André Cordeiro

MinimapSignatures: Visual Source Code Analysis Across Multiple IDEs

The proposed ecosystem automaticallygenerates anonymized source code minimaps directly from files be-ing edited across different development environments and submitsthem to deep learning models previously trained on large-scaleminimap datasets containing hundreds of thousands of source coderepresentations spanning 10 programming languages, includingJava, Python, JavaScript, C, PHP, Go, Ruby, and Kotlin. The back-end architecture supports detection of type classification, projectidentification, and author-style analysis using convolutional neuralnetworks and explainable AI techniques derived from prior researchon source code minimaps.