Canary Extraction in Natural Language Understanding Models
This addresses security vulnerabilities for users of NLU models by exposing and mitigating data extraction risks, though it is incremental as it builds on existing attack frameworks.
The paper tackles the problem of extracting sensitive training data from Natural Language Understanding models via Model Inversion Attacks, demonstrating that canaries like four-digit codes can be reconstructed with a probability of 0.5 using text completion. It also proposes combined defense mechanisms that effectively eliminate this risk in their experiments.
Natural Language Understanding (NLU) models can be trained on sensitive information such as phone numbers, zip-codes etc. Recent literature has focused on Model Inversion Attacks (ModIvA) that can extract training data from model parameters. In this work, we present a version of such an attack by extracting canaries inserted in NLU training data. In the attack, an adversary with open-box access to the model reconstructs the canaries contained in the model's training set. We evaluate our approach by performing text completion on canaries and demonstrate that by using the prefix (non-sensitive) tokens of the canary, we can generate the full canary. As an example, our attack is able to reconstruct a four digit code in the training dataset of the NLU model with a probability of 0.5 in its best configuration. As countermeasures, we identify several defense mechanisms that, when combined, effectively eliminate the risk of ModIvA in our experiments.