AI

Fixing HTTPS Misconfigurations at Scale: An Experiment with Security Notifications

Abstract

HTTPS is vital to protecting the security and privacy of users on the Internet. As the cryptographic algorithms and standards underlying HTTPS evolve to meet emerging threats, website owners are responsible for updating and maintaining their HTTPS configurations. In practice, millions of hosts have misconfigured and insecure configurations. In addition to presenting security and privacy risks, misconfigurations can harm user experience on the web, when browsers show warnings for deprecated and outdated protocols.

We investigate whether sending direct notifications to the owners of misconfigured sites can motivate them to fix or improve HTTPS misconfigurations, such as outdated ciphersuites or certificates that will expire soon. We conducted a multivariate randomized controlled experiment testing multiple variations of message content through two different notification channels. We find that security notifications alone have a moderate impact on remediation outcomes, similar to or less than notifications for other types of security vulnerabilities. We discuss how notifications can be used in conjunction with other incentives and outreach campaigns, and identify future directions for improving the security of the HTTPS ecosystem.