PII Arena
ONNX detector

pii-shield-onnx PII detection benchmark

Measured on 41 of 41 retained datasets under the same frozen scoring and masking protocol.

56.43%fully hidden annotations
69.64%detected / overlapped
38.92%extra masking
CPU characters / second

Strengths

Highest complete-masking rates by normalized data type.

Passwords, keys & tokens88.40%6,073 / 6,870
Phone numbers & email82.13%18,053 / 21,981
Logins & usernames81.68%9,297 / 11,382

Watch list

Lowest observed complete-masking rates; inspect before deployment.

Organizations16.48%3,177 / 19,274
Other sensitive attributes41.47%1,330 / 3,207
Addresses & locations44.34%17,560 / 39,602

Language coverage

The overall leader can differ from the best choice for one language slice.

LanguageDatasetsFully hidden ↑Detected ↑Extra masking ↓
English1740.53%49.99%44.87%
Russian1956.17%71.66%26.09%
Multilingual568.32%78.31%14.94%

Configuration, source and version

Family
ONNX
Revision
280f050f10a73db88fa9b864e8ca25350e73941e
Eligible datasets
41 / 41
Character F1
0.416
Untouched diagnostic
30.36% · 69,057 / 227,466
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.

DatasetGold annotationsFully hidden ↑Detected ↑Extra masking ↓
rubai-ru3,4585.12%88.29%
multiconer-ru1,20811.18%22.85%4.86%
tab-echr3,83012.40%30.60%
dialogpii-en3,08717.65%27.37%7.48%
alexen21,26122.84%95.00%9.40%
tonicai2,41725.69%27.84%0.56%
nerel24,36931.05%39.48%
gretel-multi4,48434.72%53.88%15.13%
kiji-en7,62337.58%43.46%
nym-en3,58439.23%52.06%2.69%
factrueval7,96639.88%42.02%
nemotron-pii9,39139.93%50.78%9.96%
ameau012,24246.79%58.83%6.95%
jayguard1,19547.70%68.28%6.02%
alrosait1,86252.74%87.49%3.55%
corrupt-redmadrobot5,53154.40%67.85%44.25%
redact-ru9,23055.51%70.10%6.61%
scanpatch8,70858.23%77.45%1.56%
nym-ru9,49758.92%72.92%
redmadrobot5,51659.05%67.95%10.35%
kiji-multi7,56562.49%68.16%
synth-wiki-tables13,09463.45%77.07%
privy1,89963.93%72.14%56.48%
redact-multi22,86470.23%81.56%
corrupt-hivetrace1,66771.33%87.88%15.27%
dialogpii-multi10,34574.31%82.24%
synth-ru-tickets32,09975.35%93.37%
leaky-repo9576.84%82.11%30.07%
hivetrace1,66778.82%89.74%0.28%
russian-pii-66k4,80578.88%85.64%
nym-multi8,16979.21%87.02%1.87%
leak-museum10183.17%90.10%9.91%
secrets-issues28883.68%99.65%38.24%
corrupt-secrets-issues28884.72%100.00%53.05%
synth-jira-comments2,70585.62%91.20%
creddata77692.14%96.26%40.24%
synth-secrets-ru58197.42%99.66%70.31%
synth-secrets-en58197.76%99.66%70.06%
arthur-passwords28097.86%97.86%13.24%
secrets-rules74698.53%99.87%80.36%
synth-env-configs39299.23%100.00%76.71%