Privacy analysis of voice user interfaces
Yeasmin, Farida (2020)
2020
Degree Programme in Information Technology, MSc (Tech)
Informaatioteknologian ja viestinnän tiedekunta - Faculty of Information Technology and Communication Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2020-04-08
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202003172721
https://urn.fi/URN:NBN:fi:tuni-202003172721
Tiivistelmä
A voice user interface (VUI) allows a user to interact with an application or system through voice or speech commands. A voice assistant device (VAD) primarily uses VUI to communicate with the user. The popularity of VADs is increasing due to voice is one of the most natural modes of human communication. The popularity of VADs and VUIs raises privacy concerns such as unwanted location tracking of a person using voice fingerprint data, unwanted recording of a private conversation, and unwanted habit detection of a person.
In this thesis, our research goal is to understand the emotional experiences and privacy expectations of users when they interact with VADs. We also want to identify suitable notification methods for human-to-VAD communication. In our research: (1) We categorize privacy in five types based on our background study, (2) We define four contexts based on location of the user and sight of view of the user to the VAD, (3) We analyze emotional experiences and privacy expectations of VADs based on a user study, and (4) We implement and evaluate a privacy-aware birthday pizza application that applies privacy-aware notification methods to inform users when collecting private data from conversations (e.g., birthdate).
Our evaluation shows that people have a strong preference for maintaining two types of privacy while interacting with VADs. These are location privacy and listening to private conversations. Our findings also revealed that users have several privacy expectations: 1) Consent for use of private data, 2) A feature to forget private data, 3) A function to turn off the device, and 4) Private data can be used only for positive purposes. We also find that users prefer visual and application-based notification when interacting with a VAD at home, while they prefer audio and application-based notification when at a classmate's house.
In this thesis, our research goal is to understand the emotional experiences and privacy expectations of users when they interact with VADs. We also want to identify suitable notification methods for human-to-VAD communication. In our research: (1) We categorize privacy in five types based on our background study, (2) We define four contexts based on location of the user and sight of view of the user to the VAD, (3) We analyze emotional experiences and privacy expectations of VADs based on a user study, and (4) We implement and evaluate a privacy-aware birthday pizza application that applies privacy-aware notification methods to inform users when collecting private data from conversations (e.g., birthdate).
Our evaluation shows that people have a strong preference for maintaining two types of privacy while interacting with VADs. These are location privacy and listening to private conversations. Our findings also revealed that users have several privacy expectations: 1) Consent for use of private data, 2) A feature to forget private data, 3) A function to turn off the device, and 4) Private data can be used only for positive purposes. We also find that users prefer visual and application-based notification when interacting with a VAD at home, while they prefer audio and application-based notification when at a classmate's house.