Understanding Flaws in the Deployment and Implementation of Web Encryption

In recent years, the web has switched from using the unencrypted HTTP protocol to using encrypted communications. Primarily, this resulted in increasing deployment of TLS to mitigate information leakage over the network. This development has led many web service operators to mistakenly think that migrating from HTTP to HTTPS will magically protect them from information leakage without any additional effort on their end to guar- antee the desired security properties. In reality, despite the fact that there exists enough infrastructure in place and the protocols have been “tested” (by virtue of being in wide, but not ubiquitous, use for many years), deploying HTTPS is a highly challenging task due to the technical complexity of its underlying protocols (i.e., HTTP, TLS) as well as the complexity of the TLS certificate ecosystem and this of popular client applications such as web browsers. For example, we found that many websites still avoid ubiquitous encryption and force only critical functionality and sensitive data access over encrypted connections while allowing more innocuous functionality to be accessed over HTTP. In practice, this approach is prone to flaws that can expose sensitive information or functionality to third parties. Thus, it is crucial for developers to verify the correctness of their deployments and implementations. In this dissertation, in an effort to improve users’ privacy, we highlight semantic flaws in the implementations of both web servers and clients, caused by the improper deployment of web encryption protocols. First, we conduct an in-depth assessment of major websites and explore what functionality and information is exposed to attackers that have hijacked a user’s HTTP cookies. We identify a recurring pattern across websites with partially de- ployed HTTPS, namely, that service personalization inadvertently results in the exposure of private information. The separation of functionality across multiple cookies with different scopes and inter-dependencies further complicates matters, as imprecise access control renders restricted account functionality accessible to non-secure cookies. Our cookie hijacking study reveals a number of severe flaws; for example, attackers can obtain the user’s saved address and visited websites from e.g., Google, Bing, and Yahoo allow attackers to extract the contact list and send emails from the user’s account. To estimate the extent of the threat, we run measurements on a university public wireless network for a period of 30 days and detect over 282K accounts exposing the cookies required for our hijacking attacks. Next, we explore and study security mechanisms purposed to eliminate this problem by enforcing encryption such as HSTS and HTTPS Everywhere. We evaluate each mechanism in terms of its adoption and effectiveness. We find that all mechanisms suffer from implementation flaws or deployment issues and argue that, as long as servers continue to not support ubiquitous encryption across their entire domain, no mechanism can effectively protect users from cookie hijacking and information leakage. Finally, as the security guarantees of TLS (in turn HTTPS), are critically dependent on the correct validation of X.509 server certificates, we study hostname verification, a critical component in the certificate validation process. We develop HVLearn, a novel testing framework to verify the correctness of hostname verification implementations and use HVLearn to analyze a number of popular TLS libraries and applications. To this end, we found 8 unique violations of the RFC specifications. Several of these violations are critical and can render the affected implementations vulnerable to man-in-the-middle attacks.