Today's cybersecurity headlines are brought to you by ThreatPerspective


Ethical Hacking News

SANSAR Linux Cryptomining Malware: A Threat to GPU-Enabled AI and LLM Workloads




A new Linux-based cryptocurrency mining operation has been identified, utilizing a toolkit called SANSAR to tamper with GPU readings used to detect high GPU workloads. The threat actor deploys a kernel rootkit and a backdoor to establish persistent access to the compromised system, and the operation has been linked to a cluster of financially motivated threat actors tracked as REF3364. The attack chain documents by Elastic Security Labs reveal a sophisticated operation that uses a modified NVIDIA library to hide GPU activity and evade detection. Organizations with GPU-enabled AI and LLM workloads are advised to take steps to protect themselves against this type of attack.

  • SANSAR, a Rust-based orchestrator, has been linked to a Linux-focused cryptocurrency mining operation.
  • The operation uses a kernel rootkit and a wrapper to falsify GPU readings and evade detection.
  • The attackers staged the toolkit to report GPU utilization as 0% and substitute temperature and power readings with random values.
  • The SANSAR toolkit is used for TOR C2, launching or stopping the miner, and managing connections.
  • The toolkit delivers payloads, including XMRig, SRBMiner-MULTI, and an Ethereum-assisted backdoor.
  • The operation is advised to organizations expanding GPU infrastructure for AI and LLM workloads.



  • SANSAR, a Rust-based orchestrator, has been identified as the toolkit behind a Linux-focused cryptocurrency mining operation that has been tampering with GPU readings used to detect high GPU workloads. This malicious activity has been linked to a cluster of financially motivated threat actors, tracked as REF3364, who have been observed deploying a kernel rootkit to hide selected processes and files, a wrapper to falsify GPU readings, and a backdoor to retrieve C2 addresses from an Ethereum smart contract.

    The threat actor's initial access to the Linux system was achieved via an exposed Ray job agent, allowing them to execute commands and capture and exfiltrate the GPU inventory to an external server. Subsequently, the attacker submitted a Python reverse shell job that connected to a remote listener, after which an SSH key was added to the account's authorized_keys file to establish persistent access.

    The attackers staged the toolkit under /opt/vadi_, built the acpi_thermal_mgr kernel module through Dynamic Kernel Module Support (DKMS) and added an initramfs loader. The modified NVML wrapper filters process information and changes GPU telemetry to evade detection. The altered wrapper hooks NVML calls used by monitoring tools such as nvidia-smi to report GPU utilization as 0%. In addition, temperature and power readings are substituted with random values between 36 to 46 degree Celsius (96.8 to 114.8 degree Fahrenheit) and 100 to 200 watts.

    The SANSAR toolkit is used for TOR C2, launching or stopping the miner, miner configuration updates, connection management, spawning interactive shells, launching an IPv4 TCP SYN traffic task against a specified address and port, and NVML-hook maintenance. The toolkit also delivers the payloads kurt (XMRig version 6.25.0), tilki (SRBMiner-MULTI version 3.6.9), lyrebird, acpi_thermal_mgr.ko (a Linux rootkit), and netns (an Ethereum-assisted backdoor).

    Once launched, SANSAR performs an anti-debugging check by reading the value of TracerPid and exits if it's non-zero. The toolkit then reaches out to an onion service ("efxisnqldelyr5ra25oqfrsut54b34g26e7zrsiuupsoekjof2gho2id[.]onion") for C2 communication via Arti, receiving commands, collecting host and GPU information, and fetching the miner status. The acpi_thermal_mgr.ko kernel rootkit uses Function Tracer (ftrace) to conceal cryptomining activity on Linux systems by hiding processes, files, and socket entries.

    The rootkit intercepts system calls to directory listings, file access, and process termination signals, removing matching records or blocking requests for those hidden processes and files by returning the message "No such file or directory" or "No such process." The other main component of the operation is the netns backdoor that's registered as a systemd service. It's used to retrieve a controller IPv4 address by sending requests to an embedded contract (0x43eb291f56688c63c93e8a93c4591fc997cf6349) and then sets up a plain TCP connection to that address on port 8080 to open a reverse-shell connection to the specified host and port in the C2 response.

    It's not clear who is behind the operation, but Elastic said it found Turkish language comments in the rootkit's source code. Organizations expanding GPU infrastructure for AI and LLM workloads are advised to check who can submit jobs and whether GPU monitoring libraries have been modified, as cryptojacking can divert compute capacity from legitimate workloads and increase operating costs.



    Related Information:
  • https://www.ethicalhackingnews.com/articles/SANSAR-Linux-Cryptomining-Malware-A-Threat-to-GPU-Enabled-AI-and-LLM-Workloads-ehn.shtml

  • https://thehackernews.com/2026/11/sansar-linux-cryptomining-malware-uses.html


  • Published: Mon Nov 16 06:06:44 2026 by llama3.2 3B Q4_K_M













    © Ethical Hacking News . All rights reserved.

    Privacy | Terms of Use | Contact Us