7 ms·
I appreciate the comment but I feel it fails to address mine. > is also not going to work because if they don't like it there are several other ways they can c
by godelski 3y ago
I appreciate the comment but I feel it fails to address mine.
> is also not going to work because if they don't like it there are several other ways they can copy it
>> All arguments against privacy were dismissed with claims that one can screenshot
> you have no guarantees
>> stochastically increases privacy and security
I'm not sure why we have to frame privacy and security with strict guarantees. If we need strict guarantees then we should abandon all efforts because guarantees do not exist (my preemptive response to what you are thinking is "implementation.") Fundamentally it is always stochastic as statistics is a way to capture error and uncertainty.
Everyone that advocates for deletion is well aware that one can screenshot, copy, or even write down information. It is a weird assumption to make, because it implies an exceptional level of stupidity to the person you respond to. Every 12 year old knows that you can screenshot Snapchats and they know it can be done without Snapchat warning the other person. So forgive me, because while I know you are acting in good intent (we've had enough conversations that I have that respect for you), I think I need to point out that it is easy to read such a response as indicating you did not bother to read my comment (so why comment?) or that you imply I am incredibly naive. I say this because this topic is often heated so stating this can help reduce the inference gap. I know you are not trying to do such a thing, but not all others will have that shared history to give benefit of the doubt.
There's nuance necessary beyond the existence of copy methods. People with positions similar to mine understand that the act of copying requires time and energy. That someone needs to either preemptively implement a system of record keeping or that such an action is responsive. In the latter case, having the ability to delete acts as a windowing operation. You do not know your adversaries a priori nor do all adversaries begin as adversaries. So if you can delete the information before a log is created, you have succeeded. Yes, this is stochastic. But I'd rather have a 1% of protection than a 0%, because an immutable history just means the adversary has unlimited time to strike. Basically, you are doing your adversary's job for them. tldr: the game has a temporal component and it is not turn based.
I will understand arguments about communication, of how some may assume stronger protection than received, but I'd also respond that this is a fairly universal claim and we do not apply it to many other domains as we still find utility.
I'll also add that this stochastic protection is why companies will remotely wipe your devices if they are reported lost or stolen. You wipe for protection but operate under assumption that the data was copied. This is fairly standard practice.
- AnthonyMouse 3y ago> It is a weird assumption to make, because it implies an exceptional level of stupidity to the person you respond to. Part of this is that the rest of the argument is often implied and then you say something expecting the other person to fill in the rest without actually writing the words. > stochastically increases privacy and security The question is, whose security? If you can delete the message from the recipient's device then they can't use it against you, or otherwise use it to their advantage even if their use doesn't negatively affect you (e.g. so they can refresh their recollection of what was said or confirm your assertion that you did say that). Which means it's a trade off, but also means that the other user often has the incentive to take action to defeat this, because their device is acting against their interests in favor of yours. This is how you get forks of the client, which is one of the things Signal doesn't like. > If we need strict guarantees then we should abandon all efforts because guarantees do not exist (my preemptive response to what you are thinking is "implementation.") Fundamentally it is always stochastic as statistics is a way to capture error and uncertainty. Yes and no. Everything is always stochastic because there is a 1 in 2^256 chance they could guess your AES key by random chance on the first try, but even if that's true in theory it need not be in practice. In many cases you can regard something as having a strong guarantee when the chances are low enough to be disregarded, even if they're not low enough to be literally zero in a theoretical sense. Whereas the ability of someone to screenshot a message you don't want them to be able to retain is within the capability of ordinary users, so the amount of practical protection it provides is minor and is only worth doing if it isn't a trade off against something non-trivial. And then you have the trade offs against the desire of the recipient to not have their messages deleted against their will and the desire of the foundation to not give those users an incentive to fork the client to reduce their inconvenience in retaining them. There is also the potential to mislead a naive user into relying on the feature not contemplating how easily it could be bypassed and then they give someone a permanent copy of something they only intended to provide access to temporarily. What you could do is make support for the feature optional. The sender requests deletion and then the recipient can configure their client to respect it or not. This was always within the recipient's power, the sender couldn't have enforced it anyway, but now the message can be automatically deleted if the sender and the recipient both want it to be, and no one is given a false understanding of how the system works or an incentive to try to defeat it because they're in possession of a device which is acting against their wishes.