Impactos da pandemia da COVID-19 nas conexões das redes de computadores e na cibersegurança

Abstract

Considering the restrictions imposed due to the COVID-19 pandemic, in particular those related to people’s mobility, this work has as one of its purposes to investigate the impacts of these restrictions on computer network connections and on the volume of traffic. In order to explore this topic, two opinion polls were conducted, whose main objective was to evaluate the user’s perception of the stability of their Internet connection. However, as the user’s perception may not necessarily reflect reality, an analysis of the change in traffic patterns of the Points of Presence (PoPs) of the Rede Nacional de Ensino e Pesquisa (RNP) and a monitoring of the traffic of wireless networks and mobile device connections in Radio Base Stations (RBS) for an 18-month period were performed, comparing them with information about community mobility. It was observed that Internet traffic grew. On other networks, such as those of research institutions, the volume of traffic has decreased. Only in Brazil, the IX.br (Brazil Internet Exchange), the infrastructure for interconnection between Autonomous Systems (ASs), registered an increase in traffic of around 60% between December 2019 and December 2021. In addition to bandwidth consumption, there was also a change in traffic type. For example, in the RNP’s PoPs, there was an increase of up to five times in bandwidth consumption for videoconferencing applications, although the total bandwidth consumption decreased. Upon observing significant variations in both type and volume of traffic, the possibility of impacts on cybersecurity, which is a source of concern for most organizations due to the growth of threats, was raised. In order to support risk mitigation teams with the maximum amount of information as early as possible, this work explores several methodologies based on machine learning to classify the severity of vulnerabilities and the time until they are corrected. To explore such methodologies, the textual descriptions and/or metrics of the Common Vulnerability Scoring System (CVSS) vector were used. The influence of class imbalance and software type on vulnerability classification was also analyzed. To that end, this work evaluates the efficiency of these methodologies using a dataset that comprises vulnerabilities of several different applications. However, due to the particular relevance of communication applications during the COVID-19 pandemic, this work also evaluates the difficulty of performing the predictions mentioned for this specific group of applications.