Please do not report security vulnerabilities through public GitHub issues.
Instead, report them privately in either of these ways:
- GitHub: use Report a vulnerability on the repository's Security tab. Only the maintainers can see the report.
- Email: agit8or@agit8or.net with the subject line "Depl0y Security Vulnerability".
You should receive a response within 48 hours. If for some reason you do not, please follow up via email to ensure we received your original message.
Please include the following information:
- Type of issue (e.g. buffer overflow, SQL injection, cross-site scripting, etc.)
- Full paths of source file(s) related to the manifestation of the issue
- The location of the affected source code (tag/branch/commit or direct URL)
- Any special configuration required to reproduce the issue
- Step-by-step instructions to reproduce the issue
- Proof-of-concept or exploit code (if possible)
- Impact of the issue, including how an attacker might exploit it
When we receive a security bug report, we will:
- Confirm the problem and determine the affected versions
- Audit code to find any similar problems
- Prepare fixes for all supported releases
- Release patched versions as soon as possible
We release patches for security vulnerabilities for the following versions:
| Version | Supported | Notes |
|---|---|---|
| 2.2.x | ✅ | Current release line. Security fixes land here. |
| 2.0.x – 2.1.x | Not maintained. Upgrade to the latest 2.2.x. | |
| 1.x | ❌ | End of life. Contains fixed critical RCEs — see Historical advisories. |
Always run the latest release.
When deploying Depl0y in production:
- Change the default admin password immediately after installation
- Enable 2FA (TOTP) for all admin accounts, and require it with Security → 2FA Policy → Require 2FA (for admins or for everyone). While the policy applies to an account that has not enrolled yet, that account can only enrol, view its profile and sign out: every other API call, every API key the account owns and every console session (VM/LXC console, node shell) is refused until it enrols
- Use strong, unique passwords (minimum 12 characters)
- Regularly rotate passwords and API tokens
- Use HTTPS in production (configure SSL/TLS certificates)
- Restrict network access to trusted IPs when possible
- Place Depl0y behind a firewall
- Use a VPN for remote access
- Regularly backup your database
- Encrypt sensitive data at rest
- Use encrypted connections to Proxmox hosts
- Store encryption keys securely
- Keep Depl0y updated to the latest version
- Regularly update the host operating system
- Regularly update Proxmox VE to the latest stable version
- Monitor system logs for suspicious activity
- Disable unused features and services
- Follow the principle of least privilege
- Use role-based access control (Admin, Operator, Viewer). Viewers are read-only. Operators manage guests (VMs, containers, their disks, snapshots, backups, consoles). Physical servers (BMC power, event logs, OS package updates), node evacuation, shared storage, cluster join data, Proxmox ACLs and global settings are admin-only
- Scope operators and viewers to the Proxmox hosts they need (Users → Host Access). Every API route that addresses a Proxmox host, node or VM checks that access, and lists only show those hosts; the role still decides what the user may do there. Admins reach every host. PBS servers, standalone BMCs, ISOs and cloud images are shared and not scoped per host
- Regularly audit user accounts and permissions
- Remove inactive user accounts promptly
- Use API tokens instead of passwords when possible
- Create dedicated users for Depl0y in Proxmox
- Limit permissions to only what's needed
- Enable Proxmox audit logging
- Use strong database passwords
- Restrict database access to localhost when possible
- Regularly backup the database
- Keep SQLite updated
- Use the built-in update mechanism from Settings
- Test updates in a non-production environment first
- Review release notes before applying updates
- Subscribe to release notifications on GitHub
Secret scanning and Dependabot alerts are enabled on GitHub, with push protection on. To run the same checks locally before opening a pull request:
# Secrets, in the working tree and in full history.
# .gitleaks.toml allowlists confirmed false positives — documentation
# placeholders, npm integrity hashes and browser storage key names.
gitleaks dir . --config .gitleaks.toml
gitleaks git . --config .gitleaks.toml
# Known vulnerabilities in the pinned Python dependencies.
pip-audit -r backend/requirements.txt
# Frontend dependencies.
cd frontend && npm auditBoth scans are expected to come back clean apart from advisories recorded in the changelog as having no published upstream fix.
