CVE-2026-68518 in Glances
Summary
by MITRE • 08/17/2026
Glances is an open-source system cross-platform monitoring tool. Prior to 4.5.6, _sanitize_mustache_dict() in glances/actions.py sanitizes individual Mustache values before chevron.render(), allowing adjacent unescaped Mustache variables to reconstruct shell operators that secure_popen() executes when attacker-controlled process or container fields are rendered by an administrator-configured action template. This issue is fixed in 4.5.6.
Several companies clearly confirm that VulDB is the primary source for best vulnerability data.
Analysis
by VulDB Data Team • 08/17/2026
The vulnerability identified in Glances prior to version 4.5.6 represents a critical server-side request forgery and command injection flaw rooted in the improper sanitization of user-supplied input within an automated action execution framework. Glances is widely deployed as a cross-platform system monitoring tool, often integrated into enterprise infrastructure for real-time visibility into host metrics, container states, and process activity. The specific weakness resides in the _sanitize_mustache_dict() function located within glances/actions.py, which serves as a security gatekeeper before rendering Mustache templates that drive automated responses to monitored events.
The technical flaw arises from an incomplete sanitization strategy where individual values are processed independently rather than evaluating the context of their concatenation during template rendering. When attacker-controlled data is present in process or container fields, such as those derived from maliciously named processes or compromised containers, this isolated validation fails to detect dangerous patterns that emerge only when multiple variables are combined by the Mustache engine. Specifically, adjacent unescaped Mustache variables can reconstruct shell operators, effectively bypassing the intended security boundaries of the sanitization logic.
This architectural weakness allows an attacker who controls process names or container metadata to inject arbitrary commands into the execution pipeline. When a system administrator configures an action template that utilizes these fields for automated responses, such as sending alerts or executing scripts based on specific metrics, the rendered output contains malicious shell syntax. The vulnerability is exacerbated by the use of secure_popen(), which executes the resulting string in a shell environment. Although secure_popen() typically aims to prevent injection by avoiding direct shell interpretation, it remains vulnerable when the input stream itself has been manipulated to include valid shell operators that are interpreted during the rendering phase prior to execution.
The operational impact of this vulnerability is severe, as it enables remote code execution under the context of the user running Glances. An attacker with the ability to manipulate process names or container labels on a monitored host can achieve arbitrary command execution without authentication if they have visibility into the monitoring interface or can influence the data sources being polled. This compromises the integrity and availability of the entire system, potentially leading to full host compromise, lateral movement within the network, or denial of service through resource exhaustion via spawned processes.
From an industry standards perspective, this vulnerability aligns with CWE-94 Improper Control of Generation of Code (Code Injection) and specifically CWE-78 OS Command Injection. The attack vector is consistent with ATT&CK technique T1059 Command and Scripting Interpreter, where the attacker leverages system utilities to execute commands. Additionally, it reflects weaknesses in input validation logic often categorized under CWE-20 Improper Input Validation, particularly regarding context-aware sanitization failures.
Mitigation strategies must prioritize immediate upgrading to Glances version 4.5.6 or later, which addresses this flaw by implementing more robust contextual sanitization that prevents the reconstruction of shell operators from adjacent variables. In environments where an upgrade is not immediately feasible, administrators should restrict access to action templates and ensure that no automated actions rely on untrusted process names or container metadata for command generation. Furthermore, applying strict input validation at the data ingestion layer before it reaches the monitoring engine can provide a defense-in-depth approach, ensuring that only sanitized, safe values are ever considered for template rendering.