Mareforma search_claims skips tamper checks and serves forged rows (CVE-2026-71533)
GitHub's advisory says Mareforma's search_claims never re-verifies high-trust signatures on read, so forged ESTABLISHED rows that query_claims rejects can still appear in search results. The advisory affects version 0.3.7 and is fixed in 0.3.8.
GitHub Advisory Database has published GHSA-m29v-3g7g-9c63, tracked as CVE-2026-71533, against the Mareforma project. The advisory says search_claims skips the verify-on-read step that query_claims applies to high-trust rows. [1]
According to the advisory, a runtime check showed a forged ESTABLISHED row was excluded by query() but returned by search() with its persisted trust level and no verified key. The advisory notes that a consumer therefore cannot tell the row went unchecked. [1]
The advisory's impact analysis rates the issue high severity because verify-on-read is a stated trust guarantee and search is a primary read path. Its metadata lists the GitHub severity as medium. TechDots reports both figures as the advisory gives them. [1]
The advisory names mareforma 0.3.7 as affected and says the fix is in 0.3.8. It describes a fix direction that mirrors query_claims verification in search_claims and adds a regression test. This describes the advisory's proposed remedy and does not by itself confirm the release contents. [1]
What we couldn't confirm
- No independent confirmation of the vulnerability beyond the GitHub advisory text.
- The source gives no CVSS score, so the numeric severity is unknown.
- No evidence of in-the-wild exploitation is provided.
- The source does not link the 0.3.8 release or confirm it is published on PyPI.
- The source does not say whether the project has issued a statement or whether the maintainers responded.
Sources
- GitHub Advisory Database: GHSA-m29v-3g7g-9c63 (CVE-2026-71533): Mareforma: search_claims skips verify-on-read, serving forged rowsSource published · TechDots found it
Statements labelled as a company's claim have not been independently confirmed by TechDots.
Get Cybersecurity stories by email
Related coverage
nginx2root: five new nginx bugs, patch now and lock origins to Cloudflare
F5 and nginx.org disclosed ten nginx vulnerabilities on November 18, 2026. Five memory corruption flaws, named nginx2root by Xint, can crash workers or run commands, and one HTTP/2 bug is the most widespread. Cloudflare says its edge is unaffected and that its proxy shields origins from the HTTP/2 and HTTP/3 attacks, but patching is still required.
Etherpad pre-auth file read flaw rated critical in GitHub advisory
GitHub Advisory Database has published an advisory for an unauthenticated file-read flaw in Etherpad's /static/* handler, affecting ep_etherpad-lite versions 1.8.14 up to 3.3.3 and fixed in 3.3.3. The advisory rates it critical, CVSS 9.8.