| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 1.36 MB | Adobe PDF |
Orientador(es)
Resumo(s)
Open-source software (OSS) ecosystems such as CRAN, PyPI, and npm are vital to modern
software development, providing developers with vast collections of reusable software
packages. However, the interconnectedness of these packages creates complex
interdependencies that can expose these ecosystems to systemic risks. In this work, we model
such ecosystems as directed package dependency networks and apply complex network
analysis techniques to study their robustness and vulnerability to cascading failures. We begin
by assessing how random failures and degree-based targeted attacks impact the connectivity
of the networks, revealing a high susceptibility to deliberate in-degree-based attacks, with
npm being particularly fragile. In contrast, the networks exhibit greater tolerance to random
failures and out-degree-based attacks. Then, we implement a threshold-based failure
propagation model to capture the cascading dynamicsresulting from a single package’s failure
and assess how such failures can spread through the ecosystem. The results for CRAN show
that while most cascades remain localized, a few critical packages can trigger widespread
failures. These findings reinforce the need to address the risks that result from the structural
properties of the studied repositories.
Descrição
Dissertation presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics, specialization in Data Science
Palavras-chave
Software Ecosystems Package Dependency Networks Complex Network Analysis SDG 9 - Industry, innovation and infrastructure
