro energy to generate electric power is crucial to meet the increasing energy demand of a modern economy and maintains the stability of power supply for unstable sources of power like solar, and wind. In newly constructed hydropower plants (HPPs), the trend among control systems is to adopt contemporary digital and cyber-based systems at the expense of obsolete analog hard-wired systems. Therefore, cyber-physical security is a critical issue in reliability-constrained HPPs. In this paper, we present different levels/layers of protection to manage cyber security. We adopt generalized stochastic Petri nets to quantitatively evaluate the intrusion probability. We then propose a new cyber framework and show that the proposed framework conforms to NIST cybersecurity regulations. Finally, we discuss dependability through three metrics, i.e., reliability, maintainability, and availability. A case study is presented to demonstrate that the proposed cyber framework is highly dependable through analyzing steady-state probabilities.

Lukumba Phiri, Simon Tembo

Supervisory Control and Data Acquisition (SCADA) networks are used across the globe to manage commercial and industrial control systems connected to energy, water, and telecommunications infrastructures. The connectivity provides immense benefits such as reliability, scalability, and remote connectivity, but at the same time exposes an otherwise isolated and secure system, to global cybersecurity threats. This inevitable transformation to highly connected systems thus necessitates effective security safeguards to be in place as any compromise or downtime of SCADA systems can have severe economic, safety, and security ramifications. To enhance the reliability and resilience of power grid networks there is a paradigm shift from legacy networks to smart grid networks. The Zambian power grid operator and mining companies in tandem with global players are transitioning from legacy-based protocols to Internet Protocol-based (IP-based) communications. Therefore, estimating possible cyberattack impacts and identifying system vulnerabilities are a concern in SCADA management and operations. However, it is quite difficult to plan, execute and review vulnerability analysis in critical infrastructure systems as well as in industrial control systems (such as SCADA systems) due to the complexity, and heterogeneity of these systems. A consistent domain-specific conceptual model is required to consequently establish a generic framework for cybersecurity analysis to examine and investigate security threats on smart grid systems. This paper proposes the use of Petri nets to model a framework for cyberattack response for smart grid infrastructure.

Tebepah T., Onuodu F. E., Okolai D. B.

Sending hidden messages through images has a problem of delivering error messages to the recipients. One of the ways that has been used to correct these error messages is by a process called iterating embedding and extraction. In this work, an enhanced error correction code on information extraction from images using parity test has been developed and implemented. The proposed system was built on the existing iterating embedding error correction by adding an additional error check called the Parity correction code. The Object Oriented System development methodology (OOSDM) approach was used for in this project. The system was implemented using C++ and hypertext preprocessor (PHP) for the interactive interfaces. The proposed system had an error free message delivered to the recipient at a more optimal time when compared to the system that used the iterating embedding and extraction process. This system will be of relevance to Enterprises and individuals that use cloud computing for storing information.

Damith Wickramanayaker, Col. H. A. I. U. Hewaarachchi, Sharlene Brown

The purpose of this educational research study was to investigate the effect of E-Learning in Science education at General Certificate of Education at Ordinary Level in Sri Lanka. This research examined (i) the differences of the results of students who used E-Learning and who did not (ii) whether the differences were positively or negatively related (iii) and whether the differences are statistically significant or not. This study was a quasi-experimental study design (nonequivalent control group design) which included a nonrandomized sampling design involved an experimental group (treatment group) and a control group both have a pretest and posttest assessment. The study population comprised 270 students of a girl’s school in the western province. Due to the convenience of preparation of additional E-Learning material, one English medium class was selected as the treatment group. As there was only one English medium class, the control group was selected from one of the Sinhala medium classes. Each group consisted of 37 students. This study has one independent variable, which was video lessons, and one dependable variable, which was student performance. The quantitative data was collected through pretest and posttests. A question paper was given at the pretest for both the groups at the end of one month of teaching of the chapter one lessons. The treatment group was intervened by giving a CD which contained video lessons. The posttest was conducted on both groups using the same question paper after one month. A questionnaire was given to the students of treatment group and a face interview was conducted with the control group students to obtain students’ background environment and activities at school and at home. The study has a null hypothesis (H₀) which was E-Learning does not enhance students performances and the alternative hypothesis (H₁) which was E-Learning enhances students’ performance. The directional hypothesis was tested using one tailed t-test in order to find out whether there was a statistically significant deference between the posttests means of the treatment and control groups. Three additional tests between groups’ and between pretests and posttests were also conducted in order to confirm the validity of the results. The findings showed that there was a statistically significant deference between the means of posttests of both groups. The treatment group posttest mean was higher than the control group posttest mean. Therefore, the results suggested that the intervention might have helped to increase the students’ performance.

