KeenLab Blog

Yeah on macos the communication between your app and the window server (which is conveniently called WindowServer) happens via mach ports. Most of it is undocumented, in fact anything more "low-level" than using AppKit is undocumented, although IIRC it is in principle possible to use undocumented CG* apis to create and manipulate windows yourself without going through the appkit layers. I think each CG* api is basically a thin shim that communicates to the window server, which has a corresponding CGX* implementation which does the actual logic. This article has some details https://keenlab.tencent.com/en/2016/07/22/WindowServer-The-privilege-c…

Haven't read what they're doing, but chances are they are using an adversial neural network. The job of the adversarial network is to tell apart real from fake. The job of the neural network is to fool the adverserial network. Both are trained in tandem. One could imagine training another adverserial network that isn't used to train the network itself, and so will pick up on nuances that the original adversary doesn't pick up on. Anyone could do that, I don't think it's the author's responsibility. Somewhat related: https://keenlab.tencent.com/en/2019/03/29/Tencent-Keen-Security-Lab-Ex…

Given recently revealed vulnerabilities to the Tesla autopilot [1],for example, the adversarial markers that can cause the car to change lane, is it responsible for Tesla to go even further toward self driving without some way to mitigate such vulnerabilities first? [1] https://keenlab.tencent.com/en/2019/03/29/Tencent-Keen-Security-Lab-Ex…

you can link to a specific page in pdfs by adding #page=[PageNumber] to it, for example: https://keenlab.tencent.com/en/whitepapers/Experimental_Security_Resea… Or did you mean that there is not a specific page in the pdf which gives an overview?

The actual post https://keenlab.tencent.com/en/2019/03/29/Tencent-Keen-Security-Lab-Ex…

Why link to some dude's twitter, I WONDER, when he links the [actual paper](https://keenlab.tencent.com/en/whitepapers/Experimental_Security_Resea…

It doesn't have to be vt-d: https://keenlab.tencent.com/en/2018/04/23/A-bunch-of-Red-Pills-VMware-… I can confidently tell you that your understanding of security mitigations is flawed. And I say that based on experience not just a baseless opinion. Silver bullets in security don't exist. Let's every moveit instance was run in a container in a vm and in a dmz (actually moveit transfer is usually deployed in a dmz, isolated from everything else). But the entire purpose of the software is to contain all these important files and expose them to authorized parties, basically a file server (even has sftp!). The threat actors in this case didn't even bother compromising the OS, they just got a session id as a result of abusing sqli and .net deserialization flaw and logged into the webui and downloaded the files. At no point could a vm have stopped any of this. I said your undersranding is flawed because you mindset is solution centric not data centric. If all an attacker cares about is access to your gmail, a qubes VM with strict selinux rules is useless if they get you to click on a link that exploits firefox to steal your gmail cookies, defeating any yubikey 2fa you may have.