# OpenFeature core/v0.19.3

2026-11-17 · [Original release notes](https://github.com/open-feature/flagd/releases/tag/core/v0.19.3) · [Full summary](https://ratatosk.io/en/releases/openfeature/core/v0.19.3)

Security 2 (high 1) · Breaking changes 1

## Security (2)

### [high] JSON Logic context injection fixed in flagd evaluation

- Affects: setups using the flagd binary, Go core, or Go in-process provider
- [GHSA-2wfx-5gcf-mprw](https://github.com/open-feature/flagd/security/advisories/GHSA-2wfx-5gcf-mprw), GHSA-mw4w-7w23-cg45

Before the upgrade, evaluation-context values could be interpreted as JSON Logic, allowing a crafted request to disclose context values the client did not send or crash the process. This could be triggered by flags whose targeting returns a context value from an `if` branch or uses an outer context value inside `filter`, `some`, `all`, or `none`; the dependency update fixes the flaw.

> * **security:** bump jsonlogic to v3.10.2 (GHSA-mw4w-7w23-cg45) ([#2129](https://github.com/open-feature/flagd/issues/2129)) ([bf85e04](https://github.com/open-feature/flagd/commit/bf85e04d95c1b850a28fe883cde767db969cf37d))

Security advisory:

> flagd affected by jsonlogic context injection (GHSA-mw4w-7w23-cg45)

https://github.com/open-feature/flagd/security/advisories/GHSA-2wfx-5gcf-mprw

### go-jose JWK parsing now caps RSA modulus size

- Affects: applications parsing JWKs from untrusted sources
- [CVE-2026-116587](https://www.cve.org/CVERecord?id=CVE-2026-116587), [GHSA-mgcc-cww5-fp2q](https://github.com/advisories/GHSA-mgcc-cww5-fp2q)

Before this update, crafted RSA keys in untrusted JWKs could consume excessive CPU during parsing, especially with Go 1.24 and later. Parsing now rejects RSA moduli over 4096 bits by default; a higher limit can be set with `MaxRSAModulusBits` before parsing.

> * **security:** update module github.com/go-jose/go-jose/v4 to v4.1.6 [security] ([#2114](https://github.com/open-feature/flagd/issues/2114)) ([bf167aa](https://github.com/open-feature/flagd/commit/bf167aaa7686f78bcd206f8b76365d4babfb5a5d))

## Breaking changes (1)

### flagd and flagd-proxy evaluation logging changed

- Affects: setups running flagd or flagd-proxy

In the described string-comparison cases, including missing or wrongly typed context properties, each evaluation previously produced an error log with a stack trace. These issues now log at debug, error-level logs omit stack traces except at `DPanic+`, and repeated messages are sampled at 100/100 (the first 100 identical messages per second, then every 100th); `--debug` disables sampling, and evaluation results are unchanged.

Evidence: “- logs are sampled at 100/100 (first 100 identical messages per second, then every 100th); disabled with `--debug`; applies to flagd and flagd-proxy” https://github.com/open-feature/flagd/pull/2019

> * debug logging for op errs, sample logging ([#2019](https://github.com/open-feature/flagd/issues/2019)) ([50d0f55](https://github.com/open-feature/flagd/commit/50d0f559adbd2318ffc2c2cc83ebf229ec034e70))

---

Summaries are produced by AI, so they can be wrong, and they are provided without warranty.
