What Does This Mean?
An attack in which a machine-learning model is manipulated through malicious inputs or contributions, such as harmful updates sent in federated learning, so that it behaves wrongly or carries a hidden backdoor. The phrase is used close to data poisoning, and the line between them is not always sharp.
REAL-WORLD EXAMPLE
""The security team treated the odd update from one client as possible model poisoning.""
LORE & ORIGIN
Wikipedia's adversarial machine learning article does not use the phrase "model poisoning" in the portion read; it describes data poisoning and, in federated learning, edge devices that may send harmful gradients or parameters to a central server, with backdoors possible. This is a weak entry: the exact slang use of "model poisoning" is not separately documented in opened sources. The entry describes the concept only and gives no method.
MENTIONS OVER TIME
Radar observation data not yet available for this term.
RELATED SEARCHES
TOP PLATFORMS
Not enough Radar data yet to show a platform breakdown.
WHEN DID YOU FIRST HEAR THIS?
CLICK AN OPTION BELOW TO CAST YOUR VOTE.
[ VOTE TO REVEAL COMMUNITY RESULTS ]