SIMONE: Architecture-Sensitive Near-miss Clone Detection for Simulink Models

Identification of similar fragments of software systems, or clones, has many applications in software engineering and maintenance, including quality control and improvement, standards compliance, test management and failure analysis and prevention. Code similarity analysis systems, or clone detectors, are a mature and widely used technology in tra-ditional software code maintenance. As model driven engi-neering continues to advance, technologies such as Simulink are increasingly widely used to design and implement auto-motive software systems. Automotive Simulink models are particularly prone to cloning due to the copy-paste author-ing paradigm of the Simulink IDE, and the inherent similar-ity of elements and tasks in automotive applications. Thus the ability to find Simulink model clones is equally impor-tant, but is much less thoroughly studied and used. Simulink models can be viewed as graphs, and subgraph isomorphism is the most obvious technique to implement model similarity analysis. While graph-based model clone detection techniques show good results in finding exactly similar subgraphs in graphical models, they have difficulty in finding near-miss matches, which may vary by inciden-tally added or removed blocks, lines, inputs or outputs, and largely ignore the architectural structure of the model, mak-ing analysis results difficult to communicate to practitioners. In this talk I will introduce SIMONE, a Simulink model clone detector designed to address these issues. SIMONE is a hybrid architecture / text-based clone detector that yields actionable results that can be adopted directly into the mod-elling engineer’s workflow as part of their everyday Simulink IDE interactions. SIMONE uses a two-stage analysis, begin-ning with the extraction and normalization of the architec-tural elements (subsystems) to be analyzed, and then uses the mature text-based near-miss code clone detector NICAD to efficiently identify structurally meaningful near-miss sub-system clones. By transforming the graph-based models to normalized text form for comparison, SIMONE is able to un-cover important model similarities that are difficult to fin