Evasion Attacks

Definition

Evasion Attacks is inputs crafted at inference time to make a model misclassify or choose the wrong action. In defense applications, it turns stickers, camouflage, waveforms, or text into model-level deception. The hard part is transferability across models and stealthy physical-world perturbations, especially when systems are deployed across contested links, coalition boundaries, and mixed human-machine teams. KhanBMS treats it as a threat KhanBMS counters through ensembles, sensor fusion, and operator-visible uncertainty, tying the concept back to modular command, edge execution, and auditable authority.

Reference attributes

Layer

inference-time attack

Operational value

Turns stickers, camouflage, waveforms, or text into model-level deception

Primary risk

Transferability across models and stealthy physical-world perturbations

KhanBMS role

A threat KhanBMS counters through ensembles, sensor fusion, and operator-visible uncertainty

Related terms

• Adversarial Machine Learning (AML)
  Study and defense of attacks that manipulate AI through crafted inputs, poisoned data, or model theft.
• Automatic Target Recognition (ATR)
  AI-enabled detection and classification of objects, vehicles, emitters, or activities from sensor data.
• AI Sensor Fusion
  Machine-learning methods that combine multiple sensor streams into a better estimate than any source alone.
• Explainable AI (XAI)
  Methods that show why an AI system produced a prediction, recommendation, or action.