PRESIDIO detector
presidio-ru PII detection benchmark
Measured on 2 of 41 retained datasets under the same frozen scoring and masking protocol.
47.63%fully hidden annotations
52.30%detected / overlapped
2.98%extra masking
—CPU characters / second
Strengths
Highest complete-masking rates by normalized data type.
Phone numbers & email78.69%613 / 779
Network identifiers78.12%282 / 361
People's names73.39%1,092 / 1,488
Watch list
Lowest observed complete-masking rates; inspect before deployment.
Bank accounts & cards0.54%2 / 371
Passwords, keys & tokens9.47%9 / 95
Documents & identifiers28.53%760 / 2,664
Language coverage
The overall leader can differ from the best choice for one language slice.
| Language | Datasets | Fully hidden ↑ | Detected ↑ | Extra masking ↓ |
|---|---|---|---|---|
| Russian | 2 | 47.63% | 52.30% | 2.98% |
Configuration, source and version
- Family
- PRESIDIO
- Revision
- presidio-analyzer 2.2.364, presidio-ru-recognizers 0.1.0, ru_core_news_sm 3.8.0
- Eligible datasets
- 2 / 41
- Character F1
- 0.594
- Untouched diagnostic
- 47.70% · 3,426 / 7,183
- Benchmark version
- 1.0.3 · 2026-09-09
- CPU seconds / 10k chars
- Not measured
- GPU seconds / 10k chars
- Not measured
Results by dataset
Complete masking is primary; detection shows whether any character overlapped.
| Dataset | Gold annotations | Fully hidden ↑ | Detected ↑ | Extra masking ↓ |
|---|---|---|---|---|
| redmadrobot | 5,516 | 41.70% | 45.14% | 3.35% |
| hivetrace | 1,667 | 67.25% | 76.00% | 0.54% |