New Linux Malware Fakes GPU Activity to Stay Hidden
TL;DR: A new Linux malware called SANSAR is hiding its cryptomining activity by modifying an NVIDIA library. This trick fakes GPU usage reports, allowing it to operate undetected on servers and cloud infrastructure, bypassing standard monitoring tools.
Key facts
- Category
- Cybersecurity
- Impact
- High
- Published
- Source
- The Hacker News
Full summary
A new Linux malware fakes GPU usage reports by modifying an NVIDIA library, making it nearly invisible to standard security monitoring tools.
Security researchers have identified a new, sophisticated Linux malware designed for cryptocurrency mining that uses a novel technique to remain undetected. According to a report from The Hacker News, a financially motivated group tracked as REF3364 is deploying a toolkit called SANSAR. This malware is specifically engineered to hide its activity by tampering with the very tools system administrators use to monitor GPU performance. The operation targets Linux-based systems, which form the backbone of most cloud computing and on-premise server environments. By making its resource-intensive mining operations invisible, the malware can steal valuable computing power for extended periods without raising alarms, running up significant costs for its victims. The core of the toolkit is a Rust-based orchestrator that communicates over the Tor network, making it difficult to trace and shut down.
The malware’s primary innovation lies in how it achieves stealth. Instead of just hiding its process or obfuscating its code, SANSAR directly attacks the monitoring infrastructure. It targets the NVIDIA Management Library (NVML), a critical component that provides an API for monitoring and managing NVIDIA GPU devices. This is the same library that powers the widely used `nvidia-smi` command-line utility, which administrators rely on to check GPU utilization, temperature, and power consumption. The malware replaces the legitimate `libnvidia-ml.so.1` file with its own malicious, modified version. This trojanized library acts as a man-in-the-middle for monitoring queries. When an administrator or an automated tool requests GPU status, the malicious library intercepts the call and returns fabricated data suggesting the GPU is idle or lightly used, even when it is actually running at maximum capacity to mine cryptocurrency.
This technique represents a significant escalation in the cat-and-mouse game between attackers and defenders. While cryptojacking has been a persistent threat for years, early variants were often noisy and easily detectable through high CPU or GPU usage. Attackers adapted by using techniques to hide processes or limit resource consumption to stay below monitoring thresholds. The SANSAR malware takes this a step further by corrupting the source of truth itself. This approach is philosophically similar to traditional rootkits, which modify the operating system kernel to hide files, processes, and network connections from the administrator. By applying this rootkit-like principle to the hardware monitoring layer, the attackers demonstrate a deep understanding of modern infrastructure and the specific tools used to manage it. It shows a clear trend toward more targeted and sophisticated evasion tactics that undermine the fundamental assumptions of system monitoring.
For CTOs, IT managers, and security teams, this development is a critical warning: performance metrics from a potentially compromised system cannot be trusted implicitly. Relying solely on GPU utilization alerts to detect cryptojacking is no longer a viable strategy. A defense-in-depth approach is essential. Teams must now correlate monitoring data with other signals to spot anomalies. This includes looking for unexpected network connections, especially to known Tor nodes, analyzing power consumption data at the physical or hypervisor level if possible, and scrutinizing cloud bills for unexplained cost increases in GPU-intensive instances. Furthermore, implementing file integrity monitoring (FIM) on critical system libraries, including hardware-specific ones like NVML, can help detect unauthorized modifications. This incident forces a shift from simply watching for high usage to actively hunting for signs that the monitoring tools themselves have been subverted.
Why it matters
This malware's technique of modifying a core monitoring library (NVML) represents a significant evolution in evasion. It renders GPU utilization metrics unreliable, forcing security and infrastructure teams to question the integrity of their own monitoring tools and find new ways to detect resource abuse.
Business impact
Undetected cryptojacking directly translates to increased operational costs from stolen compute resources, especially on expensive GPU-enabled cloud instances. This SANSAR malware increases financial risk by making it harder to spot and stop unauthorized use of company infrastructure, impacting budgets.
Tags
Related on Notifire
Related stories
Primary source: The Hacker News
