Thank you, very cool. Maybe you should go ahead and reread the post you replied to.
I've read it the first time, dude. And you're still wrong: ACLs don't work at "low level" - they operate several layers above what is truly "low level" in a mass-storage stack. As proof, ACLs aren't even a real thing if you don't use NTFS, for example. And it's exactly the same for Unix, obviously.
Everything related to isolation between users (including ACLs) is done,
in fine, through a security token, because it would cost way too much time and CPU to check credentials each time. Obviously, the "normal way" to get one is to use credentials; therefore you'll need a login and password. But that's the "normal way", and, as I've already said it numerous times, you're definitively thinking too much inside the box to speak about security. I'll tell it another time: YOU MUST BE PARANOID TO CHECK SECURITY FEATURES. The problem lies not in the anticipated and expected threats, but rather in the unforeseen and/or unknown ones. Forget any already anticipated threat or problem, as it is precisely already foreseen. One must think outside the box, or for events classified as "improbable" or even "impossible".
But once you get a token, the process can use it to impersonate ANY access allowed by this token. Little by little, you may then obtain an admin's token, then a domain admin's one, and then, you're the king of the network. Because the "human factor" is such that most people do NOT reboot their computer once an admin came to unlock something they needed - whatever it was. Including through remote control - and there isn't ANY reliable way to do it, whatever protocol / software you can use. If you can get remote access to a computer, someone else can TOO. Full stop.
The "drama" here is that most computers aren't rebooted often, in particular servers, therefore the granted security tokens aren't flushed away. And most people (and even admins...) set their validity time to a way too long period: near 2 months isn't uncommon (2^32 milliseconds, in fact, so roughly 50 days).
To get back to thinking out of the box: that's why we use "dead man's switch" type of securities, for instance. Because death from a heart attack or an aneurysm is NOT a risk normally anticipated for a security position, since there are always TWO guards together... But both can die at the same time, either due to misfortune, or because of an attack! So, if the guards die suddenly, the alarm triggers a few minutes later, at worst. For the same reason, a train always defaults to applying brakes, and a constant action is required to allow its wheels to move: if the train driver dies, the train stops automatically a few kilometers further.
Applied to computer security, a dead man switch means something like UAC or start/stop validities - you ask for a security token, and you explicitely release it once you used it, with a default timeout set to a very short period, like 30 seconds at most. Because EVERYTHING done by user can be replayed by a process injecting messages/triggers inside other processes, excepted when something like the "Secure Desktop" used by UAC is triggered... And even that can be hacked, theoretically, if you managed to get inside kernel! It should rely on another CPU running a ROMmed software to be "perfect", this CPU being used to enable/disable physical address and data lines on the main CPU. Unfortunately, this cannot be done on a modular computer but exclusively on locked machines where access to peripherals is fixed by construction.
So, you CAN break into ACLs even on a live OS. And as I already proved, even your grandma can do it on an offline OS. OK, the trainee secretary can't. But you do realize that several people, including myself, are "a bit" above this level of competence regarding computers security?
Just for you to know: common OS are rated CC-EAL4. There are THREE levels above this one (and three below). You probably never encountered anything better than CC-EAL4, while in my domain, it's the MINIMUM rating, the one used for... non-critical display panels. We're not playing in the same league.
Very much a TL;DR but no, users cannot access data from other users in Windows via the terminal while the OS is in use and ACLs are active. This includes both files and directories. VeraCrypt protects data at rest.
Seriously, you're a complete joke, you know?
So, tell me: how the fuck can a ransomware infect and spread across a network, in this case? I'm using this example because, nowadays, it's the most common type of hack encountered...
And, for FLOSS software: who is responsible for their review and analysis? What are their qualifications? What is the deadline for submitting the analysis report? Is the code reviewed 100%? Is static analysis or, better yet, dynamic analysis used? Who takes responsibility for ensuring the safety of such software, and is there any guarantee? Code that is not reviewed by anyone, except malicious hackers, is by no means more reliable than proprietary code... I have seen more than enough bugs in "free" software to not have an exaggerated trust in them, especially when you can struggle to get any guarantee - be it in terms of corrections or diagnostic assistance!
One always gets what they pay for. If you pay nothing, draw your own conclusions, knowing that the engineers who worked on it, on the other hand, need to eat and have a roof over their heads.
