CVE-2026-34109: Guardian Language-System Unauthenticated OS Command Injection via id Parameter in speech.php
Guardian language-system passes the id GET parameter directly into a PHP exec() call in speech.php (line 18) without sanitization: exec(\"php jobs/speech_audio.php \".$login_session.\" \".$_GET['id'].\" ...\"). No authentication is required. An unauthenticated remote attacker can append shell metacharacters to execute arbitrary OS commands on the server.
Metrics
- CVSS v4.0
- 9.3
- Severity
- CRITICAL
- Fixed in
- —
- Affected Products
- 1
HarborGuard Analysis
Synopsis
OS command injection in the Guardian language-system web application allows an unauthenticated remote attacker to execute arbitrary operating system commands on the server by appending shell metacharacters to the id GET parameter in speech.php. The vulnerable code passes user-supplied input directly into a PHP exec() call with no sanitization or authentication check, reachable over the network with no credentials required. Successful exploitation gives the attacker full command execution on the host, enabling data theft, file modification, and service disruption. No fix version has been published; HarborGuard tracks this advisory and will make a patched-image rebuild available as soon as upstream ships a patch.
HarborGuard Coverage
Detection of CVE-2026-34109 is available across every HarborGuard environment: the CVE is ingested from upstream feeds within minutes of publication and matched against images in customer registries, CI/CD pipelines, and custom-built images derived from guardian/language-system. Any image layer containing the affected commit (up to and including e42c395ec4b03fe62973a669c9209a673838b8a4) is flagged automatically.
AvailableTriage is available with a CVSS v4.0 score of 9.3 (Critical), surfaced alongside each customer org's compliance policy weighting to prioritize severity relative to their environment. Findings are routed to the appropriate team inbox based on per-environment notification rules, so the right engineers see the alert without manual filtering.
AvailableBecause no upstream fix version exists, HarborGuard re-checks the advisory on every ingest cycle and will make a patched-image rebuild available automatically the moment a remediated release is published. In the interim, customers with compensating-control policies can apply network-policy isolation rules or egress filtering to restrict access to speech.php through HarborGuard's policy controls.
Pending upstreamExploit Conditions
- Network reachabilityRequired
The attacker must reach the speech.php endpoint over the network; the vulnerability is exposed on any host where the web application is publicly or internally reachable.
- AuthenticationNot required
No credentials or session token of any kind are needed; the vulnerable endpoint accepts unauthenticated GET requests.
- Victim interactionNot required
The attacker sends a crafted HTTP request directly and receives command output without any action from a logged-in user.
- Attack complexityDetail
Exploitation is reliable and condition-free; appending standard shell metacharacters to the id parameter is sufficient with no race conditions or environment-specific prerequisites.
Blast Radius
- The attacker executes arbitrary OS commands as the web server process user, gaining an interactive or scripted foothold on the host.
- All files readable by the web server process are exposed, including application source code, configuration files, and stored credentials or API keys.
- The attacker can write or overwrite files on disk, injecting backdoors, altering application logic, or corrupting stored data.
- The web server process and dependent services can be terminated or resource-exhausted, taking the application offline.
How HarborGuard Handles This
Available on HarborGuard: because no upstream patch exists for CVE-2026-34109, HarborGuard continuously re-evaluates the advisory on every ingest cycle and will trigger a patched-image rebuild automatically once a fixed version is published upstream. For customers with auto-remediation enabled, that rebuild will be followed by a regression-test run and a PR opened against affected workloads, with no manual intervention required. While no patch is available, HarborGuard's policy controls can be used to apply compensating controls: network-policy isolation to restrict external access to the speech.php endpoint, egress filtering to limit the blast radius of a successful command injection, and flagging of any pipeline promotion of the affected image as a policy violation requiring manual approval. Customers should treat any image containing the affected commit as critical-risk and consider whether the application needs to remain exposed until a fix is available.
- guardian / language-system≤ e42c395ec4b03fe62973a669c9209a673838b8a4
CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N