Understanding emerging client-Side web vulnerabilities using dynamic program analysis

Marius Steffens

Abstract: Today's Web heavily relies on JavaScript as it is the main driving force behind the plethora of Web applications that we enjoy daily. The complexity and amount of this client-side code have been steadily increasing over the years. At the same time, new vulnerabilities keep being uncovered, for which we mostly rely on manual analysis of security experts. Unfortunately, such manual efforts do not scale to the problem space at hand. Therefore in this thesis, we present techniques capable of finding vulnerabilities automatically and at scale that originate from malicious inputs to postMessage handlers, polluted prototypes, and client-side storage mechanisms. Our results highlight that the investigated vulnerabilities are prevalent even among the most popular sites, showing the need for automated systems that help developers uncover them in a timely manner. Using the insights gained during our empirical studies, we provide recommendations for developers and browser vendors to tackle the underlying problems in the future. Furthermore, we show that security mechanisms designed to mitigate such and similar issues cannot currently be deployed by first-party applications due to their reliance on third-party functionality. This leaves developers in a no-win situation, in which either functionality can be preserved or security enforced.