Analysis of Certificate Transparency using Probabilistic Verification

点击翻译按钮获取中文摘要

Public key certificates are vital for internet security as they help users know which websites have been properly authenticated. However, there have been reports of certificates being issued incorrectly by supposedly trusted third parties. This led to them being exploited in man-in-the-middle attacks with devastating results. To improve the auditing of issued certificates, one idea proposed is to have them be submitted to a public, append-only log that anybody can view. This is part of an open framework called certificate transparency. However, a problem with this system is that a subset of users can be made to exclusively contact a `fake' log that contains fraudulent certificates. To prevent this, clients across the network could be made to gossip with each other and verify if they are seeing the same version of the log. Researchers such as Chuat et al have designed gossip protocols which disseminate information whenever misbehaviour from the log is noticed. At present it is not known how effective these protocols are in detecting attacks. For this research project, we would like to apply probabilistic model checking to analyse how these protocols behave and look for potential weaknesses. We also seek to propose any recommendations to improve the performance and detectability of these protocols. Finally, we will apply our methods to other systems that use gossip protocols to help users verify consistency for other transparency logs.