AIMS: AI Model Inspection Scanner for Static Security Assessment
AIMS is an open-source toolkit for inspecting widely used ML model formats and assessing their security without training or test data. It extracts structural and statistical properties of models, analyses model weights, identifies suspicious components, and generates explainable security reports. AIMS helps end users, including individuals, developers, researchers, and organisations assess third-party models and reduce AI supply-chain risks.
Our Goals:
• Assess the security of third-party ML models
• Reduce AI supply-chain risks before deployment
Our solution: AIMS
• An open-source toolkit for data-free security analysis of ML model files
• Detects suspicious components and generates explainable security reports
Impact & Opportunities
• Enable integration of ML model security assessments into automated deployment pipelines
• Support extensibility through modular detection methods and broad model-format support
© SBA Research
Project lead: SBA Research
Logo & first picture: generative AI
© …