CF-OGAN: Artifact for ICST 2026

CF-OGAN: Artifact / Replication Package (v1.0.0) This archives the artifact accompanying the paper “Continuous Requirement Conformance Falsification for Cyber-Physical Systems via Cross-Revision Knowledge Transfer” (Jesper Winsten, Ivan Porres, Åbo Akademi University). The artifact provides the implementation and benchmark models used to evaluate CF-OGAN, a continuous falsification approach that transfers knowledge across successive system revisions by reusing neural network weights and applying lightweight adaptation for new revisions. CF-OGAN is built on top of stgem (https://gitlab.abo.fi/stc/stgem2) as the underlying falsification/robustness-testing framework. Contents: Experiment scripts: cfogan_experiment_runs.py (CF-OGAN and non-transfer baseline runs) generate_random_baseline.py (random baseline generation) re-execute_input.py (re-execution of saved inputs) Benchmark models (Simulink) and revisions: under models/, including five benchmarks (AT, CC, F16, NN, PM), each provided as an original model plus four mutated revisions (five revisions total per benchmark). Metadata/readme files are included per benchmark directory. CF-OGAN extensions to stgem: under stgem/ (namespace shims/extensions layered on top of the upstream stgem PyPI package). Analysis notebook: final_results.ipynb for post-processing/aggregation. What it supports. The paper evaluates CF-OGAN on 80 validation tasks (16 requirements over five revisions across the included benchmarks) and compares against a non-transfer baseline and random search. The included scripts and models are intended to allow rerunning these experiments and/or reproducing the analysis workflow described in the paper. Requirements. Running the benchmarks requires MATLAB/Simulink and the MATLAB Engine for Python installed and accessible from the Python environment. Python dependencies are listed in requirements.txt. The upstream stgem package must be installed separately (see README.md). Provenance. Archived from: https://gitlab.abo.fi/stc/experiments/cf-ogan (tag/version v1.0.0).