arXiv subjects

Michael Backes

Publications and source records attributed to Michael Backes.

At least 91 records · Page 5Linked to original sources

arXiv2024

Breaking Agents: Compromising Autonomous LLM Agents Through Malfunction Amplification

Recently, autonomous agents built on large language models (LLMs) have experienced significant development and are being deployed in real-world applications. These agents can extend the base LLM's capabilities in multiple ways. For example, a well-built agent using GPT-3.5-Turbo as its core can outperform the more advanced GPT-4 model by leveraging external components. More importantly, the usage of tools enables these systems to perform actions in the real world, moving from merely generating text to actively interacting with their environment. Given the agents' practical applications and their ability to execute consequential actions, it is crucial to assess potential vulnerabilities. Such autonomous systems can cause more severe damage than a standalone language model if compromised. While some existing research has explored harmful actions by LLM agents, our study approaches the vulnerability from a different perspective. We introduce a new type of attack that causes malfunctions by misleading the agent into executing repetitive or irrelevant actions. We conduct comprehensive evaluations using various attack methods, surfaces, and properties to pinpoint areas of susceptibility. Our experiments reveal that these attacks can induce failure rates exceeding 80\% in multiple scenarios. Through attacks on implemented and deployable agents in multi-agent scenarios, we accentuate the realistic risks associated with these vulnerabilities. To mitigate such attacks, we propose self-examination detection methods. However, our findings indicate these attacks are difficult to detect effectively using LLMs alone, highlighting the substantial risks associated with this vulnerability.

cs.CR↗

arXiv2024

ICLGuard: Controlling In-Context Learning Behavior for Applicability Authorization

In-context learning (ICL) is a recent advancement in the capabilities of large language models (LLMs). This feature allows users to perform a new task without updating the model. Concretely, users can address tasks during the inference time by conditioning on a few input-label pair demonstrations along with the test input. It is different than the conventional fine-tuning paradigm and offers more flexibility. However, this capability also introduces potential issues. For example, users may use the model on any data without restriction, such as performing tasks with improper or sensitive content, which might violate the model policy or conflict with the model owner's interests. As a model owner, it is crucial to establish a mechanism to control the model's behavior under ICL, depending on the model owner's requirements for various content. To this end, we introduce the concept of "applicability authorization" tailored for LLMs, particularly for ICL behavior, and propose a simple approach, ICLGuard. It is a fine-tuning framework designed to allow the model owner to regulate ICL behavior on different data. ICLGuard preserves the original LLM and fine-tunes only a minimal set of additional trainable parameters to "guard" the LLM. Empirical results show that the guarded LLM can deactivate its ICL ability on target data without affecting its ICL ability on other data and its general functionality across all data.

cs.CR↗

arXiv2024

SOS! Soft Prompt Attack Against Open-Source Large Language Models

Open-source large language models (LLMs) have become increasingly popular among both the general public and industry, as they can be customized, fine-tuned, and freely used. However, some open-source LLMs require approval before usage, which has led to third parties publishing their own easily accessible versions. Similarly, third parties have been publishing fine-tuned or quantized variants of these LLMs. These versions are particularly appealing to users because of their ease of access and reduced computational resource demands. This trend has increased the risk of training time attacks, compromising the integrity and security of LLMs. In this work, we present a new training time attack, SOS, which is designed to be low in computational demand and does not require clean data or modification of the model weights, thereby maintaining the model's utility intact. The attack addresses security issues in various scenarios, including the backdoor attack, jailbreak attack, and prompt stealing attack. Our experimental findings demonstrate that the proposed attack is effective across all evaluated targets. Furthermore, we present the other side of our SOS technique, namely the copyright token -- a novel technique that enables users to mark their copyrighted content and prevent models from using it.

