Electronic Thesis and Dissertation Repository

Thesis Format



Master of Science


Computer Science


Haque, Anwar


Network failure is the unintentional interruption of internet services, resulting in widespread client frustration. It is especially true for time-sensitive services in the healthcare industry, smart grid control, and mobility control, among others. In addition, the COVID-19 pandemic has compelled many businesses to operate remotely, making uninterrupted internet access essential. Moreover, Internet Service Providers (ISPs) lose millions of dollars annually due to network failure, which has a negative impact on their businesses. Currently, redundant network equipment is used as a restoration technique to resolve this issue of network failure. This technique requires a strategy for failure identification and prediction to run faultlessly and without delay. However, we lack a suitable generic network failure identification and prediction system due to the unavailability of publicly accessible failure data. This study simulates network traffic to gather failure data based on a general network failure guideline. Furthermore, various state-of-the-art Machine Learning and Deep Learning methods were applied to the generated data. Notably, our proposed Deep Learning model for failure identification provides accuracy, precision, recall, and F1 scores in the range of 97% to 99% for three different demonstration networks. Additionally, our proposed Long Short Term Memory model gives low root mean square error rates of 0.9751 for failure prediction.

Summary for Lay Audience

Internet services may be interrupted for various reasons, such as device failure, connection issues, natural disasters, etc. When a network error inadvertently disrupts internet services, customers become frustrated. Especially in a medical emergency, personal safety, or transportation-related emergency, a lack of network connectivity can be catastrophic. In addition, since the outbreak of the COVID-19 pandemic, many businesses require a constant internet connection. In reality, every aspect of our lives depends on the Internet, including emergency needs, banking, entertainment, healthcare, socializing, and creative work. The Internet is now such an integral part of our lives that its absence will be immediately noticeable.

Additionally, Internet Service Providers (ISPs) are impacted by service interruptions from a business perspective, as they lose millions annually due to network failure. As a result, ISPs utilize redundant network equipment as backups if a network device fails. However, the disadvantage of this approach is that it requires time to identify the type of failure and then fix the problem with an appropriate backup. Constructing a system that can predict the type of failure occurring next to using backups effectively would be preferable. But building this type of system will necessitate a record of past network failures to determine what types of failures may occur in the future. Nevertheless, the absence of publicly accessible failure data is responsible for lacking a suitable general network failure identification and prediction system.

This research simulates network traffic to obtain failure data based on a generic failure guideline to address this issue. Furthermore, cutting-edge Artificial Intelligence (AI) techniques on that data yield promising results.