SERSEM: Selective Entropy-Weighted Scoring for Membership Inference in Code Language Models
Abstract
As Large Language Models (LLMs) for code increasingly utilize massive, often non-permissively licensed datasets, evaluating data contamination through Membership Inference Attacks (MIAs) has become critical. We propose SERSEM (Selective Entropy-Weighted Scoring for Membership Inference), a novel white-box attack framework that suppresses uninformative syntactical boilerplate to amplify specific memorization signals. SERSEM utilizes a dual-signal methodology: first, a continuous character-level weight mask is derived through static Abstract Syntax Tree (AST) analysis, spellchecking-based multilingual logic detection, and offline linting. Second, these heuristic weights are used to pool internal transformer activations and calibrate token-level Z-scores from the output logits. Evaluated on a 25,000-sample balanced dataset, SERSEM achieves a global AUC-ROC of 0.7913 on the StarCoder2-3B model and 0.7867 on the StarCoder2-7B model, consistently outperforming the implemented probability-based baselines Loss, Min-K% Prob, and PAC. Our findings demonstrate that focusing on human-centric coding anomalies provides a significantly more robust indicator of verbatim memorization than sequence-level probability averages.
Cite
@article{arxiv.2604.01147,
title = {SERSEM: Selective Entropy-Weighted Scoring for Membership Inference in Code Language Models},
author = {Kıvanç Kuzey Dikici and Serdar Kara and Semih Çağlar and Eray Tüzün and Sinem Sav},
journal= {arXiv preprint arXiv:2604.01147},
year = {2026}
}
Comments
Accepted to the FSE 2026 Poisoned Chalice Competition