Artificial neural network-based numerical modeling of virus transmission in computer networks
No Thumbnail Available
Date
2023
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
UMT Lahore
Abstract
In this thesis, artificial neural networks (NN) are used to predict virus transmission patterns in computer networks, introducing a new and data-driven strategy for improving network security. This is a difficult challenge because of how complicated the model is in real life. By training the neural networks (NN) on historical network data and virus transmission records, the model learns to recognize underlying patterns and factors that influence the spread of viruses in a network environment. By contrasting the efficiency of neural networks with traditional numerical techniques for resolving complex equations, this study explores a novel viewpoint. To illustrate the potential of the neural network-based technique in defining viral transmission via computer networks, specifically employ the Susceptible, Exposed, Infected, and Removed (SEIR) model. The neural network methodology is compared to widely used numerical methods like the Euler method and the Fourth Order Runge-Kutta method (RK-4) in this study on the dynamics of viral propagation through computer networks. The numerical methods, Euler and RK-4, are really good at staying stable and providing correct answers, but they run into issues with certain step sizes. Once the step size gets too big, both Euler and RK-4 stop working correctly. However, the neural network method works consistently for all step sizes. The importance of this study is seen in its comparison of the Euler and RK-4, and neural network approaches. The neural network method stands out as the best option since it continuously produces estimates that are more accurate, no matter the circumstance. The work highlights the capacity of neural networks to provide more accurate results, highlighting their promising role in boosting network security against emerging cyber threats, even though numerical approaches still serve their purpose well.