Our paper was accepted to CHI 2023. The paper was conducted during my Ph.D. at Tel Aviv University, advised by Eran Toch. More details you can find in our paper.
Video from CHI 2023.
Abstract
People regularly rely on social support from family, friends, and the public when mitigating security and privacy risks, even if mainstream technologies hardly support these interactions. In this paper, we evaluated Meerkat, a mobile application that allows users to receive support through screenshot capturing, marking, and messaging (see the image of the app below). In a field experiment (n = 65), we tested how Meerkat helps users face phishing attempts and examined it by receiving help from close social connections and community volunteers. Our findings show that while users could learn from both types of helpers, they were significantly more willing to rely on advice from close connections. We evaluate several criteria for successful support interactions, showing that learning is significantly correlated with specific properties of the support interaction, such as the length of the messages. We conclude the paper by discussing how our findings can be used to design community-based applications.
Meerkat
This paper presents a user study based on Meerkat, a mobile application for peer support in Android smartphones (see the image above). The app allows users to ask for and receive help when encountering challenging interactions with their mobile phones. Users can have multiple roles: seekers, users who receive technical support, and helpers who provide technical assistance. Meerkat allows seekers to receive support from contacts (e.g., family and friends) or other users that form a community of volunteers. When users want help, they can capture a screenshot and ask another user for explanations and advice by doodling and writing over the screenshot.
Video from the CHI2022 demo.
Research Questions
We investigate our research questions in the scope of combating mobile phishing attacks. We specifically focus on two types of social support architectures: ones that rely on existing social connections and those that use community member volunteers who are anonymous to the support seeker. Our definition of successful support interactions is based on the seeker and helper's perceptions of reliance, learning, privacy exposure, and user satisfaction. These perceptions can be different if the helper is a close connection or from a community of volunteers. Our research question is, what are the differences in perceptions of reliance, learning, and privacy exposure between close connections and a community of volunteers?
Experiment
The experiment was conducted in the field with participants installing the Meerkat app on their phones. During the five-day course of the study, participants received ten tasks, each representing a phishing attempt. The manipulated variable was the helper that answered the seeker’s questions, a social connection, or a community member (see below). The result is a within-subject design, where participants were randomized to different relationship types for each of the ten messages, with repeated measures for ten randomized phishing messages.
We recruited paired participants by posting on the university’s message boards. We stated in recruitment materials that we were looking for paired participants who live together or have strong social ties. Participants must be 18 years old and own an Android smartphone device. We verified eligibility through a short online questionnaire. We explained that the experiment required them to install an app from the Google Play store and perform ten tasks for up to five days. We offered them to select a date from four predefined dates to start the experiment. We recruited 65 participants (32 pairs) who persisted through the whole duration of the study.
Results
I present short results section about reliance, learning, and privacy exposure. The full analysis can be found in our paper.
Reliance
Seekers are willing to rely on social connections 8% more than volunteers.
Helpers feel their advice is 12% more valuable to close social connections than strangers.
Seekers’ Learning and Satisfaction
Seekers feel they learn 8% more from their close connections than volunteers.
Seekers expose less sensitive information by 17% to their close connections than volunteers.
The helpfulness of responses positively correlates with the length of the messages.
Helpers’ Perceptions
Helpers perceive their teaching abilities to be higher by 13% for close social connections than the volunteers.
Conclusion
People can rely on and learn from close connections and community volunteers.
Support from close connections has higher reliance and learning perceptions than a community of volunteers.
A community of volunteers increases seekers’ exposure to sensitive information perceptions more than close connections.
Some features may augment automatic support in the future.
Our full paper: https://doi.org/10.1145/3544548.3581183
Our app is available in Google Play.
