C5 / 10.53% of index
Correctness
Is detection accuracy measured against named ground truth?
What is correctness?
Correctness measures whether the discovery system finds real cryptographic assets while controlling false negatives, false positives and partial matches.
An inventory can look comprehensive while silently missing assets. Named corpora, independent ground truth and reproducible evaluators make accuracy claims auditable.
Questions to ask a vendor
- What corpus was used?
- Who constructed and labeled the ground truth?
- Are precision, recall and partial-match rules published per detector?
What this edition actually supports
Weight 2 of 19. How many findings are right, and what did the scanner miss?
A useful accuracy study fixes a corpus and independently adjudicated labels before detection, then reports false positives and false negatives by surface. QScout scores 6 in edition 2026.7. Its published 1,368-case corpus is vendor labelled, with detector-shaped vocabulary and a precision construction artifact; it does not prove market-leading field accuracy. Missing peer metrics are unknown, not automatic product failure.
How to investigate this criterion
Look for a frozen denominator, label provenance, replayable cases and confidence intervals; compare peers on the same cases only.
Primary and original inputs: api.qtonicquantum.com · pqc-index.org. Review all dated cell source checks and the 1 October per-cell limits.
The 2026.7 cells below are public-evidence scores. A vendor page documents a claim; an artifact check tests only its stated scope. Missing product execution and inaccessible sources remain explicit unknowns. The score snapshot has not been recomputed from later pages.
Correctness scores in edition 2026.7
Equal scores share the same criterion position; the next position skips the tied places. The highest value in this column is a published rubric cell, not a measured product winner. A partial or unverified source anchor remains visible even when its numeric score is high.
| Position on criterion | Product | Correctness score | Overall score | Internal 1 October source review | Evidence |
|---|---|---|---|---|---|
| 1 | QScout Pulse GoldQtonic Quantum Corp | 6 | 8.00 | Partial or qualified supportPublic corpus arithmetic recomputes but labels are partly detector-derived, legacy, and narrow; it is not an independently adjudicated product miss rate.Exact cell and citations | S27S28S29S30 |
| 2 | CBOM SecureEncryption Consulting | 4 | 7.47 | Partial or qualified supportVendor says false positives are eliminated but gives no method or precision/recall metric; anchor is only a documented claim.Exact cell and citations | S04S05S06 |
| 2 | Keyfactor AgileSec + CommandKeyfactor | 4 | 6.53 | Narrow feature documentedVendor records false-positive correction for secure key exchange, without published precision/recall.Exact cell and citations | S08S25 |
| 2 | QCecuring CBOMQCecuring | 4 | 6.53 | Partial or qualified supportParsing limitations describe potential misses, not a measured false-positive rate; score four remains a rubric judgment.Exact cell and citations | S19 |
| 2 | AppViewX Quantum Trust HubAppViewX | 4 | 6.37 | Claim not verified from accessible evidenceZero reflects no documented metric in a 403-constrained search, not evidence of no accuracy work.Exact cell and citations | S07S24 |
| 2 | SandboxAQ AQtive GuardSandboxAQ | 4 | 6.21 | Partial or qualified supportVendor says false positives reduced but provides no method or metric in cited source.Exact cell and citations | S12 |
| 2 | DigiCert Quantum CentralDigiCert | 4 | 5.37 | Narrow feature documentedVendor documents manual handling for false positives, without a published accuracy metric.Exact cell and citations | S18 |
| 8 | Fortanix Key Insight / PQC CentralFortanix | 0 | 5.74 | Claim not verified from accessible evidenceZero reflects no metric in bounded accessible docs, not an observed correctness result.Exact cell and citations | S15 |
| 8 | IBM Guardium + Quantum SafeIBM | 0 | 5.58 | Partial or qualified supportFinal zero correctly excludes a CBOMkit benchmark not tied to Guardium/Quantum Safe; absence of IBM-specific metric is bounded.Exact cell and citations | S09S10S26 |
| 8 | O3 SecurityO3 Security | 0 | 5.47 | Claim not verified from accessible evidenceZero is a bounded no-metric finding; docs portal access limits prevent complete correctness review.Exact cell and citations | S17 |
| 8 | ISARA AdvanceISARA | 0 | 5.42 | Claim not verified from accessible evidenceZero is a bounded lack of public accuracy metric; source marketing claim is not a benchmark.Exact cell and citations | S14 |
| 8 | QuSecure QuProtect R3QuSecure | 0 | 5.21 | Claim not verified from accessible evidenceVendor says always accurate but provides no benchmark; zero means not documented under rubric, not measured inaccuracy.Exact cell and citations | S13 |
| 8 | TYCHON Quantum CommandTYCHON | 0 | 5.16 | Claim not verified from accessible evidenceVendor pages returned 403; zero is bounded no metric, not a product accuracy result.Exact cell and citations | S11 |
| 8 | CryptoNext COMPASSCryptoNext | 0 | 4.68 | Claim not verified from accessible evidenceZero is a bounded absence of published accuracy evidence, not measured error rate; archived no-hit search was not reproduced.Exact cell and citations | S16 |