Thieves on Sesame Street: Model Extraction of BERT-based APIs

Research question

Can fine-tuned language models be stolen with nonsensical queries?

Key findings

Random word-sequence queries labelled by the victim suffice to distil a comparable BERT model - Watermarking and membership-based defences offer limited protection

Sources

arxiv.orghttps://arxiv.org/abs/1910.12366

Contribute

Spot an error, or know a record we should add? Members can propose additions and submit corrections.

Submit research Report a correction or request removal

Related in Research Paper Library

Structured Decomposition for Reliable LLM-Generated Access Control Policies

This paper presents an LLM-based system that translates natural-language access control policies (NLACPs) into executable Rego code for Open Policy Agent (OPA). It provides a modular, end-to-end pipeline for policy…

Model extraction and stealing2026

Secrets That Survive Everything: Runtime Credential Exposure in Production Web Applications

Pre-deployment secret scanning operates only on source code, never on what a production application serves. We document two exploitation chains in which Azure AD client credentials and APIM subscription keys from…

Model extraction and stealing2026

End-to-End Hard-Label Cryptanalytic Model Extraction Using Efficient Sign Recovery

The importance of deep neural networks (DNNs) is widely recognized, and the parameters obtained through training are regarded as valuable assets. Recently, attacks that extract these parameters using only oracle…

Model extraction and stealing2026

Thieves on Sesame Street! Model Extraction of BERT-based APIs

Research question

Key findings

Sources

Contribute

Related in Research Paper Library

Structured Decomposition for Reliable LLM-Generated Access Control Policies

Secrets That Survive Everything: Runtime Credential Exposure in Production Web Applications

End-to-End Hard-Label Cryptanalytic Model Extraction Using Efficient Sign Recovery