On the Effectiveness of Membership Inference in Targeted Data Extraction from Large Language Models
This work addresses privacy concerns for users of large language models by evaluating the practical utility of membership inference attacks in real-world data extraction scenarios, though it is incremental as it builds on prior research linking extraction and inference.
The study tackled the problem of privacy risks from training data memorization in LLMs by integrating multiple membership inference attack techniques into a data extraction pipeline to benchmark their effectiveness, finding that their performance in this integrated setting differs from conventional benchmarks.
Large Language Models (LLMs) are prone to memorizing training data, which poses serious privacy risks. Two of the most prominent concerns are training data extraction and Membership Inference Attacks (MIAs). Prior research has shown that these threats are interconnected: adversaries can extract training data from an LLM by querying the model to generate a large volume of text and subsequently applying MIAs to verify whether a particular data point was included in the training set. In this study, we integrate multiple MIA techniques into the data extraction pipeline to systematically benchmark their effectiveness. We then compare their performance in this integrated setting against results from conventional MIA benchmarks, allowing us to evaluate their practical utility in real-world extraction scenarios.