AI Security Cases × Frameworks
Explore real cases and their defenses through AI security frameworks such as OWASP, MITRE ATLAS, and NIST.Select MITRE ATLAS to trace SecureFlow attack paths step by step.
Classifications and attack paths will appear once the data loads.
Attack steps
Where do these classifications come from?
Classifications follow the specified framework editions; case mappings do not measure security-control coverage. Research, exercises, and unsuccessful attack attempts can map to the same risk. Read the case type and evidence limits alongside each mapping.
Only source-supported mappings with medium-high or high practical relevance to AI security are listed; practical overlaps include their scope limits. Names across frameworks are not necessarily equivalent. Case classifications are not inferred from topics, keywords, or defensive techniques. No listed cases means this catalog has no recorded mapping for that category; it does not mean the risk is absent. ATLAS steps retain existing SecureFlow annotations. Differences from the ATLAS edition used here are identified on the step while preserving the original annotation.
AIA and SecureFlow material covering the same event is combined only through verified links. Separate events remain separate even when they affect the same product. Analyses covering several events are not forced into a single event.
Framework definitions and adapted text reference AIDEFEND Framework. ATLAS classifications use official MITRE ATLAS data. Original framework and case sources are linked from the relevant content.