Devirt Corpus: Obfuscated JavaScript with Deobfuscation Metrics

A corpus of obfuscated JavaScript samples paired with their deobfuscated outputs and readability scores. Each sample includes metrics on input and output size, and the percentage of original code kept after deobfuscation. The dataset is authored by devirt-dev and was last updated on Hugging Face in June 2026.

Use Cases

• Benchmarking deobfuscation tools based on output size and code retention metrics
• Training machine learning models to detect or reverse code obfuscation techniques
• Analyzing the relationship between obfuscation methods and resulting code readability scores
• Studying the effectiveness of deobfuscators on samples from different sources (e.g., generated vs. real)

Strengths

• Includes per-sample deobfuscation metrics such as input/output byte sizes and a 'kept_pct' score
• Distinguishes samples by source technique and origin (e.g., generated/strings_rc4, real/tranco)
• Provides both raw JSONL data and a browsable Dataset Viewer on Hugging Face

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment
• Freshness should be verified as the last update timestamp is from the future (2026-06-09)

Provenance

Source

devirt-dev on Hugging Face

Collection Method

Samples are run through a deobfuscator; the process likely involves automated collection and transformation.

Freshness

Last updated 2026-06-09 12:49:59