Security of open source and closed source software: An empirical comparison of published vulnerabilities

摘要

Reviewing literature on open source and closed source security reveals that the discussion is often determined by biased attitudes toward one of these development styles. The discussion specifically lacks appropriate metrics, methodology and hard data. This paper contributes to solving this problem by analyzing and comparing published vulnerabilities of eight open source software and nine closed source software packages, all of which are widely deployed. It provides an extensive empirical analysis of vulnerabilities in terms of the mean time between vulnerability disclosures, the development of disclosure over time, and the severity of vulnerabilities, and allows for validating models provided in the literature. The investigation reveals that (a) the mean time between vulnerability disclosures was lower for open source software in half of the cases, while the other cases showed no differences, (b) 14 out of 17 software packages showed a significant linear or piecewise linear correlation between the time and the number of published vulnerabilities, and (c) no significant differences in the severity of vulnerabilities were found between open source and closed source software.