Toolkit for stylometry, authorship verification, and writing-style analysis of text and source code.
Code for "Detection of LLM-Generated Java Code Using Discretized Nested Bigrams" (arXiv:2502.15740). Achieves state-of-the-art performance in distinguishing human vs. LLM-written Java.
Explainable multi-modal ML for detecting malicious PyPI packages. Three-modality detection (metadata + AST static analysis + code stylometry), SHAP-driven Ladisa taxonomy mapping (7 attack vectors), real-time CLI scanner, and live PyPI monitoring. F1=0.9993 on 18.5K packages.
What ties your anonymous code to you: the names both projects carry, the lines both have that nobody else does, and the rare words both reach for, set against 784 real repositories. Read only, no network, no model.
To associate your repository with the code-stylometry topic, visit your repo's landing page and select "manage topics."