This is a company that’s been reported to use the dwell time of you mouse over a product as a potential indicator of interest. Something like a Citrix remote desktop is extremely chatty trying to keep the origin and server in sync with every move of a mouse or keystroke. If the ACKs from the origin confirming the receipt of screen change data took an abnormally long time it could show in system performance metrics pretty easily.
Probably a remote kvm system with QOS monitoring. Many secure systems won’t let you connect directly to sensitive resources from your personal workstation.
But the lag from the infiltrator to that is what was detected. There was presumably no Amazon software installed on the infiltrator’s computer so how can the lag be measured?
Keyboard input over kvm is pretty awful. It’s possible the kvm software was enforcing a delay between keystrokes to make sure they are delivered in order. Seeing keys consistently pressed with 500ms separation would be odd.
Perhaps something like time between key pressed and key released being abnormally high? Or erratic mouse movement?
I know whenever a PC I’m using is being remotely controlled, the mouse jerks around instead of moving smoothly around the screen. I’d imagine that gets even worse with ping/more layers of remote connections.
Amazon security experts took a closer look at the flagged ‘U.S. remote worker’ and determined that their remote laptop was being remotely controlled – causing the extra keystroke input lag.
With access to the final remote desktop, and access to the workers laptop you know the delay from these two so if there is more delay, then you can infer it’s coming from somewhere else? I’m sure there are more paths too but access to the North Koreans hardware doesn’t seem required
Also worth pointing out that this was a flagged employee (probably from something like data access logs) so they would be under more scrutiny and surveillance than the average employee
No. I’m talking about measuring the time in-between inputs being received over the remote connection. Purely observation from the receiver side of the connection.
Network overhead + dropped and re-sent packets, introducing unusual lag in between commands/keystrokes.
A key being pressed and key being released are two separate events that get transmitted separately and usually happen pretty close together. That gap getting larger, due to the long-distance connection introducing lag, could be what they were looking at.
How did Amazon know the lag?
This is a company that’s been reported to use the dwell time of you mouse over a product as a potential indicator of interest. Something like a Citrix remote desktop is extremely chatty trying to keep the origin and server in sync with every move of a mouse or keystroke. If the ACKs from the origin confirming the receipt of screen change data took an abnormally long time it could show in system performance metrics pretty easily.
That’s common internet marketing practice and the main reason to use an adblocker.
But didn’t they start to eye-track their drivers? Stuff like that.
Probably a remote kvm system with QOS monitoring. Many secure systems won’t let you connect directly to sensitive resources from your personal workstation.
But the infiltrator hacked the remote worker’s computer.
The actual worker may have been require to use…
https://aws.amazon.com/workspaces/desktop-as-a-service/
But the lag from the infiltrator to that is what was detected. There was presumably no Amazon software installed on the infiltrator’s computer so how can the lag be measured?
Keyboard input over kvm is pretty awful. It’s possible the kvm software was enforcing a delay between keystrokes to make sure they are delivered in order. Seeing keys consistently pressed with 500ms separation would be odd.
Perhaps something like time between key pressed and key released being abnormally high? Or erratic mouse movement?
I know whenever a PC I’m using is being remotely controlled, the mouse jerks around instead of moving smoothly around the screen. I’d imagine that gets even worse with ping/more layers of remote connections.
Yes but you need access to the culprit’s computer for the lag measurement.
With access to the final remote desktop, and access to the workers laptop you know the delay from these two so if there is more delay, then you can infer it’s coming from somewhere else? I’m sure there are more paths too but access to the North Koreans hardware doesn’t seem required
also, time between key presses on the compromised machine could indicate network lag to what is actually a Remote Desktop.
Also worth pointing out that this was a flagged employee (probably from something like data access logs) so they would be under more scrutiny and surveillance than the average employee
No. I’m talking about measuring the time in-between inputs being received over the remote connection. Purely observation from the receiver side of the connection.
Network overhead + dropped and re-sent packets, introducing unusual lag in between commands/keystrokes.
A key being pressed and key being released are two separate events that get transmitted separately and usually happen pretty close together. That gap getting larger, due to the long-distance connection introducing lag, could be what they were looking at.