project

Caliber

An output-perturbation defense for machine-learning APIs that return confidence scores, introduced in a 2026 arXiv preprint. It adds Gaussian noise to model logits with the variance normalized by the model's squared median top-two logit margin.

Known aliases

Caliber: Cross-Architecture Extraction-Cost Control for Score-Returning APIs

Relationships

No evidence-backed relationships are recorded.

Current clusters

build1 publisher

Caliber scales its extraction noise by each model's median top-two logit margin

An arXiv defense adds Gaussian noise to a served model's logits and solves for the scale that hits a chosen accuracy target, and it bounds how many repeated queries an attacker needs to average that noise away.

Publishers:arxiv.org

Reality

Evidence44
Adoption8
Hype gap+12
Incentives: Insufficient
Confidence42