Search

A large language model refers to a deep learning model characterized by extensive parameters and pretraining on a large-scale corpus, utilized for processing natural language text and generating high-quality text output. The increasing deployment of large language models has brought significant attention to their associated privacy and security issues. Recent experiments have demonstrated that training data can be extracted from these models due to their memory effect. Initially, research on large language model training data extraction focused primarily on non-targeted methods. However, following the introduction of targeted training data extraction by Carlini et al., prefix-based extraction methods to generate suffixes have garnered considerable interest, although current extraction precision remains low. This paper focuses on the targeted extraction of training data, employing various methods to enhance the precision and speed of the extraction process. Building on the work of Yu et al., we conduct a comprehensive analysis of the impact of different suffix generation methods on the precision of suffix generation. Additionally, we examine the quality and diversity of text generated by various suffix generation strategies. The study also applies membership inference attacks based on neighborhood comparison to the extraction of training data in large language models, conducting thorough evaluations and comparisons. The effectiveness of membership inference attacks in extracting training data from large language models is assessed, and the performance of different membership inference attacks is compared. Hyperparameter tuning is performed on multiple parameters to enhance the extraction of training data. Experimental results indicate that the proposed method significantly improves extraction precision compared to previous approaches. [ABSTRACT FROM AUTHOR]

With the rapid development of artificial intelligence, privacy threats are already getting the spotlight. One of the most common privacy threats is the membership inference attack (MIA). Existing MIAs can effectively explore the potential privacy leakage risks of deep neural networks. However, DNNs are usually compressed for practical use, especially for edge computing, MIA will fail due to changes in DNNs' structure or parameters during the compression. To address this problem, we propose CM-MIA, an MIA against compression models, which can effectively determine their privacy leakage risks before deployment. In specific, firstly we use a variety of compression methods to help build shadow models for different target models. Then, we use these shadow models to construct sample features and identify abnormal samples by calculating the distance between each sample feature. Finally, based on the hypothesis test, we determine whether the abnormal sample is a member of the training dataset. Meanwhile, only abnormal samples are used for membership inference, which reduces time costs and improves attack efficiency. Extensive experiments are conducted on 6 datasets to evaluate CM-MIA's attack capacity. The results show that CM-MIA achieves the state-of-the-art attack performance in most cases. Compared with baselines, the attack success rate of CM-MIA is increased by 10.5% on average. [ABSTRACT FROM AUTHOR]

As communication technology advances with 5G, the amount of data accumulated online is explosively increasing. From these data, valuable results are being created through data analysis technologies. Among them, artificial intelligence (AI) has shown remarkable performances in various fields and is emerging as an innovative technology. In particular, machine learning and deep learning models are evolving rapidly and are being widely deployed in practical applications. Meanwhile, behind the widespread use of these models, privacy concerns have been continuously raised. In addition, as substantial privacy invasion attacks against machine learning and deep learning models have been proposed, the importance of research on privacy-preserving AI is being emphasized. Accordingly, in the field of differential privacy, which has become a de facto standard for preserving privacy, various mechanisms have been proposed to preserve the privacy of AI models. However, it is unclear how to calibrate appropriate privacy parameters, taking into account the trade-off between a model’s utility and data privacy. Moreover, there is a lack of research that analyzes the relationship between the degree of differential privacy guarantee and privacy invasion attacks. In this paper, we investigate the resistance of differentially private AI models to substantial privacy invasion attacks according to the degree of privacy guarantee, and analyze how privacy parameters should be set to prevent the attacks while preserving the utility of the models. Specifically, we focus on generative adversarial networks (GAN), which is one of the most sophisticated AI models, and on the membership inference attack, which is the most fundamental privacy invasion attack. In the experimental evaluation, by quantifying the effectiveness of the attack based on the degree of privacy guarantee, we show that differential privacy can simultaneously preserve data privacy and the utility of models with moderate privacy budgets.