SECAI Core · phase 11 of 15 · noun

Model inversion attack

A security attack where an attacker trains the model with a specific set of records and uses it to inquire the data used to train the model

The Explain card

Plain English
A model inversion attack uses a model's responses to reconstruct or infer the data it was trained on. The attacker probes with crafted queries and works backwards from the outputs.
Example
By repeatedly querying a face recognition API and optimizing inputs against its confidence scores, an attacker reconstructs a recognizable image of a person in the training set.
Why it matters
Models can leak their training data through their answers alone. Defenders limit confidence-score exposure, rate-limit queries, and apply privacy-preserving training techniques.
Hook
Running the model in reverse to see what it saw.

Word knowledge

How the word is built, where it came from, and what it sits beside in memory.

In a sentence

The privacy team ran a model inversion attack against a face-recognition API, then compared the reconstructed portraits with protected enrollment images.

Why these words
  • model Italian modello, from Latin modulus, small measure identifies the computational system targeted
  • inversion Latin inversio, a turning upside down names the reversal from observable results toward concealed information
  • attack French attaque, from Italian attaccare, to join battle classifies the phrase as a hostile security technique
Where it came from
Origin
English security-research coinage formed from model, inversion, and attack, introduced by Matthew Fredrikson and colleagues
Entered the language
2014
What changed
Inversion first described physical or logical reversal, then acquired a specialized machine-learning sense within this security phrase.
How it is spelled
Pattern
The phrase is written as three separate words.
Pattern
Inversion uses the noun ending -sion.
Breaks the pattern
Invert becomes inversion by replacing the final t sound with s before -ion.

Spelled like

  • membership inference attack
  • model extraction attack
  • conversion
  • reversion
Broken into chunks
  • model
    • modeling
    • modeler
  • invers
    • inverse
    • inversely
  • attack
    • attacker
    • attacking
What it sits beside

Same subject

  • membership inference attack
  • model extraction attack
  • data poisoning attack
  • adversarial example

Same shape

  • membership inference attack
  • model extraction attack
  • data poisoning attack

privacy attacks

  • model inversion attack
  • membership inference attack

machine-learning security attacks

Where it sits in the deck

Phase 11: Threat Landscape: Attack Vectors and Adversarial Techniques

With defences named, learn what they must defend against — the full catalogue of attack techniques targeting AI systems, their inputs, outputs, training pipelines, and supply chains.

Neighbours in this phase