Open-source Large Language Models (LLMs) often employ safety alignment methods to resist harmful instructions.However, recent research shows that maliciously fine-tuning these LLMs on harmful data can easily bypass these safeguards.To counter this, we theoretically uncover why malicious fine-tuning succeeds and identify potential defense strategies.Building on the theoretical analysis, we introduce the Self-Degraded Defense (SDD) framework.SDD encourages LLMs to produce high-quality but irrelevant responses to harmful prompts.When attackers attempt malicious fine-tuning, the general capability of the LLM aligned by SDD will significantly decrease, rendering it incapable of following harmful instructions.Our experimental results confirm SDD's effectiveness against such attacks.Our code is available at https://github.com/ZeroNLP/SDD.

Multimodal Large Language Models (MLLMs) have serious security vulnerabilities.While safety alignment using multimodal datasets consisting of text and data of additional modalities can effectively enhance MLLM's security, it is costly to construct these datasets.Existing low-resource security alignment methods, including textual alignment, have been found to struggle with the security risks posed by additional modalities.To address this, we propose Synthetic Embedding augmented safety Alignment (SEA), which optimizes embeddings of additional modality through gradient updates to expand textual datasets.This enables multimodal safety alignment training even when only textual data is available.Extensive experiments on image, video, and audio-based MLLMs demonstrate that SEA can synthesize a high-quality embedding on a single RTX3090 GPU within 24 seconds.SEA significantly improves the security of MLLMs when faced with threats from additional modalities.To assess the security risks introduced by video and audio, we also introduced a new benchmark called VA-SafetyBench.High attack success rates across multiple MLLMs validate its challenge.Our code and data will be available at https://github.com/ZeroNLP/SEA.This paper contains harmful data and modelgenerated content that can be offensive in nature.

Unsupervised domain adaptation-based fault diagnosis methods have been extensively studied due to their powerful knowledge transferability under different working conditions. Despite their encouraging performance, most of them cannot sufficiently account for the temporal dimension of the vibration signal, resulting in incomplete feature information used in the domain alignment procedure. To alleviate the limitation, we present a self-supervised domain adaptation fault diagnosis network (SDAFDN), which considers two temporal dependencies to improve the transferability of the learned representations. Specifically, we first design a down-sampling and interaction network that considers the temporal dependency among subsequences with low temporal resolution in feature space. Then, we combine domain adversarial learning with feature mapping to achieve domain alignment. Finally, we introduced a self-supervised learning module, which considers the temporal dependency between the past and future temporal segments via classification tasks. Extensive experiments on public Paderborn University and PHM data sets demonstrate the superiority of the proposed SDAFDN and the effectiveness of considering temporal dependencies in domain alignment.