• We've moved. Our community is now at ownd.sh — latest discussions and updates live there.

KERNEL ROOTKIT DEVELOPMENT IN 2026 - WHY EVERYTHING YOU'VE READ IS OBSOLETE

Technotron

Elite member
Elite member
Technotron Rep
0
0
Rep
31
Technotron Vouches
2
0
0
Vouches
2
Posts
920
Likes
86
Bits
11 MONTHS
11 11 MONTHS OF SERVICE
Leave a rep and a like
handsome.png


 

Hidden Content

Most rootkit content floating around was written in the 2010s era and it's genuinely useless now. If you've read tutorials about SSDT hooking or patching the syscall table, forget all of it. On any modern Windows 11 machine that stuff gets you a BSOD within minutes because of PatchGuard. It watches the SSDT, the IDT, the GDT, the LSTAR MSR — basically everything the old tutorials taught you to touch. And HVCI makes kernel memory non-executable anyway, so even if you got a driver loaded, your code can't run.

So how does this actually work in real intrusions these days? Nobody writes drivers anymore, they steal them. The technique is called BYOVD — bring your own vulnerable driver. There are hundreds of legitimately signed drivers floating around that have known arbitrary read/write bugs. There's a whole database of them at loldrivers.io. The Capcom.sys driver is the classic example everyone learns on — signed driver with a function that lets user mode read and write kernel memory directly. Microsoft knows about all of this of course, which is why they shipped the vulnerable driver blocklist. But here's the thing, the blocklist only works when HVCI is on, so the first move of any serious implant is getting HVCI turned off, and that itself is something a decent EDR should be screaming about. Most don't.

Once you've got kernel read/write, the modern play isn't hooking anything. It's callback abuse. Windows has all these notification APIs — process creation callbacks, thread callbacks, object callbacks, registry callbacks. A malicious driver just registers them like any AV product would, and then uses them to do the dirty work. The spicy part is you can use the same APIs to REMOVE the callbacks that the EDR registered. That's the "EDR blinding" stuff that MDSec published research on. No hooks, no patches, nothing for PatchGuard to complain about. Just the driver politely unregistering your protection.

File hiding works the same way now — instead of hooking NtQueryDirectoryFile like the old days, you just register a minifilter with the filter manager and quietly return STATUS_NO_SUCH_FILE when someone asks about your files. You're a legitimate participant in the filesystem stack. Nothing to detect from a code perspective.

Process hiding via DKOM still works, interestingly. Unlinking an EPROCESS from ActiveProcessLinks isn't covered by PatchGuard so task manager never sees you. But ETW still fires the process creation event, so hiding from the process list alone is pointless against anything modern — which is why the real ones pair it with ETW tampering.

And above all of this sits the hypervisor layer. If you can get a thin hypervisor installed (which BYOVD makes possible), the actual Windows kernel becomes your guest and you can snoop on EDR processes, hide pages, redirect interrupts. Joanna Rutkowska's Blue Pill concept from 2006 was exactly this, way ahead of its time. Detection is basically timing analysis — RDTSC deltas — and it's noisy enough that you can't trust it.

For the defenders reading this, the useful part: baseline your kernel callbacks. There are open source tools that dump all registered process/thread/image/object callbacks. If your AV's callbacks suddenly disappear from that list, that's your intrusion. Memory forensics with Volatility 3 still rips most of this apart too — the plugins for callbacks and SSDT inspection will show you everything that's out of place. The whole game for a modern implant isn't surviving forensics, it's preventing the memory dump from ever happening.

If you want to actually learn this stuff properly, Pavel Yosifovich's Windows Kernel Programming book is the real entry point, and build everything in a VM with testsigning enabled. First rule of kernel dev — your first ten drivers will BSOD the machine. Don't do it on the machine you care about.

 



 



@Techn0tron
 

Attachments

  • handsome.png
    handsome.png
    514 bytes · Views: 1
Technotron's signature

uptain2542

Member
uptain2542 Rep
0
0
Rep
0
uptain2542 Vouches
0
0
0
Vouches
0
Posts
42
Likes
2
Bits
1 MONTH
1 1 MONTH OF SERVICE

Attachments

  • handsome.png
    handsome.png
    514 bytes · Views: 1
  • kek.png
    kek.png
    696 bytes · Views: 1
  • sheepe.gif
    sheepe.gif
    4.5 KB · Views: 1

zyn1282

Well-known member
zyn1282 Rep
0
0
Rep
0
zyn1282 Vouches
0
0
0
Vouches
0
Posts
103
Likes
0
Bits
2 YEARS
2 YEARS OF SERVICE

QwertyNeo

Member
QwertyNeo Rep
0
0
Rep
0
QwertyNeo Vouches
0
0
0
Vouches
0
Posts
13
Likes
0
Bits
2 YEARS
2 YEARS OF SERVICE
thank you for showing this guide thanks a lot
 

179,168

126,023

1,405,739

Top