Jay P. Narain

The solution of Partial differential equations has been of considerable interest lately because of interest in Machine Learning methods [1,2,3]. The use of artificial neural network to solve ordinary and elliptic partial differential equations has been elaborately described in these papers. Presently some of the salient features of a new simpler artificial neural network has been investigated. While in this neural network one can introduce random grid points and modelling errors, the solutions presented in the paper show that the new ANN scheme is both able to minimize the modelling errors and increase solution efficiency of elliptic partial differential second order equations.

Jay P. Narain

The solution of Partial differential equations has been of considerable interest lately because of interest in Machine Learning methods. The use of artificial neural network to solve ordinary and partial differential equations has been elaborately described in the works of Lagaris, Likas and Fotiadis [1]. Although finite difference, finite element, and other numerical and analytical methods are computationally more efficient, the machine learning methods like artificial neural network offer the hope of solving complicated non-linear equations without deriving analytical methods. The neural network method [1] uses boundary conditions embedded with neural network. Some neural network experience is required to use such method. Presently a very simple neural network scheme is presented where boundary conditions are explicitly applied in the loss function. The neural network differentiation method is based on the work of Luis Angel [2].

Prashant Tyagi, Sharada Devi P. P.

Hybrid-Cloud Environment is a promising area of research, considering the greater flexibility and agility that it provides to businesses for moving their workloads and applications onto the cloud environment, as business requirements and budgets fluctuate. Governments, nonprofits, educational institutions, Product and Service Oriented Companies are all leveraging a hybrid-cloud deployment model to modernize their Information Technology (IT) platform, improve service availability, expand IT capacity, build highly reliable, resilient fault tolerant applications, systems and infrastructure on cloud and innovate faster. This paper discusses best practices on how organizations can seamlessly integrate their on-premises infrastructure with cloud native solutions from various cloud vendors for extending and optimizing their hybrid-cloud environments. In response to the complexities involved in the hybrid-cloud deployment, the work in this paper also discusses some of the popular use cases and how organizations can leverage cloud analytics, artificial intelligence tools and scalability of the cloud whilst also maintaining greater control and operational flexibility over their data for compliance purposes.

Balagopal Ramdurai

In the current years, due to the significant developments in technologies in almost every domain, the standard of living has been improved. Emergence of latest innovations, advanced machinery and equipment especially in the healthcare domain, have simplified the diagonalizing process to a wide extent. Smart techniques employed in medical applications resolved the detection and rectification of various diseases. This work involves analysing how usage of Artificial intelligence practice can improve patient experience across the gamut of the healthcare industry primarily in the Out Patient (OP) Segment. As per industry statistics 88% of the patients are willing to shift the healthcare provider without hesitation. A 2018 study by The Beryl Institute found that 91 percent of patients think patient experience is extremely or very important, and it has an impact on the healthcare decisions they make.

Sandeep Jayashankar, Subin Thayyile Kandy

This document specifies a method of creation or generation of software logs that would further assist in building more granular access control definitions. The technique relies on including an authorization token within each log record, which is generated using a signed JSON Web Token (JWT). Each authorization token embeds all factual information that lets the log viewers set these access constraints, either using an external access control list or applying an access control list outlined in the log management platform.