You’re muted — or are you? Videoconferencing apps may listen even when mic is off

Kassem Fawaz’s brot­her was on a vide­o­con­fe­rence with the microp­hone muted when he noti­ced that the microp­hone light was still on — indi­ca­ting, inex­pli­cably, that his microp­hone was being acces­sed.

Alar­med, he asked Fawaz, an expert in online privacy and an assis­tant profes­sor of elec­tri­cal and compu­ter engi­ne­e­ring at the Univer­sity of Wiscon­sin–­Ma­di­son, to look into the issue.

Fawaz and gradu­ate student Yucheng Yang inves­ti­ga­ted whet­her this “mic-off-light-on” pheno­me­non was more wides­pread. They tried out many diffe­rent vide­o­con­fe­ren­cing appli­ca­ti­ons on major opera­ting systems, inclu­ding iOS, Android, Windows and Mac, chec­king to see if the apps still acces­sed the microp­hone when it was muted.

“It turns out, in the vast majo­rity of cases, when you mute your­self, these apps do not give up access to the microp­hone, ” says Fawaz. “And that’s a problem. When you’re muted, people don’t expect these apps to collect data.”

After their initial testing, Fawaz and Yang, along with colle­a­gues from Loyola Univer­sity Chicago, conduc­ted a more formal inves­ti­ga­tion of just what happens when vide­o­con­fe­ren­cing soft­ware microp­ho­nes are muted. They will present their results at the Privacy Enhan­cing Tech­no­lo­gies Sympo­sium in July.

First, the team conduc­ted a user study, asking 223 vide­o­con­fe­ren­cing app users how they unders­tand the func­tion of the mute button and how they think it should handle audio data. While the parti­ci­pants were split about whet­her they thought the chat appli­ca­ti­ons were acces­sing their microp­ho­nes when muted, most beli­e­ved that the apps should not be able to collect data while set to mute.

For the second part of the study, the team inves­ti­ga­ted the actual beha­vior of the mute button on many popu­lar apps, deter­mi­ning what type of data is collec­ted and whet­her it could reveal perso­nal infor­ma­tion.

They used runtime binary analy­sis tools to trace raw audio in popu­lar vide­o­con­fe­ren­cing appli­ca­ti­ons as the audio trave­led from the app to the compu­ter audio driver and then to the network while the app was muted.

They found that all of the apps they tested occa­si­o­nally gather raw audio data while mute is acti­va­ted, with one popu­lar app gathe­ring infor­ma­tion and deli­ve­ring data to its server at the same rate regard­less of whet­her the microp­hone is muted or not.

The rese­ar­chers then deci­ded to see if they could use data collec­ted on mute from that app to infer the types of acti­vi­ties taking place in the back­ground. Using machine lear­ning algo­rithms, they trai­ned an acti­vity clas­si­fier using audio from YouTube videos repre­sen­ting six common back­ground acti­vi­ties, inclu­ding cooking and eating, playing music, typing and clea­ning. Applying the clas­si­fier to the type of tele­metry packets the app was sending, the team could iden­tify the back­ground acti­vity with an average of 82% accu­racy.

“When you’re cooking, the acous­tic signa­ture is diffe­rent from some­one who is driving or watching a video, ” says Fawaz. “So these types of acti­vi­ties can be distin­guis­hed just based on this acous­tic finger­print that was actu­ally sent out to the cloud.”

Whet­her or not the data is being acces­sed or used, the findings raise privacy concerns.

“With a camera, you can turn it off or even put your hand over it, and no matter what you do, no one can see you, ” says Fawaz. “I don’t think that exists for microp­ho­nes.”

Turning off a microp­hone is possi­ble in most device opera­ting systems, but it usually means navi­ga­ting through seve­ral menus. Instead, the team suggests the solu­tion might lie in deve­lo­ping easily acces­si­ble soft­ware “swit­ches” or even hard­ware swit­ches that allow users to manu­ally enable and disa­ble their microp­ho­nes.

Other authors include George K. Thiru­vat­hu­kal and Neil Klin­gens­mith of Loyola Univer­sity Chicago as well as Loyola gradu­ate student Jack West, who will join Fawaz’s lab the fall.

 

Photo: UW–Ma­di­son engi­ne­e­ring profes­sor Kassem Favaz (left) and gradu­ate student Yucheng Yang analy­zed the way popu­lar video confe­ren­cing apps collect data, finding that at least one appe­ars to collect all the audio from a user’s microp­hone even when “muted.” Contri­bu­ted photo