QuantDinger is a local-first, self-hosted quantitative trading system.
This policy defines the supported versions, security scope, vulnerability reporting process, remediation targets, and public acknowledgments for the project.
QuantDinger is under active development.
At this stage:
- The
mainbranch is the only supported version for security updates. - Older commits, forks, or modified deployments are not actively supported.
Users are strongly encouraged to stay up to date with the latest release or commit when running QuantDinger in production environments.
QuantDinger is designed to run entirely under the user’s control.
We consider the following areas in scope for security review:
- Source code vulnerabilities in this repository
- Authentication and authorization logic within QuantDinger
- Handling of API keys, secrets, and credentials by the application
- Strategy execution logic and isolation boundaries
- Default configuration security issues
QuantDinger treats uploaded indicator and Strategy API V2 source as untrusted. Static source validation is a policy layer, not the sole security boundary. Supported deployments execute indicator code, strategy discovery, backtests, and live strategy callbacks in child processes that:
- start with a clean environment and no inherited file descriptors;
- receive only explicitly serialized market data, parameters, state, and bounded runtime context;
- exchange data with the trusted host through a typed JSON protocol;
- have no database handle, exchange or broker credential, application secret, or order-submission client;
- apply timeout, memory, file-size, and output-size limits; and
- on Linux, activate fail-closed seccomp and Landlock restrictions before untrusted handlers run.
Trusted API, evaluator, backtest, and trading workers retain authentication, durable-state, reconciliation, credential, and exchange-order authority. An order intent returned by a strategy sandbox is not authorization to trade; it must pass the normal ownership, account, risk, lease, and fencing checks. QuantDinger does not fall back to in-process untrusted-code execution when the kernel boundary cannot be activated.
The following are outside the scope of this security policy:
- Misconfigured user environments (OS, Docker, firewall, cloud host)
- Compromised user machines or infrastructure
- Third-party services, exchanges, or APIs
- Modified or unofficial builds of QuantDinger
Submit suspected security vulnerabilities through the private reporting channel below.
Please do not open a public GitHub issue for security vulnerabilities.
Instead, report privately via email:
- Email:
[email protected] - Subject:
[Security] Brief description of the issue
Please include:
- a clear description of the vulnerability
- steps to reproduce (if applicable)
- potential impact
- any suggested mitigations (optional)
We aim to:
- acknowledge reports within 72 hours
- provide a preliminary assessment within 7 days
- provide a remediation or mitigation target within 14 days
Confirmed vulnerabilities are prioritized using impact, exploitability, and exposure. Our normal remediation targets are:
- critical: mitigation or release within 7 days
- high: mitigation or release within 30 days
- medium: mitigation or release within 60 days
- low: scheduled according to risk and maintenance capacity
Timelines may vary depending on the complexity and severity of the issue.
If a report is accepted, we will coordinate a fix and, when appropriate, a responsible public disclosure.
We ask reporters to keep vulnerability details private until a fix or agreed mitigation is available. QuantDinger will maintain communication with the reporter, prepare upgrade guidance, and publish an advisory or a security entry in this document when disclosure is safe. Publicly known exploitable vulnerabilities of medium or higher severity should not remain unpatched for more than 60 days.
Confirmed vulnerabilities are disclosed through GitHub Security Advisories, release notes, or the Security Advisories section below. Credit is provided unless the reporter requests anonymity.
We ask security researchers to:
- avoid exploiting vulnerabilities beyond proof of concept
- allow reasonable time for remediation before public disclosure
- act in good faith and with respect for users
We are happy to acknowledge responsible disclosures in release notes or documentation, if desired.
A critical vulnerability affected deployments that used the previously
published default SECRET_KEY. Under those conditions, an unauthenticated
attacker could forge an HS256 access token. The former validation path also
allowed a token without a token_version claim to bypass session-version
checks and used the role carried in the token for authorization decisions.
The issue was resolved on July 21, 2026. QuantDinger now rejects missing, known-default, and fewer-than-10-byte signing secrets (32 or more random bytes remain strongly recommended); requires the JWT identity and session claims; and validates the current user status, token version, and role against the authoritative database record before granting access.
Administrators of deployments created before this fix should update to the
latest supported revision, replace SECRET_KEY with a unique high-entropy
value, and require users to sign in again. If an affected API was reachable by
untrusted networks, review access logs and rotate any broker, exchange, or
provider credentials that may have been exposed.
Deployments that accepted untrusted strategy or indicator code before the
August 2026 hardening update should assume application environment secrets may
have been exposed. Upgrade to the latest main revision, review access logs,
and rotate JWT/session secrets, database credentials, provider tokens, and
connected broker or exchange credentials. Preserve access to credentials
encrypted with the previous CREDENTIAL_ENCRYPTION_KEY until they have been
re-encrypted or re-entered; changing that key without a migration makes stored
credentials unreadable.
The current supported isolation model also separates manifest discovery, indicator execution, backtest callbacks, and live callbacks from trusted application processes. Operators upgrading from an older build must recreate the backend and worker containers so the new child-process boundary is active; restarting only the web frontend is insufficient.
Two authenticated security issues involving broker endpoint validation and strategy compilation resource controls were resolved on September 29, 2026. The update restricts broker connections to approved endpoints and enforces process isolation and resource limits during strategy discovery.
Operators should upgrade to the latest supported revision. Deployments that allowed untrusted users to access broker configuration before the update should review relevant logs and rotate affected broker credentials as a precaution. Additional technical details are intentionally limited here to support responsible disclosure.
The following researchers reported security issues through the coordinated disclosure process. Descriptions are intentionally limited to information that is appropriate for public release.
| Researcher | Area | Disclosure date |
|---|---|---|
| Risma Ajul | JWT authentication and authorization | July 2026 |
| Satrio | Strategy and indicator execution isolation | August 2026 |
| Jo Arsy | Proxy-header trust and IP-based controls | September 2026 |
| Dan Aridor and the SPR{K3 security research team | Broker endpoint validation and strategy compilation resource controls | September 2026 |
| Nikhil Agrawal, Mettle Labs | Authorization and tenant-isolation controls | October 2026 |
| Varkey Denny | Strategy-code sandbox validation | November 2026 |
| goranai | Credential redaction in agent-facing interfaces | November 2026 |
| Liam Giannini, webcracy.ch | Role authorization for trading operations; credited as finder | November 2026 |
| Aleff | Strategy and indicator sandbox process and file-access isolation | November 2026 |
QuantDinger is provided as-is, without warranty.
As a self-hosted system, users are responsible for:
- securing their own environments
- protecting API keys and credentials
- complying with applicable laws and regulations