Continual relation extraction (CRE) aims to continuously learn relations in new tasks without forgetting old relations in previous tasks.Current CRE methods are all rehearsal-based which need to store samples and thus may encounter privacy and security issues.This paper targets rehearsal-free continual relation extraction for the first time and decomposes it into task identification and within-task prediction sub-problems. Existing rehearsal-free methods focus on training a model (expert) for within-task prediction yet neglect to enhance models’ capability of task identification.In this paper, we propose an Ensemble-of-Experts (EoE) framework for rehearsal-free continual relation extraction. Specifically, we first discriminatively train each expert by augmenting analogous relations across tasks to enhance the expert’s task identification ability. We then propose a cascade voting mechanism to form an ensemble of experts for effectively aggregating their abilities.Extensive experiments demonstrate that our method outperforms current rehearsal-free methods and is even better than rehearsal-based CRE methods.
论文检索
输入标题、作者或关键词，从 2,111 篇学术成果中精准定位
How Johnny Can Persuade LLMs to Jailbreak Them: Rethinking Persuasion to Challenge AI Safety by Humanizing LLMs
PDF ↗Most traditional AI safety research views models as machines and centers on algorithm-focused attacks developed by security experts. As large language models (LLMs) become increasingly common and competent, non-expert users can also impose risks during daily interactions. Observing this, we shift the perspective, by treating LLMs as human-like communicators to examine the interplay between everyday language interaction and AI safety. Specifically, we study how to persuade LLMs to jailbreak them. First, we propose a persuasion taxonomy derived from decades of social science research. Then, we apply the taxonomy to automatically generate persuasive adversarial prompts (PAP) to jailbreak LLMs. Results show that persuasion significantly increases the jailbreak risk across all risk categories: PAP consistently achieves an attack success rate of over 92% on Llama-2-7b-Chat, GPT-3.5, and GPT-4 in 10 trials, surpassing recent algorithm-focused attacks. On the defense side, we explore various mechanisms against PAP, find a significant gap in existing defenses, and advocate for more fundamental solutions for AI safety.
Security practitioners maintain vulnerability reports (e.g., GitHub Advisory) to help developers mitigate security risks. An important task for these databases is automatically extracting structured information mentioned in the report, e.g., the affected software packages, to accelerate the defense of the vulnerability ecosystem.However, it is challenging for existing work on affected package identification to achieve high precision. One reason is that all existing work focuses on relatively smaller models, thus they cannot harness the knowledge and semantic capabilities of large language models.To address this limitation, we propose VulLibGen, the first method to use LLM for affected package identification. In contrast to existing work, VulLibGen proposes the novel idea to directly generate the affected package. To improve the precision, VulLibGen employs supervised fine-tuning (SFT), retrieval augmented generation (RAG) and a local search algorithm. The local search algorithm is a novel post-processing algorithm we introduce for reducing the hallucination of the generated packages. Our evaluation results show that VulLibGen has an average precision of 0.806 for identifying vulnerable packages in the four most popular ecosystems in GitHub Advisory (Java, JS, Python, Go) while the best average precision in previous work is 0.721. Additionally, VulLibGen has high value to security practice: we submitted 60 <vulnerability, affected package> pairs to GitHub Advisory (covers four ecosystems) and 34 of them have been accepted and merged.
Textual data is often represented as real-numbered embeddings in NLP, particularly with the popularity of large language models (LLMs) and Embeddings as a Service (EaaS). However, storing sensitive information as embeddings can be susceptible to security breaches, as research shows that text can be reconstructed from embeddings, even without knowledge of the underlying model. While defence mechanisms have been explored, these are exclusively focused on English, leaving other languages potentially exposed to attacks. This work explores LLM security through multilingual embedding inversion. We define the problem of black-box multilingual and crosslingual inversion attacks, and explore their potential implications. Our findings suggest that multilingual LLMs may be more vulnerable to inversion attacks, in part because English-based defences may be ineffective. To alleviate this, we propose a simple masking defense effective for both monolingual and multilingual models. This study is the first to investigate multilingual inversion attacks, shedding light on the differences in attacks and defenses across monolingual and multilingual settings.
