CVE-2026-120699
Description
CubeSandbox is an instant, concurrent, secure, and lightweight sandbox for AI agents. Prior to 0.5.1, any authenticated tenant permitted to call POST /sandboxes can supply metadata["host-mount"] with an arbitrary existing host path because the control plane performs no authorization check or allowed-prefix validation for the path. injectHostDirMounts forwards the path to prepareHostDirVolume, where Cubelet uses mount --rbind through the function to perform a recursive bind mount as root and exposes the host subtree to the tenant's MicroVM through virtio-fs, read-write by default. This allows host-root file access and modification, host code execution, access to the container runtime, and compromise of co-resident tenants without a kernel or virtual-machine-monitor flaw. This issue is fixed in version 0.5.1.
AI Remediation Playbook Pro+
Get step-by-step remediation playbooks with specific commands, verification steps, and rollback procedures.
Upgrade to Pro+AI Attack Path AnalysisPro+
Visualize potential attack chains and lateral movement paths with AI-powered analysis.
AI Risk ForecastingPro+
Predict exploitation likelihood with AI-powered threat intelligence forecasting.
AI Compliance MappingPro+
Map vulnerabilities to compliance frameworks with AI-assisted gap analysis.
MITRE ATT&CK TechniquesThreat Intelligence
How attackers may exploit this vulnerability
Adversaries exploit vulnerabilities in internet-facing applications to gain initial access.
Adversaries use compromised credentials to bypass access controls.
Adversaries exploit software vulnerabilities to elevate privileges.
Adversaries bypass mechanisms designed to control elevated privileges.
Data source: MITRE ATT&CK® Framework —attack.mitre.org