cs.CR↗

arXiv2024

Memorization in Self-Supervised Learning Improves Downstream Generalization

Self-supervised learning (SSL) has recently received significant attention due to its ability to train high-performance encoders purely on unlabeled data-often scraped from the internet. This data can still be sensitive and empirical evidence suggests that SSL encoders memorize private information of their training data and can disclose them at inference time. Since existing theoretical definitions of memorization from supervised learning rely on labels, they do not transfer to SSL. To address this gap, we propose SSLMem, a framework for defining memorization within SSL. Our definition compares the difference in alignment of representations for data points and their augmented views returned by both encoders that were trained on these data points and encoders that were not. Through comprehensive empirical analysis on diverse encoder architectures and datasets we highlight that even though SSL relies on large datasets and strong augmentations-both known in supervised learning as regularization techniques that reduce overfitting-still significant fractions of training data points experience high memorization. Through our empirical results, we show that this memorization is essential for encoders to achieve higher generalization performance on different downstream tasks.

cs.LG↗

arXiv2024

Transferable Availability Poisoning Attacks

We consider availability data poisoning attacks, where an adversary aims to degrade the overall test accuracy of a machine learning model by crafting small perturbations to its training data. Existing poisoning strategies can achieve the attack goal but assume the victim to employ the same learning method as what the adversary uses to mount the attack. In this paper, we argue that this assumption is strong, since the victim may choose any learning algorithm to train the model as long as it can achieve some targeted performance on clean data. Empirically, we observe a large decrease in the effectiveness of prior poisoning attacks if the victim employs an alternative learning algorithm. To enhance the attack transferability, we propose Transferable Poisoning, which first leverages the intrinsic characteristics of alignment and uniformity to enable better unlearnability within contrastive learning, and then iteratively utilizes the gradient information from supervised and unsupervised contrastive learning paradigms to generate the poisoning perturbations. Through extensive experiments on image benchmarks, we show that our transferable poisoning attack can produce poisoned samples with significantly improved transferability, not only applicable to the two learners used to devise the attack but also to learning algorithms and even paradigms beyond.

cs.CR↗

arXiv2024

Voice Jailbreak Attacks Against GPT-4o

Recently, the concept of artificial assistants has evolved from science fiction into real-world applications. GPT-4o, the newest multimodal large language model (MLLM) across audio, vision, and text, has further blurred the line between fiction and reality by enabling more natural human-computer interactions. However, the advent of GPT-4o's voice mode may also introduce a new attack surface. In this paper, we present the first systematic measurement of jailbreak attacks against the voice mode of GPT-4o. We show that GPT-4o demonstrates good resistance to forbidden questions and text jailbreak prompts when directly transferring them to voice mode. This resistance is primarily due to GPT-4o's internal safeguards and the difficulty of adapting text jailbreak prompts to voice mode. Inspired by GPT-4o's human-like behaviors, we propose VoiceJailbreak, a novel voice jailbreak attack that humanizes GPT-4o and attempts to persuade it through fictional storytelling (setting, character, and plot). VoiceJailbreak is capable of generating simple, audible, yet effective jailbreak prompts, which significantly increases the average attack success rate (ASR) from 0.033 to 0.778 in six forbidden scenarios. We also conduct extensive experiments to explore the impacts of interaction steps, key elements of fictional writing, and different languages on VoiceJailbreak's effectiveness and further enhance the attack performance with advanced fictional writing techniques. We hope our study can assist the research community in building more secure and well-regulated MLLMs.

cs.CR↗

arXiv2024

Instruction Backdoor Attacks Against Customized LLMs

The increasing demand for customized Large Language Models (LLMs) has led to the development of solutions like GPTs. These solutions facilitate tailored LLM creation via natural language prompts without coding. However, the trustworthiness of third-party custom versions of LLMs remains an essential concern. In this paper, we propose the first instruction backdoor attacks against applications integrated with untrusted customized LLMs (e.g., GPTs). Specifically, these attacks embed the backdoor into the custom version of LLMs by designing prompts with backdoor instructions, outputting the attacker's desired result when inputs contain the pre-defined triggers. Our attack includes 3 levels of attacks: word-level, syntax-level, and semantic-level, which adopt different types of triggers with progressive stealthiness. We stress that our attacks do not require fine-tuning or any modification to the backend LLMs, adhering strictly to GPTs development guidelines. We conduct extensive experiments on 6 prominent LLMs and 5 benchmark text classification datasets. The results show that our instruction backdoor attacks achieve the desired attack performance without compromising utility. Additionally, we propose two defense strategies and demonstrate their effectiveness in reducing such attacks. Our findings highlight the vulnerability and the potential risks of LLM customization such as GPTs.