Recently, the powerful large language models (LLMs) have been instrumental in propelling the progress of recommender systems (RS). However, while these systems have flourished, their susceptibility to security threats has been largely overlooked. In this work, we reveal that the introduction of LLMs into recommendation models presents new security vulnerabilities due to their emphasis on the textual content of items. We demonstrate that attackers can significantly boost an item’s exposure by merely altering its textual content during the testing phase, without requiring direct interference with the model’s training process. Additionally, the attack is notably stealthy, as it does not affect the overall recommendation performance and the modifications to the text are subtle, making it difficult for users and platforms to detect. Our comprehensive experiments across four mainstream LLM-based recommendation models demonstrate the superior efficacy and stealthiness of our approach. Our work unveils a significant security gap in LLM-based recommendation systems and paves the way for future research on protecting these systems.
The swift detection of multimedia fake news has emerged as a crucial task in combating malicious propaganda and safeguarding the security of the online environment. While existing methods have achieved commendable results in modeling entity-level inconsistency, addressing event-level inconsistency following the inherent subject-predicate logic of news and robustly learning news representations from poor-quality news samples remain two challenges. In this paper, we propose an Event-diven fake news detection framework (Event-Radar) based on multi-view learning, which integrates visual manipulation, textual emotion and multimodal inconsistency at event-level for fake news detection. Specifically, leveraging the capability of graph structures to capture interactions between events and parameters, Event-Radar captures event-level multimodal inconsistency by constructing an event graph that includes multimodal entity subject-predicate logic. Additionally, to mitigate the interference of poor-quality news, Event-Radar introduces a multi-view fusion mechanism, learning comprehensive and robust representations by computing the credibility of each view as a clue, thereby detecting fake news. Extensive experiments demonstrate that Event-Radar achieves outstanding performance on three large-scale fake news detection benchmarks. Our studies also confirm that Event-Radar exhibits strong robustness, providing a paradigm for detecting fake news from noisy news samples.
CoGenesis: A Framework Collaborating Large and Small Language Models for Secure Context-Aware Instruction Following
PDF ↗With the advancement of language models (LMs), their exposure to private data is increasingly inevitable, and their deployment (especially for smaller ones) on personal devices, such as PCs and smartphones, has become a prevailing trend. In contexts laden with user information, enabling models to both safeguard user privacy and execute commands efficiently emerges as an essential research imperative. In this paper, we propose CoGenesis, a collaborative generation framework integrating large (hosted on cloud infrastructure) and small models (deployed on local devices) to address privacy concerns logically. Initially, we design a pipeline to create personalized writing instruction datasets enriched with extensive context details as the testbed of this research issue. Subsequently, we introduce two variants of CoGenesis based on sketch and logits respectively. Our experimental findings, based on our synthesized dataset and two additional open-source datasets, indicate that: 1) Large-scale models perform well when provided with user context but struggle in the absence of such context. 2) While specialized smaller models fine-tuned on the synthetic dataset show promise, they still lag behind their larger counterparts. 3) Our CoGenesis framework, utilizing mixed-scale models, showcases competitive performance, providing a feasible solution to privacy issues.
Transferable Embedding Inversion Attack: Uncovering Privacy Risks in Text Embeddings without Model Queries
PDF ↗This study investigates the privacy risks associated with text embeddings, focusing on the scenario where attackers cannot access the original embedding model. Contrary to previous research requiring direct model access, we explore a more realistic threat model by developing a transfer attack method. This approach uses a surrogate model to mimic the victim model’s behavior, allowing the attacker to infer sensitive information from text embeddings without direct access. Our experiments across various embedding models and a clinical dataset demonstrate that our transfer attack significantly outperforms traditional methods, revealing the potential privacy vulnerabilities in embedding technologies and emphasizing the need for enhanced security measures.
