CVE-2026-55407: Anthropics Buffa memory exhaustion in protobuf decoder

CVE-2026-55407 · Severity: medium · CVSS 4 · Published 2026-07-16

Buffa is a Rust protobuf code generator used to serialize and deserialize message data. The library's default behavior of preserving unknown fields in decoded messages allows an attacker to send specially crafted protobuf payloads that consume excessive heap memory without triggering size limits. A small malicious message can force the decoder to allocate many times its own size in memory, enabling an unauthenticated remote attacker to crash services through memory exhaustion with minimal network traffic.

Vendors: Anthropic, crates.io.

Quick answers

Is CVE-2026-55407 exploited in the wild?
No. Junglewise Threat Intelligence has no report of CVE-2026-55407 being exploited in the wild.
How severe is CVE-2026-55407?
Medium severity, with a CVSS score of 4, published 16 July 2026.
Which products and versions are affected?
Anthropic Buffa < 0.8.0; buffa (crates.io) SEMVER:introduced=0,fixed=0.8.0.
Is there a fix for CVE-2026-55407?
Yes. buffa (crates.io): upgrade to 0.8.0 or later. Junglewise finds every place you run it, drafts the upgrade and applies it once you approve.
How is CVE-2026-55407 attacked?
It is attacked over the network, in a software library.
Am I affected by CVE-2026-55407?
Check whether you run Anthropic Buffa or buffa (crates.io) at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

How to fix

  • buffa (crates.io): upgrade to 0.8.0 or later
    cargo update -p buffa

Technical details

The decode_unknown_field function in buffa's protobuf decoder fails to enforce allocation limits on heap memory consumed during decoding of unknown fields from untrusted protobuf messages. By default, generated code preserves unknown fields, and the top-level message size cap does not account for in-decode memory amplification. Nested unknown fields within a StartGroup can achieve approximately 22× amplification (e.g., a 64 MiB input forcing ~1.4 GB allocation); length-delimited unknown fields can be sized arbitrarily, offering even greater amplification. The vulnerability is reachable from default decode APIs without authentication or user interaction. Version 0.8.0 fixes this by enforcing per-message unknown field count limits (default 1 million fields, ~40 MiB overhead). Mitigation for older versions includes setting preserve_unknown_fields=false or implementing external size limits (e.g., connectrpc's 4 MiB default still allows ~88 MiB amplification).

Affected products

  • Anthropics buffa < 0.8.0

Identifiers

  • CVE-2026-55407
  • GHSA-f9qc-qg88-7pq5

Timeline

  • 2026-07-01: disclosed
  • 2026-07-16: advisory
  • 2026-08-28: patched

References

This page as Markdown. Free to reuse under CC BY 4.0 with a link back.