cs.CR↗

arXiv2024

"Do Anything Now": Characterizing and Evaluating In-The-Wild Jailbreak Prompts on Large Language Models

The misuse of large language models (LLMs) has drawn significant attention from the general public and LLM vendors. One particular type of adversarial prompt, known as jailbreak prompt, has emerged as the main attack vector to bypass the safeguards and elicit harmful content from LLMs. In this paper, employing our new framework JailbreakHub, we conduct a comprehensive analysis of 1,405 jailbreak prompts spanning from December 2022 to December 2023. We identify 131 jailbreak communities and discover unique characteristics of jailbreak prompts and their major attack strategies, such as prompt injection and privilege escalation. We also observe that jailbreak prompts increasingly shift from online Web communities to prompt-aggregation websites and 28 user accounts have consistently optimized jailbreak prompts over 100 days. To assess the potential harm caused by jailbreak prompts, we create a question set comprising 107,250 samples across 13 forbidden scenarios. Leveraging this dataset, our experiments on six popular LLMs show that their safeguards cannot adequately defend jailbreak prompts in all scenarios. Particularly, we identify five highly effective jailbreak prompts that achieve 0.95 attack success rates on ChatGPT (GPT-3.5) and GPT-4, and the earliest one has persisted online for over 240 days. We hope that our study can facilitate the research community and LLM vendors in promoting safer and regulated LLMs.

cs.CR↗

arXiv2024

Link Stealing Attacks Against Inductive Graph Neural Networks

A graph neural network (GNN) is a type of neural network that is specifically designed to process graph-structured data. Typically, GNNs can be implemented in two settings, including the transductive setting and the inductive setting. In the transductive setting, the trained model can only predict the labels of nodes that were observed at the training time. In the inductive setting, the trained model can be generalized to new nodes/graphs. Due to its flexibility, the inductive setting is the most popular GNN setting at the moment. Previous work has shown that transductive GNNs are vulnerable to a series of privacy attacks. However, a comprehensive privacy analysis of inductive GNN models is still missing. This paper fills the gap by conducting a systematic privacy analysis of inductive GNNs through the lens of link stealing attacks, one of the most popular attacks that are specifically designed for GNNs. We propose two types of link stealing attacks, i.e., posterior-only attacks and combined attacks. We define threat models of the posterior-only attacks with respect to node topology and the combined attacks by considering combinations of posteriors, node attributes, and graph features. Extensive evaluation on six real-world datasets demonstrates that inductive GNNs leak rich information that enables link stealing attacks with advantageous properties. Even attacks with no knowledge about graph structures can be effective. We also show that our attacks are robust to different node similarities and different graph features. As a counterpart, we investigate two possible defenses and discover they are ineffective against our attacks, which calls for more effective defenses.

