Replication Package for Ansible Sensitive Parameter Prediction Study

Replication package for the paper entitled "Smelling Secrets: Leveraging Machine Learning and Language Models for Sensitive Parameter Detection in Ansible Security Analysis", accepted for publication at the 25th IEEE International Conference on Source Code Analysis & Manipulation (SCAM 2025).
Contents00_data_collection — Scripts and results for data collection and ground truth construction.RQ1_ml — Scripts, results, and models for RQ1 (performance of machine learning classifiers)RQ2_lm — Scripts, results, and models for RQ2 (performance of language model classifiers)RQ3_comparison — Scripts and results for RQ3 (comparison of best models from RQ1 and RQ2 against baselines)RQ4_unseen_params — Scripts and results for RQ4 (prediction of unannotated parameters)LicenceData contained within this replication package is licenced under CC BY-NC 4.0.
The source code contained within this package is dual-licenced. You are free to use it for non-commercial purposes under the conditions of the GPL 3.0 license.
If you wish to use this product or any of its code commercially, contact us at: [e-mail]