Adversarial Evasion

Evasion Attack - An evasion attack is one of the most common attacks on machine learning models performed during production. It involves designing an input that appears normal to a human but is misclassified by ML models, affecting their behavior; a typical example includes altering pixels in an image to cause an image recognition system to fail in classification.

Class Information

JSON JSON-LD OWL XML Markdown

Identification

Label (rdfs)
Evasion Attack
Preferred Label
Adversarial Evasion
Alternative Labels
Deceptive Input Attack, Evasion, Misclassification Attack
Identifier
N/A

Definition and Examples

Definition
An evasion attack is one of the most common attacks on machine learning models performed during production. It involves designing an input that appears normal to a human but is misclassified by ML models, affecting their behavior; a typical example includes altering pixels in an image to cause an image recognition system to fail in classification.
Examples
  • N/A

Translations

N/A

Class Relationships

Sub Class Of
Parent Class Of
  • N/A
Is Defined By
N/A
See Also
N/A

Additional Information

Comment
N/A
Description
N/A
Notes
  • N/A
Deprecated
False

Metadata

History Note
N/A
Editorial Note
N/A
In Scheme
N/A
Source
N/A
Country
N/A

Graph