A sophisticated Linux malware called Koske, discovered in July 2025, hides malicious code within innocent-looking panda bear JPEG images to deploy cryptocurrency miners and establish persistent system access[^1]. Security researchers at AquaSec believe Koske was developed using artificial intelligence, based on its adaptive behaviors and code structure[^2].
The malware exploits misconfigured JupyterLab instances to gain initial access, then downloads two panda images containing separate payloads - a C-based rootkit and a shell script[^3]. Rather than using steganography, Koske employs polyglot files that function as both valid images and executable scripts[^1].
Once executed, the malware:
- Deploys CPU and GPU-optimized miners for 18 different cryptocurrencies
- Establishes persistence through cron jobs and systemd services
- Uses LD_PRELOAD to hide malicious processes and files
- Manipulates DNS settings and network configurations
- Automatically switches mining pools if one becomes unavailable[^1]
“Impersonation and psychological warfare will be a big thing in the coming years,” warns Rem Dudas from Palo Alto Networks, noting how AI enables malware to mimic other threat actors’ techniques[^4].
[^1]: BleepingComputer - New Koske Linux malware hides in cute panda images
[^2]: The420 - How Is A “Panda” Becoming a Persistent Threat?
[^3]: Securitricks - AI-Generated Malware in Panda Image Hides Persistent Linux Threat
[^4]: BetaNews - Hackers are using AI and panda images to infect Linux machines
…and this is where sanitizing inputs becomes even more important…
baod_rate@programming.dev 1 day ago
Is it just me or is this not a very convincing rationale.
AmbitiousProcess@piefed.social 1 day ago
Not whatsoever.
Practically any mining software would allow you to change a pool whenever you felt like it, and making a script that just goes "oh, x.x.x.x isn't responding anymore, I should point my hashrate to y.y.y.y now" is... not hard, to say the least.