Resumen
Intelligent voice assistants are internet-connected devices, which listen to their environment and react to spoken user commands in order to retrieve information from the internet, control appliances in the household, or notify the user of incoming messages, reminders, and the like. With their increasing ubiquity in smart homes, their application seems only limited by the imagination of developers, who connect these off-the-shelf devices to existing apps, online services, or appliances. However, since their inherent nature is to observe the user in their home, their ubiquity also raises concern of security and user privacy. To justify the trust placed into the devices, the devices must be secure from unauthorized access and the back-end infrastructure tasked with speech-to-text analysis, command interpretation, and connection to other services and appliances must maintain confidentiality of data. To investigate existing possible vulnerabilities, approaches to mitigate them, as well as general considerations in this emerging field, we supplement the findings of a recent study with results from a systematic literature review. We were able to compile a list of six main types of user privacy vulnerabilities, partially confirming previous findings, but also finding additional issues. We discuss these vulnerabilities, their associated attack vectors, and possible mitigations users can take to protect themselves.