Skip to content

Dataset · bigcode

Labeled source code for PII detection

bigcode/bigcode-pii-dataset

Code snippets in 31 languages annotated with password, key, and personal data leaks.

What's inside

The dataset contains 12,099 code snippets, each about 50 lines long, across 31 programming languages for the token-classification task. Annotations were performed by 1,399 crowdworkers from 35 countries on the Toloka platform, identifying names, logins, emails, IP addresses, keys, passwords, and identifiers.

What you can do with it

  • Training NER and token classification models to find secrets and PII in code
  • Creating repository cleaning filters before training code LLMs
  • Evaluating the quality of automated source code security audit systems
Tasks
Token classification
Languages
Code
Size
10K–100K rows
Downloads this month
143
Likes
62
Access
on request on Hugging Face

How to load it

from datasets import load_dataset

ds = load_dataset("bigcode/bigcode-pii-dataset")