Metaphors considered harmful? An exploratory study of the effectiveness of functional metaphors for end-to-end encryption

Abstract

Abstract-Background: Research has shown that users do not use encryption and fail to understand the security properties which encryption provides. We hypothesise that one contributing factor to failed user understanding is poor explanations of security properties, as the technical descriptions used to explain encryption focus on structural mental models. Purpose: We methodically generate metaphors for end-to-end (E2E) encryption that cue functional models and develop and test the metaphors' effect on users' understanding of E2E-encryption. Data: Transcripts of 98 interviews with users of various E2E-encrypted messaging apps and 211 survey responses. Method: First, we code the user interviews and extract promising explanations. These user-provided explanations inform the creation of metaphors using a framework for generating metaphors adapted from literature. The generated metaphors and existing industry descriptions of E2E-encryption are analytically evaluated. Finally, we design and conduct a survey to test whether exposing users to these descriptions improves their understanding of the functionality provided by E2E-encrypted messaging apps. Results: While the analytical evaluation showed promising results, none of the descriptions tested in the survey improve understanding; descriptions frequently cue users in a way that undoes their previously correct understanding. Metaphors developed from user language are better than existing industry descriptions, in that ours cause less harm. Conclusion: Creating explanatory metaphors for encryption technologies is hard. Short statements that attempt to cue mental models do not improve participants' understanding. Better solutions should build on our methodology to test a variety of potential metaphors, to understand both the improvement and harm that metaphors may elicit.

Vision: A Noisy Picture or a Picker Wheel to Spin? Exploring Suitable Metaphors for Differentially Private Data Analyses

“…We investigated the media outlets 2 and companies applying differential privacy to see how differential privacy is conveyed to users. We found that differential privacy is conveyed to people using an example of tossing a coin for changing responses to sensitive questions, 3 noisy sound waves of radio channels, 4 and a noisy portrait 5 from the media outlets. Exploring how companies described differential privacy to their users did not result in any further metaphors which we could analyse and use in our study.…”

“…Two of the important steps in the framework proposed by Alty et al [1] are identifying system functionality and analysing metaphor-system pairings. To analyse the metaphors of E2E encryption, Demjaha et al [4] benefited from a metaphor evaluation matrix. The template of the metaphor evaluation matrix we use for analysing metaphor-system pairings is shown in Table 1 and is adapted from [4].…”

“…To analyse the metaphors of E2E encryption, Demjaha et al [4] benefited from a metaphor evaluation matrix. The template of the metaphor evaluation matrix we use for analysing metaphor-system pairings is shown in Table 1 and is adapted from [4]. As we have three different scenarios, we adapted the template table to evaluate whether a metaphor is suitable for each of our scenarios.…”

“…To address our research question, similar to Demjaha et al [4] who benefited from the framework proposed by Alty et al [1] to generate and evaluate their explanatory metaphors for E2E encryption, we employed and adapted the framework to fit our objective. Our approach consists of three phases: 1) metaphor generation, 2) metaphor analytical evaluations based on expert analyses, and 3) metaphor empirical evaluations involving lay users.…”

Making Encryption Feel Secure: Investigating how Descriptions of Encryption Impact Perceived Security

“…The authors suggest to implement a user interface that makes E2E encryption processes more graspable for the user and increases transparency about the business model and the encryption protocol, which is not publicly available yet. The creation of metaphors with the objective of improving user understanding of encryption also seems to be a promising direction for future research, however, Demjaha et al, [10] showed that using metaphors can sometimes do more harm than good, and the authors underline the difficulties of explaining encryption to users. Similar problems are pointed out by Abu-Salma and colleagues [1], who analyzed the user interface of the secure messaging app Telegram.…”