MathGenie: Generating Synthetic Data with Question Back-translation for Enhancing Mathematical Reasoning of LLMs
PDF ↗Large language models (LLMs) have exhibited great potential in mathematical reasoning. However, there remains a performance gap in this area between existing open-source models and closed-source models such as GPT-4. In this paper, we introduce MathGenie, a novel method for generating diverse and reliable math problems by leveraging the ground-truth solutions of the seed data. We augment these ground-truth solutions and use a specially finetuned model to translate these augmented solutions back into new questions. Subsequently, we generate code-integrated solutions for these questions. To ensure the correctness of the code-integrated solutions, we employ rationale-based verification for filtering. Then, we finetune various pretrained models, ranging from 7B to 70B, on the newly curated data, resulting in a family of models known as MathGenie. These models consistently outperform previous open-source models across five representative mathematical reasoning datasets, achieving state-of-the-art performance. In particular, MathGenie-InternLM2 achieves an accuracy of 87.7% on GSM8K and 55.7% on MATH, securing the best overall score.
RLHFPoison: Reward Poisoning Attack for Reinforcement Learning with Human Feedback in Large Language Models
PDF ↗Reinforcement Learning with Human Feedback (RLHF) is a methodology designed to align Large Language Models (LLMs) with human preferences, playing an important role in LLMs alignment. Despite its advantages, RLHF relies on human annotators to rank the text, which can introduce potential security vulnerabilities if any adversarial annotator (i.e., attackers) manipulates the ranking score by up-ranking any malicious text to steer the LLM adversarially. To assess the red-teaming of RLHF against human preference data poisoning, we propose RankPoison, a poisoning attack method on candidates’ selection of preference rank flipping to reach certain malicious behaviors (e.g., generating longer sequences, which can increase the computational cost). With poisoned dataset generated by RankPoison, we can perform poisoning attacks on LLMs to generate longer tokens without hurting the original safety alignment performance. Moreover, applying RankPoison, we also successfully implement a backdoor attack where LLMs can generate longer answers under questions with the trigger word. Our findings highlight critical security challenges in RLHF, underscoring the necessity for more robust alignment methods for LLMs.
Deep Learning · Large Language Models
Large language models (LLMs) have gained considerable attention for their excellent natural language processing capabilities. Nonetheless, these LLMs present many challenges, particularly in the realm of trustworthiness. This paper introduces TrustLLM, a comprehensive study of trustworthiness in LLMs, including principles for different dimensions of trustworthiness, established benchmark, evaluation, and analysis of trustworthiness for mainstream LLMs, and discussion of open challenges and future directions. Specifically, we first propose a set of principles for trustworthy LLMs that span eight different dimensions. Based on these principles, we further establish a benchmark across six dimensions including truthfulness, safety, fairness, robustness, privacy, and machine ethics. We then present a study evaluating 16 mainstream LLMs in TrustLLM, consisting of over 30 datasets. Our findings firstly show that in general trustworthiness and capability (i.e., functional effectiveness) are positively related. Secondly, our observations reveal that proprietary LLMs generally outperform most open-source counterparts in terms of trustworthiness, raising concerns about the potential risks of widely accessible open-source LLMs. However, a few open-source LLMs come very close to proprietary ones, suggesting that open-source models can achieve high levels of trustworthiness without additional mechanisms like *moderator*, offering valuable insights for developers in this field. Thirdly, it is important to note that some LLMs may be overly calibrated towards exhibiting trustworthiness, to the extent that they compromise their utility by mistakenly treating benign prompts as harmful and consequently not responding. Besides these observations, we've uncovered key insights into the multifaceted trustworthiness in LLMs. We emphasize the importance of ensuring transparency not only in the models themselves but also in the technologies that underpin trustworthiness. We advocate that the establishment of an AI alliance between industry, academia, the open-source community to foster collaboration is imperative to advance the trustworthiness of LLMs.
