parquet-dos-poc

Parquet Model File Vulnerability PoC

Vulnerability

DoS via Schema Complexity, Column Explosion, and Row Group Metadata Abuse in Parquet files

Files

• poc_many_columns.parquet — 10,000 columns causing CPU/memory exhaustion on schema parsing
• poc_deep_nested.parquet — 20-level nested struct schema causing recursive parsing overhead
• poc_many_rowgroups.parquet — 1,000 row groups causing metadata bloat and parsing overhead
• benign.parquet — Clean file for comparison

Reproduce

import pyarrow.parquet as pq
# CPU exhaustion from schema complexity:
table = pq.read_table('poc_many_columns.parquet')  # Slow parsing of 10K columns

# Recursive parsing overhead:
table = pq.read_table('poc_deep_nested.parquet')  # 20-level nested structs