Researchers Drop a Public Exploit That Roots AnyDesk Linux Before Anyone Approves the Call

Remote desktop tools are built around one comforting moment: someone has to click accept. A newly published exploit for AnyDesk on Linux skips that moment entirely. Researchers have released a full working attack, dubbed AnyPwn, that gives an attacker root access before anyone approves the incoming connection.

The flaw affects AnyDesk for Linux version 8.0.2 on x86_64 machines, and the public proof of concept targets that exact build. Windows and macOS versions are not affected.

How the attack works

The bug sits in the code that handles a particular type of packet in AnyDesk's session protocol, known as mode 5 stream packets. The software never properly checks the payload length it is given.

Feed it a value like 0xFFFFFFF0, add the 0x10 the code tacks on, and the number wraps around to something tiny. AnyDesk allocates a minuscule buffer, then writes far more data into it than it can hold. From that heap overflow, the exploit chains together existing code fragments to run a command as root.

The attack works over a direct TCP connection on port 7070. It is not a sure thing every time. The exploit is probabilistic, and if the memory on the heap is not laid out the way the attacker expects, the AnyDesk service crashes instead of handing over control.

The researchers also say the vulnerable code can be reached through AnyDesk's relay servers, which they triggered using the Frida instrumentation tool. They did not demonstrate a full exploit over relays, though.

A quiet patch with no warning

AnyDesk actually fixed this months ago. Version 8.0.3, released in June 2026, closed the hole. The changelog described it only as having "fixed a bug that could lead to a crash." There was no CVE, no formal security advisory and nothing telling Linux admins they had a root level problem on their hands.

The researchers announced the issue on June 22, and AnyDesk acknowledged it the following day. The company said the vulnerability was "limited to direct connections on Linux (connections that do not go through our relays)." As of the October 9 coverage, no CVE had been assigned.

That framing matters. A "crash fix" is the kind of update that sits in a queue for months on servers and workstations nobody thinks about. A pre-authentication root exploit is the kind that gets patched the same afternoon. After the researchers posted a video of the proof of concept, AnyDesk appears to have pulled the 8.0.2 download.

The discovery came from Rick de Jager of the V12 security team, who found it using V12's own code review engine.

What admins should do

Any Linux machine still running AnyDesk 8.0.2 should be updated to 8.0.3 or later. The latest version noted in coverage is 8.1.0. Where an update is not possible right away, restricting access to TCP port 7070 cuts off the direct connection path the public exploit uses.

Remote access software is a prized foothold because it already has deep control of the machine it runs on. The same logic drives attacks on devices people trust implicitly, as the Ledger hardware implant that drained $86 million showed. Once the trusted layer is compromised, every safeguard above it stops meaning much.

AnyDesk has had a rough run on security. A separate overflow, CVE-2025-27918, was fixed in version 7.0.0 in April 2025, though that one involved processing user images and is unrelated. The company also suffered a production breach in 2024 that forced it to revoke certificates. This time, the fix existed for months. What was missing was anyone saying how serious it was.

Some offers on this page may be paid placements or contain affiliate links.