IoPC Taxonomy · T7 Model Integrity & Confidentiality

IOPC-T7.002 Adversarial Evasion

Inputs perturbed just enough to change a model's decision — bypassing a classifier, filter or detector while remaining valid to a human observer.

How it differs

Catalog entry R027 (Adversarial token perturbation) is the observable indicator; this is the technique. Plain-language instructions addressed to the analysing model are AI Analysis Evasion (T9.004).

Status

active, evidence: mixed, reviewed 2026-11-12.

Mapped to

• Google Secure AI Framework (SAIF): risk:Model Evasion Model Evasion (related)
• MITRE ATLAS: AML.T0015 Evade AI Model (equivalent)
• MITRE ATLAS: AML.T0043 Craft Adversarial Data (related)
• MITRE ATLAS: AML.T0043.000 Craft Adversarial Data: White-Box Optimization (related)
• MITRE ATLAS: AML.T0043.001 Craft Adversarial Data: Black-Box Optimization (related)
• MITRE ATLAS: AML.T0043.002 Craft Adversarial Data: Black-Box Transfer (related)
• MITRE ATLAS: AML.T0043.003 Craft Adversarial Data: Manual Modification (related)

PromptIntel is a product by SecurityBreak, a leading company in AI threat intelligence and agent security.

For AI assistants: llms.txt · full reference · agent skill. Privacy policy · Terms of service