As people rely more heavily on online communication, privacy becomes an increasingly critical concern. Users of communication services (e.g., email and messaging) risk breaches of confidentiality due to attacks on the service from outsiders or rogue employees, or even government subpoenas and network surveillance. End-to-end encryption, in which anyone cannot read the user's content, is the only way to fully protect their online communications from malicious attackers, rogue company employees, and government surveillance. Although in recent years we have witnessed considerable efforts to push end-to-end encryption into broader adoption, and indeed several popular messaging tools have adopted end-to-end encryption, some obstacles still remain which hinder general users from proactively and confidently adopting end-to-end encrypted communication tools and acknowledge their security benefits.

In this dissertation, we investigated the adoption of end-to-end encrypted communication from a variety of user-centered perspectives. In the first part, we conducted a lab study (n=52), evaluating how general users understand the balance between the usability and security for different key management models in end-to-end encryption. We found that participants understood the models well and made coherent assessments about when different tradeoffs might be appropriate. Our participants recognized that the less-convenient exchange model was more secure overall, but found the security of the key-directory based model to be "good enough" for many everyday purposes.

In the second part, we explored how general users value the usability and security tradeoffs for different approaches of searching over end-to-end encrypted messages. After systematizing these tradeoffs to identify key feature differences, we used these differences as a basis for a choice-based conjoint analysis experiment (n=160). We found that users indicated high relative importance for increasing privacy and minimizing local storage requirements. While privacy was more important overall, after the initial improvement was made, further improvement was considered less valuable. Also, local storage requirement was more important than adding marginal privacy.

Since significant research indicated that non-expert users' mental models about end-to-end encryption led them to make mistakes when using these tools, in the third part of this dissertation, we took the first step to tackle this problem by providing high-level, roughly correct information about end-to-end encryption to non-expert users. In a lab study, participants (n=25) were shown one of several variations on a short tutorial. Participants were asked about their understanding of end-to-end encryption before and after the tutorial, as well as which information they found most useful and surprising. Overall, participants effectively learned many benefits and limitations of end-to-end encryption; however, some concerns and misconceptions still remained, and our participants even developed new ones. The results provided insight into how to structure new educational materials for end-to-end encryption.