The proliferation of Large Language Models (LLMs) has raised concerns over training data privacy. Membership Inference Attacks (MIA), aiming to identify whether specific data was used for training, pose significant privacy risks. However, existing MIA methods struggle to address the scale and complexity of modern LLMs. This paper introduces OR-MIA, a novel MIA framework inspired by model optimization and input robustness. First, training data points are expected to exhibit smaller gradient norms due to optimization dynamics. Second, member samples show greater stability, with gradient norms being less sensitive to controlled input perturbations. OR-MIA leverages these principles by perturbing inputs, computing gradient norms, and using them as features for a robust classifier to distinguish members from non-members. Evaluations on LLMs (70M to 6B parameters) and various datasets demonstrate that OR-MIA outperforms existing methods, achieving over 90% accuracy. Our findings highlight a critical vulnerability in LLMs and underscore the need for improved privacy-preserving training paradigms.

With the widespread application of large language models (LLM), concerns about the privacy leakage of model training data have increasingly become a focus. Membership Inference Attacks (MIAs) have emerged as a critical tool for evaluating the privacy risks associated with these models. Although existing attack methods, such as LOSS, Reference-based, min-k, and zlib, perform well in certain scenarios, their effectiveness on large pre-trained language models often approaches random guessing, particularly in the context of large-scale datasets and single-epoch training. To address this issue, this paper proposes a novel ensemble attack method that integrates several existing MIAs techniques (LOSS, Reference-based, min-k, zlib) into an XGBoost-based model to enhance overall attack performance (EM-MIAs). Experimental results demonstrate that the ensemble model significantly improves both AUC-ROC and accuracy compared to individual attack methods across various large language models and datasets. This indicates that by combining the strengths of different methods, we can more effectively identify members of the model’s training data, thereby providing a more robust tool for evaluating the privacy risks of LLM. This study offers new directions for further research in the field of LLM privacy protection and underscores the necessity of developing more powerful privacy auditing methods.

ShuffleNet, an efficient neural network architecture, has gained prominence in computer vision for its high accuracy with low computational cost. However, its sensitivity to noise and perturbations limits robustness and feature representation. Existing methods like data augmentation and ensemble lack efficiency in noise reduction. This paper introduces MIR-ShuffleNet, enhancing ShuffleNet with regularized mutual information for noise reduction. MIR- ShuffleNet employs mutual information to enhance effective post-convolution features, reducing noise impact and redundancy. Extensive experiments on five datasets demonstrate MIR- ShuffleNet's efficiency and robustness superiority over existing ShuffleNet variants.

Designing an effective representation learning method for multimodal sentiment analysis is a critical research area. The primary challenge is capturing shared and private information within a comprehensive modal representation, especially when dealing with uniform multimodal labels and raw feature fusion.To overcome this challenge, we propose a novel deep modal shared information learning module that utilizes the covariance matrix to capture shared information across modalities. Additionally, we introduce a label generation module based on a self-supervised learning strategy to capture the private information specific to each modality. Our module can be easily integrated into multimodal tasks and offers flexibility by allowing parameter adjustment to control the information exchange relationship between modes, facilitating the learning of private or shared information as needed. To further enhance performance, we employ a multi-task learning strategy that enables the model to focus on modal differentiation during training. We provide a detailed formulation derivation and feasibility proof for the design of the deep modal shared information learning module.To evaluate our approach, we conduct extensive experiments on three common multimodal sentiment analysis benchmark datasets. The experimental results validate the reliability of our model, demonstrating its effectiveness in capturing nuanced information in multimodal sentiment analysis tasks.