Social Aspects · Trustworthy Machine Learning
In this paper, we introduce a novel class of fast, beam search-based adversarial attack (BEAST) for Language Models (LMs). BEAST employs interpretable parameters, enabling attackers to balance between attack speed, success rate, and the readability of adversarial prompts. The computational efficiency of BEAST facilitates us to investigate its applications on LMs for jailbreaking, eliciting hallucinations, and privacy attacks. Our gradient-free targeted attack can jailbreak aligned LMs with high attack success rates within one minute. For instance, BEAST can jailbreak Vicuna-7B-v1.5 under one minute with a success rate of 89% when compared to a gradient-based baseline that takes over an hour to achieve 70% success rate using a single Nvidia RTX A6000 48GB GPU. BEAST can also generate adversarial suffixes for successful jailbreaks that can transfer to unseen prompts and unseen models such as GPT-4-Turbo. Additionally, we discover a unique outcome wherein our untargeted attack induces hallucinations in LM chatbots. Through human evaluations, we find that our untargeted attack causes Vicuna-7B-v1.5 to produce $\sim$15% more incorrect outputs when compared to LM outputs in the absence of our attack. We also learn that 22% of the time, BEAST causes Vicuna to generate outputs that are not relevant to the original prompt. Further, we use BEAST to generate adversarial prompts in a few seconds that can boost the performance of existing membership inference attacks for LMs. We believe that our fast attack, BEAST, has the potential to accelerate research in LM security and privacy.
Social Aspects · Fairness, Equity, Justice and Safety
Large language models (LLMs) have exhibited impressive capabilities in comprehending complex instructions. However, their blind adherence to provided instructions has led to concerns regarding risks of malicious use. Existing defence mechanisms, such as model fine-tuning or output censorship methods have proven to be fallible at ensuring that LLMs do not return semantically impermissible responses. We present fundamental limitations of verifying the semantic properties of LLM outputs and identifying compositional threats, illustrating inherent challenges of current approaches to censoring LLM outputs. Specifically, we demonstrate that semantic censorship can be perceived as an undecidable problem, and semantic properties of LLM outputs can become impossible to verify when the LLM is capable of providing "encrypted" outputs. We further show challenges of censorship can extend beyond just semantic censorship, as attackers can reconstruct impermissible outputs from a collection of permissible ones. Consequently, we call for a re-evaluation of the problem of censorship and its goals, stressing the need for new definitions and approaches to censorship. In addition, we provide an initial attempt toward achieving this goal through syntactic censorship, drawing from a security perspective to design censorship methods that can provide guarantees.
Deep Learning · Large Language Models
Modern language models (LMs) have gained widespread acceptance in everyday and professional contexts, particularly in programming. An essential procedure enabling this adoption is instruction tuning, which substantially enhances LMs' practical utility by training them to follow user instructions and human preferences. However, existing instruction tuning schemes overlook a crucial aspect: the security of generated code. As a result, even the state-of-the-art instruction-tuned LMs frequently produce unsafe code, posing significant security risks. In this work, we introduce SafeCoder to address this gap. SafeCoder performs security-centric fine-tuning using a diverse and high-quality dataset that we collected using an automated pipeline. We integrate the security fine-tuning with standard instruction tuning, to facilitate a joint optimization of both security and utility. Despite its simplicity, we show that SafeCoder is effective across a variety of popular LMs and datasets. It is able to drastically improve security (by about 30%), while preserving utility.
Deep Learning · Robustness
Black-box finetuning is an emerging interface for adapting state-of-the-art language models to user needs. However, such access may also let malicious actors undermine model safety. To demonstrate the challenge of defending finetuning interfaces, we introduce covert malicious finetuning, a method to compromise model safety via finetuning while evading detection. Our method constructs a malicious dataset where every individual datapoint appears innocuous, but finetuning on the dataset teaches the model to respond to encoded harmful requests with encoded harmful responses. Applied to GPT-4, our method produces a finetuned model that acts on harmful instructions 99% of the time and avoids detection by defense mechanisms such as dataset inspection, safety evaluations, and input/output classifiers. Our findings question whether black-box finetuning access can be secured against sophisticated adversaries.
Deep Learning · Robustness
In guaranteeing the absence of adversarial examples in an instance's neighbourhood, certification mechanisms play an important role in demonstrating neural net robustness. In this paper, we ask if these certifications can compromise the very models they help to protect? Our new *Certification Aware Attack* exploits certifications to produce computationally efficient norm-minimising adversarial examples $74$% more often than comparable attacks, while reducing the median perturbation norm by more than $10$%. While these attacks can be used to assess the tightness of certification bounds, they also highlight that releasing certifications can paradoxically reduce security.
