Plagiarism detection in programming education is critical for preserving academic integrity and supporting fair assessment. While established systems such as MOSS and JPlag effectively detect syntactic and token-level similarities, they often provide opaque outputs and can be evaded by modern obfuscation tools. This paper proposes a multimethod plagiarism detection framework that combines token fingerprinting, abstract syntax trees based structural comparison, rarity-weighted scoring, and semantic/graph-based analysis. The contribution is a design that emphasises multidimensional evidence, visual analytics for explainability, and robustness against evasion. We present the motivation, architecture, and evaluation plan positioning this work as a bridge between robust detection and educator-centred interpretability.