Personal Virtual Assistant (“DANI”)
A list of six main types of user privacy risks, which slightly confirm previous findings, but also find additional problems are compiled, and the limits users can take to protect themselves are data retention, processing and storage of data.
Abstract
-Smart voice assistants are gadgets that connects to the Internet, listen to their location and also respond to the instructions given by users to extract information online, manage resources, or notify the user with incoming messages, reminders, and the like. With their growing visibility in smart homes, their app seems to be limited only by the imagination of developers, who connect this off-the-shelf device to existing programs, Internet resources, or electronics. However, as their location is easy to use in their home, their location also raises concerns about user safety and privacy. In order to justify the trust placed on devices, devices must be protected from unauthorized access and the background infrastructure provided for speech and text analysis, command interpretation, and communication of other services and electronic devices must maintain data privacy. To investigate current risks, mitigation measures, and common assumptions in this emerging field, add recent research findings to the results of systematic reviews. We were able to compile a list of six main types of user privacy risks, which slightly confirm previous findings, but also find additional problems. We discuss these threats, their associated attack vectors, and the limits users can take to protect themselves are data retention, processing and storage of data, Side Channel Attacks, Attacks on Voice Models using Adversarial Samples.