CVE-2026-55673: PowSyBl: Command Injection in LocalCommandExecutor-s
Impact
Description Both AbstractLocalCommandExecutor OS-dependent implementations are subject to CWE-78 (OS Command Injection), with a secondary CWE-88 (Argument Injection) concern via environment variables.
The local command executor build command strings via concatenation and executes them (through bash -c for Unix, or cmd /c for Windows). Any string argument or environment variable value reaching this executor can break out of the intended command and execute arbitrary shell code as the JVM user.
The sink is reachable through public APIs that accept List<String> arguments without any documented indication that elements undergo shell interpretation.
Any caller — CLI, library integration, or embedding service — that forwards data from a less-trusted source into these APIs inherits the vulnerability.
Vulnerable elements The vulnerable elements are: - Public methods: - UnixLocalCommandExecutor.execute(...) - WindowsLocalCommandExecutor.execute(...) - LocalComputationManager.execute(...) - ParallelLoadFlowActionSimulator.run(...) - ActionSimulatorTool.run(...) - AmplModelRunner.run(...) - AmplModelRunner.runAsync(...) - itools commands: - action-simulator, when the task-count option is defined - security-analysis, when the external option is defined - dynamic-security-analysis
Characteristics
The vulnerability characteristics are the following: - It allows arbitrary shell command execution as the JVM user (read, write, execute any file; spawn processes; exfiltrate data; etc.) - There are multiple independent injection vectors within a single vulnerable call: - args elements (weakly escaped — bypassable via $(...), backticks, escaped quotes) - Environment variable values (unescaped — bypassable via ;, \n, or shell metacharacters) - It is silent from the caller's perspective: the List<String> API gives no indication that shell interpretation occurs. - It exposes downstream projects: any service embedding powsybl-core (REST front-ends, pypowsybl tooling, multi-tenant grid analysis platforms) that exposes contingency IDs or computation parameters to external input is exploitable without needing to touch powsybl's code directly.
Am I impacted?
You are vulnerable if you make direct calls to one of the vulnerable end-user methods or itools commands listed in the "Vulnerable elements" section, with user-provided parameters, without controlling them.
Patches
com.powsybl:powsybl-computation-local:7.2.2 and higher
Workarounds
If you cannot update your powsybl-computation-local version, you can check the content of the user-provided arguments when calling the vulnerable methods, by forbidding (if possible) the following characters: - On Unix/Linux systems: - \, !, #, $, ^, &, , (, ), {, }, |, [, ], \, ;, \, ", ,, <, >, ?, - On Windows: - %, !, ", \n, \r, ^
Other sources
PowSyBl (Power System Blocks) is a framework to build power system oriented software. Prior to 7.2.2, UnixLocalCommandExecutor and WindowsLocalCommandExecutor concatenate command arguments and environment variables into strings interpreted through bash -c or cmd /c without sufficient escaping. Attacker-controlled values reaching UnixLocalCommandExecutor.execute, WindowsLocalCommandExecutor.execute, LocalComputationManager.execute, ParallelLoadFlowActionSimulator.run, ActionSimulatorTool.run, AmplModelRunner.run, or AmplModelRunner.runAsync can break out of the intended command and execute arbitrary shell commands as the JVM user. The affected itools paths include action-simulator with task-count, security-analysis with external, and dynamic-security-analysis. Downstream CLI tools, libraries, REST front ends, and multi-tenant grid-analysis services that forward less-trusted contingency identifiers or computation parameters into these APIs can expose the injection remotely. This issue is fixed in version 7.2.2.
— MITRE
Affected Software
Remediation
Recommended actions to resolve this vulnerability, in priority order.
- Upgrade
Upgrade
maven/com.powsybl:powsybl-computation-localto a version that resolves this vulnerability.Fixed in 7.2.2 - Upgrade
Upgrade
com.powsybl:powsybl-computation-localto a version that resolves this vulnerability.Fixed in 7.2.2 - Configuration
If you cannot update beyond 7.2.2, validate/reject any user-provided string arguments (including environment-variable values) reaching the vulnerable methods/commands so they cannot contain shell metacharacters that enable command injection (e.g., ;, newline/control characters, $, backticks/$(...), quotes, %, !, \, ^, and the other listed characters).
Local command inputs to the vulnerable itools (action-simulator task-count, security-analysis external, dynamic-security-analysis) Disallow command-injection metacharacters in user-controlled strings = forbid characters ;, \n, \r, $, `, $(...), backticks, %, !, ", \, ^, #, &, *, (, ), {, }, |, [, ], ,, <, >, ?, space
Event History
Frequently Asked Questions
Which deployments are realistically exposed?
Any CLI, library integration, or embedding service that forwards data from a less-trusted source into the affected public APIs is exposed. The command runs with the privileges of the JVM user.
What input must an attacker control to exploit this issue?
An attacker needs control over any string argument or environment-variable value that reaches a local command executor. Such values can escape the intended command and execute shell code.
Which application paths should be reviewed for reachable execution?
Review uses of UnixLocalCommandExecutor.execute, WindowsLocalCommandExecutor.execute, LocalComputationManager.execute, ParallelLoadFlowActionSimulator.run, ActionSimulatorTool.run, AmplModelRunner.run, and AmplModelRunner.runAsync. Also review itools command paths mentioned by the advisory.
Does the execution method differ by operating system?
Yes. Unix execution uses bash -c, while Windows execution uses cmd /c; both implementations construct command strings through concatenation and are affected.