- Proxmox credentials are encrypted using Fernet symmetric encryption
- Encryption key must be kept secure and backed up
- Loss of encryption key means loss of stored credentials
- API uses JWT tokens for authentication
- Access tokens last the configured session timeout (8 hours by default); refresh tokens are httpOnly cookies
- API keys (
X-API-Key: dk_...) act as their owner and follow the same role and mandatory-2FA rules; they cannot open console sessions - Console WebSockets (
/api/v1/pve-console/ws/...) never carry the access token. The page first callsPOST /api/v1/pve-console/ws-ticket, which returns a random ticket that works once, for about 30 seconds, for that one user and that one VM, container or node. Only the ticket appears in the socket URL, so proxy and server access logs hold nothing reusable
- Connections to Proxmox, PBS, BMCs and VMware without "Verify SSL" are pinned to an approved certificate (Settings → TLS certificates); SSH connections to nodes and BMCs are pinned to an approved host key (Settings → SSH host keys). A changed certificate or key is always refused before credentials are sent
- By default the first certificate or key seen for a new host is trusted
automatically (trust on first use). Such entries are flagged unconfirmed
on those settings pages until an admin compares the fingerprint with the host
(
openssl x509 -noout -fingerprint -sha256 -in /etc/pve/local/pve-ssl.pem,ssh-keygen -lf /etc/ssh/ssh_host_ed25519_key.pub) and clicks Confirm. Each first-use trust is also written to the audit log - If depl0y may be on the same network as an attacker when it first contacts a host, set "Unknown hosts" to Refuse until an admin approves on both pages before adding hosts. New hosts are then held as pending until approved
- ISO and cloud image uploads are restricted to specific directories
- File size limits are enforced
- File types are validated
- Malicious files should be prevented, but additional scanning is recommended
Subscribe to releases on GitHub to receive notifications about security updates: https://github.com/agit8or1/Depl0y/releases
From 2.2.93 every release is signed with minisign.
Each release has three assets: depl0y.tar.gz, SHA256SUMS and SHA256SUMS.minisig.
install.sh and the in-app updater check the signature against the release key below
before using the checksum, and refuse unsigned releases. The signature's trusted comment
names the release (depl0y vX.Y.Z), so a signature cannot be reused for another version.
Release public key (also in minisign.pub):
RWSr/WW0BJrhEcOJU3TiPs935jc6Qqy7yjLeuu5l+jaG0jlZ0CaHHRjH
To check a release by hand:
V=vX.Y.Z
for f in depl0y.tar.gz SHA256SUMS SHA256SUMS.minisig; do
curl -fsSLO "https://github.com/agit8or1/Depl0y/releases/download/$V/$f"
done
minisign -V -P RWSr/WW0BJrhEcOJU3TiPs935jc6Qqy7yjLeuu5l+jaG0jlZ0CaHHRjH -m SHA256SUMS # expect: Trusted comment: depl0y $V
sha256sum -c SHA256SUMSMultiple remote code execution paths were fixed in v1.3.8: command injection in
the setup and system-update APIs and in the deployment service, plus sensitive
data exposure in logs and a weak cryptographic key. All were resolved by
eliminating shell=True, applying shlex.quote() to dynamic parameters,
validating input with strict patterns, adding subprocess timeouts, and
generating strong keys. Fixing commit:
3786c83.
This is retained for the record. The 1.x line is end of life — if you are still
running it, upgrade to the current 2.2.x release rather than following the
1.3.8-era instructions, which assumed /opt/depl0y was a git checkout. It is
not; see the README's upgrade section.
If you have suggestions on how this process could be improved, please submit a pull request or open an issue.