It's not clear to me from this thread whether those messages only become public if one of the parties to the messages then download the chat, or even when a third party downloads the chat. One seems more serious than the other.
> IMPORTANT: I can only find evidence that Zoom by default creates a log of private messages WITH THE HOST. This happens regardless of whether the session is recorded. It isn't a good design decision from a privacy perspective. But it's not the same as recording all* messages.*
https://twitter.com/rcalo/status/1244404664260411392
It's not super clear what a "host" is in Zoom parlance. Is it the meeting creator, or do they mean the host as in the person you were chatting with?
Hosts are privileged to have control over the meeting, e.g. can toggle recording or shut it down (so everyone gets disconnected and sees "was ended by host" or something like that), etc.
[1] https://support.zoom.us/hc/en-us/articles/115004792763-Savin...
Yes, you want to account for people's actual behavior. This isn't going to rise above the level of "minor" if viewed from a security perspective, because it's a self-only attack -- nobody gets any powers they didn't already have, and Alice is hurting herself, not someone else.
(She might inadvertently hurt Carl, if Carl was sending her messages making fun of Bob, but she was allowed to do that anyway.)
A usability or operational perspective might object to the behavior here more strongly.
It's not about hurting people's feelings, it's about information leakage. And to cut off anyone that says passwords shouldn't be shared in a private chat, that's irrelevant. Good infosec security practices in one place do not preclude criticism of bad practices elsewhere. Security is about layers or protection, so any layer with problems should be noted. If that layer happens to be a third party application that mixes private and public channels in some instances and if there isn't warning as to this happening, it deserves to be called out.
Another way to look at it is that any minor information leakage can have a major impact if the information leaked is very important.
It's not irrelevant. There are phone apps with no other purpose than to publicize your location. If you should happen to be a fugitive, using such an app would be a bad move. Does that make the privacy leak in the app a security problem? No, how could it? If you don't want your location publicized, the answer isn't to remove the only feature from a location-sharing app so you can run NOPs in peace. It's to stop using the app.
Your misuse of a feature that performs exactly as advertised can't justify calling that feature a security problem. The people responsible for the feature don't know how you're using it. The use pattern is the security problem, and it needs to be addressed by people who (1) know what it is, and/or (2) are responsible for it. Zoom fulfills neither criterion.
That's clearly not the case here. It's irrelevant in that in the context of this specific discussion, which this application, which is marketed towards enterprises as secure, an argument that a sharing of private information accidentally though bad UI is the problem of the user for putting private information in that private channel is irrelevant.
> Your misuse of a feature that performs exactly as advertised can't justify calling that feature a security problem.
That depends on how we classify "exactly as advertised". If the overall claims of the product as to being secure are easily and often circumvented on accident through poor UI, then that may be a security or privacy problem. If that place it's advertised is not something that most users will encounter during normal usage, then how it's advertised is of little consequence.
One extreme end of this would be something hidden deep in the EULA or privacy policy that advertised how this works, and the other extreme end would be an alert every time you use it that explains this. I think one is obviously a problem (to the point that it looks purposeful), and the other obviously isn't, but the only difference between them is where the information is placed or how assured you can be that the user has encountered and hopefully understood it. I think this clearly indicates that the problem is not whether certain behavior is advertised, but how aware users are made of how the application functions.
I was under the mistaken impression that this would include one-on-one chats between the _non-host_ members of the meeting.
According to my reading, she will by default include everything she typed and any messages sent to her. That's a horrible default if true.
A fair number of use cases benefit from being able to see private chats - webinar replays it's nice to see the questions folks are asking, but usually host in a large webinar doesn't want to chat spam everyone by allowing 300 people to post to everyone. So they can set it host only, download, then edit and include to whomever cares what questions are getting asked.