Deep Learning · Graph Neural Networks
Given the critical role of graphs in real-world applications and their high-security requirements, improving the ability of graph neural networks (GNNs) to detect out-of-distribution (OOD) data is an urgent research problem. The recent work GNNSAFE proposes a framework based on the aggregation of negative energy scores that significantly improves the performance of GNNs to detect node-level OOD data. However, our study finds that score aggregation among nodes is susceptible to extreme values due to the unboundedness of the negative energy scores and logit shifts, which severely limits the accuracy of GNNs in detecting node-level OOD data. In this paper, we propose NODESAFE: reducing the generation of extreme scores of nodes by adding two optimization terms that make the negative energy scores bounded and mitigate the logit shift. Experimental results show that our approach dramatically improves the ability of GNNs to detect OOD data at the node level, e.g., in detecting OOD data induced by Structure Manipulation, the metric of FPR95 (lower is better) in scenarios without (with) OOD data exposure are reduced from the current SOTA by 28.4% ( 22.7% ). The code is available via https://github.com/ShenzhiYang2000/NODESAFE-Bounded-and-Uniform-Energy-based-Out-of-distribution-Detection-for-Graphs.
Deep Learning · Graph Neural Networks
Graph Attention Networks (GATs) are designed to provide flexible neighborhood aggregation that assigns weights to neighbors according to their importance. In practice, however, GATs are often unable to switch off task-irrelevant neighborhood aggregation, as we show experimentally and analytically. To address this challenge, we propose GATE, a GAT extension that holds three major advantages: i) It alleviates over-smoothing by addressing its root cause of unnecessary neighborhood aggregation. ii) Similarly to perceptrons, it benefits from higher depth as it can still utilize additional layers for (non-)linear feature transformations in case of (nearly) switched-off neighborhood aggregation. iii) By down-weighting connections to unrelated neighbors, it often outperforms GATs on real-world heterophilic datasets. To further validate our claims, we construct a synthetic test bed to analyze a model's ability to utilize the appropriate amount of neighborhood aggregation, which could be of independent interest.
Social Aspects · Privacy-preserving Statistics and Machine Learning
Data is the lifeblood of the modern world, forming a fundamental part of AI, decision-making, and research advances. With increase in interest in data, governments have taken important steps towards a regulated data world, drastically impacting data sharing and data usability and resulting in massive amounts of data confined within the walls of organizations. While synthetic data generation (SDG) is an appealing solution to break down these walls and enable data sharing, the main drawback of existing solutions is the assumption of a trusted aggregator for generative model training. Given that many data holders may not want to, or be legally allowed to, entrust a central entity with their raw data, we propose a framework for collaborative and private generation of synthetic tabular data from distributed data holders. Our solution is general, applicable to any marginal-based SDG, and provides input privacy by replacing the trusted aggregator with secure multi-party computation (MPC) protocols and output privacy via differential privacy (DP). We demonstrate the applicability and scalability of our approach for the state-of-the-art select-measure-generate SDG algorithms MWEM+PGM and AIM.
Social Aspects · Trustworthy Machine Learning
The White House Executive Order on Artificial Intelligence highlights the risks of large language models (LLMs) empowering malicious actors in developing biological, cyber, and chemical weapons. To measure these risks, government institutions and major AI labs are developing evaluations for hazardous capabilities in LLMs. However, current evaluations are private and restricted to a narrow range of malicious use scenarios, which limits further research into reducing malicious use. To fill these gaps, we release the Weapons of Mass Destruction Proxy (WMDP) benchmark, a dataset of 3,668 multiple-choice questions that serve as a proxy measurement of hazardous knowledge in biosecurity, cybersecurity, and chemical security. To guide progress on unlearning, we develop RMU, a state-of-the-art unlearning method based on controlling model representations. RMU reduces model performance on WMDP while maintaining general capabilities in areas such as biology and computer science, suggesting that unlearning may be a concrete path towards reducing malicious use from LLMs. We release our benchmark and code publicly at https://wmdp.ai.