VulnBench
Menu

Dataset catalog

Reconstruct the evidence

Every public VulnBench release links its narrative to immutable source artifacts, documented metrics, and downloadable static data.

Published datasets

Release-scoped and versioned

Dataset versions remain attached to their release protocol. A correction creates a new patch version instead of silently replacing the historical source.

  1. Dataset 2.0.0

    Snyk VulnBench 2.0

    A blind study of nine productized security configurations, from the Snyk Code static analyzer to Claude Code, Codex Security and DeepSec, on 108 hand-curated, attacker-reachable vulnerabilities in 20 open-source web applications.

    Data and provenanceMetric definitions
    Status
    current
    Published
    Dataset
    2.0.0
    Last updated

  2. Dataset 1.0.0

    Snyk VulnBench JS 1.0

    A controlled study of how repeatably agentic LLM security reviews find vulnerabilities in the same JavaScript code, prompt, and benchmark harness.

    Data and provenanceMetric definitions
    Status
    superseded
    Published
    Dataset
    1.0.0
    Last updated

Data policy

Durable by design

  • Static assets require no runtime API or account.
  • Published source revisions remain immutable.
  • Zero, missing, and not-applicable values stay distinct.
  • Metric definitions include units, aggregation, and caveats.
  • Corrections preserve stable release URLs and prior versions.