Ad Astra is almost entirely funded

[zakiyatasnim]

Lea Schönherr, Maximilian Golla, Jan Wiele, Thorsten Eisenhofer, Dorothea Kolossa and Thorsten Holz from the Ruhr University Bochum and the Max Planck Institute for Security and Privacy have released a research paper titled “This is unacceptable. Where is my privacy?” Here are their conclusions:

"We were able to identify over a thousand sequences that falsely trigger smart speakers. Examples are presented in the video . In our study, we analyze a variety of audio sources, study the influence of gender and different languages, and measure the reproducibility of the detected triggers. We also describe a method for artificially creating such expressions. Reverse decoding the Amazon Echo communication channel allows us to gain insight into how commercial companies work with these triggers in practice. In addition, we analyze the impact of random triggers on privacy and discuss mechanisms that can protect the personal data of smart speaker users."

The researchers analyzed voice assistants from Amazon, Apple, Google, Microsoft, and Deutsche Telekom, as well as three Chinese models — Xiaomi, Baidu, and Tencent. A recent publication examined austria number data devices from the first four manufacturers. Representatives from Apple, Google, and Microsoft did not immediately comment. Amazon later issued the following statement:

“Unfortunately, we have not yet been given access to the methodology of this study, so we cannot confirm the judgments expressed in it. For our part, we can assure you that privacy is an important part of the Alexa service, and our devices are activated only after pronouncing the appropriate word. Users speak to Alexa billions of times a month, and false positives are extremely rare. Our speech recognition systems improve every day as customers use these devices. We continue to improve these technologies and encourage researchers to disclose their research methodology to us so that we can answer in more detail.”

The researchers declined to provide a copy of the full paper before publication. However, the data presented suggests that voice assistants can be privacy-invading without users noticing. Those concerned about this issue are advised to use digital assistants only when absolutely necessary and to turn them off at other times. Alternatively, they can opt out of these devices altogether.

MORE INTERESTING:

Understanding difficulties: why smart voice assistants have such a hard time learning new languages
The dark side of voice assistants
Listen to the voice of genderless artificial intelligence