cs.CR↗

arXiv2024

Prompt Stealing Attacks Against Text-to-Image Generation Models

Text-to-Image generation models have revolutionized the artwork design process and enabled anyone to create high-quality images by entering text descriptions called prompts. Creating a high-quality prompt that consists of a subject and several modifiers can be time-consuming and costly. In consequence, a trend of trading high-quality prompts on specialized marketplaces has emerged. In this paper, we perform the first study on understanding the threat of a novel attack, namely prompt stealing attack, which aims to steal prompts from generated images by text-to-image generation models. Successful prompt stealing attacks directly violate the intellectual property of prompt engineers and jeopardize the business model of prompt marketplaces. We first perform a systematic analysis on a dataset collected by ourselves and show that a successful prompt stealing attack should consider a prompt's subject as well as its modifiers. Based on this observation, we propose a simple yet effective prompt stealing attack, PromptStealer. It consists of two modules: a subject generator trained to infer the subject and a modifier detector for identifying the modifiers within the generated image. Experimental results demonstrate that PromptStealer is superior over three baseline methods, both quantitatively and qualitatively. We also make some initial attempts to defend PromptStealer. In general, our study uncovers a new attack vector within the ecosystem established by the popular text-to-image generation models. We hope our results can contribute to understanding and mitigating this emerging threat.

cs.CR↗

arXiv2024

Composite Backdoor Attacks Against Large Language Models

Large language models (LLMs) have demonstrated superior performance compared to previous methods on various tasks, and often serve as the foundation models for many researches and services. However, the untrustworthy third-party LLMs may covertly introduce vulnerabilities for downstream tasks. In this paper, we explore the vulnerability of LLMs through the lens of backdoor attacks. Different from existing backdoor attacks against LLMs, ours scatters multiple trigger keys in different prompt components. Such a Composite Backdoor Attack (CBA) is shown to be stealthier than implanting the same multiple trigger keys in only a single component. CBA ensures that the backdoor is activated only when all trigger keys appear. Our experiments demonstrate that CBA is effective in both natural language processing (NLP) and multimodal tasks. For instance, with $3\%$ poisoning samples against the LLaMA-7B model on the Emotion dataset, our attack achieves a $100\%$ Attack Success Rate (ASR) with a False Triggered Rate (FTR) below $2.06\%$ and negligible model accuracy degradation. Our work highlights the necessity of increased security research on the trustworthiness of foundation LLMs.

cs.CR↗

arXiv2024

Efficient Data-Free Model Stealing with Label Diversity

Machine learning as a Service (MLaaS) allows users to query the machine learning model in an API manner, which provides an opportunity for users to enjoy the benefits brought by the high-performance model trained on valuable data. This interface boosts the proliferation of machine learning based applications, while on the other hand, it introduces the attack surface for model stealing attacks. Existing model stealing attacks have relaxed their attack assumptions to the data-free setting, while keeping the effectiveness. However, these methods are complex and consist of several components, which obscure the core on which the attack really depends. In this paper, we revisit the model stealing problem from a diversity perspective and demonstrate that keeping the generated data samples more diverse across all the classes is the critical point for improving the attack performance. Based on this conjecture, we provide a simplified attack framework. We empirically signify our conjecture by evaluating the effectiveness of our attack, and experimental results show that our approach is able to achieve comparable or even better performance compared with the state-of-the-art method. Furthermore, benefiting from the absence of redundant components, our method demonstrates its advantages in attack efficiency and query budget.

cs.CR↗

arXiv2024

Chasing Gravitational Waves with the Cherenkov Telescope Array

The detection of gravitational waves from a binary neutron star merger by Advanced LIGO and Advanced Virgo (GW170817), along with the discovery of the electromagnetic counterparts of this gravitational wave event, ushered in a new era of multimessenger astronomy, providing the first direct evidence that BNS mergers are progenitors of short gamma-ray bursts (GRBs). Such events may also produce very-high-energy (VHE, > 100GeV) photons which have yet to be detected in coincidence with a gravitational wave signal. The Cherenkov Telescope Array (CTA) is a next-generation VHE observatory which aims to be indispensable in this search, with an unparalleled sensitivity and ability to slew anywhere on the sky within a few tens of seconds. New observing modes and follow-up strategies are being developed for CTA to rapidly cover localization areas of gravitational wave events that are typically larger than the CTA field of view. This work will evaluate and provide estimations on the expected number of of gravitational wave events that will be observable with CTA, considering both on- and off-axis emission. In addition, we will present and discuss the prospects of potential follow-up strategies with CTA.

astro-ph.HE↗

arXiv2024

MGTBench: Benchmarking Machine-Generated Text Detection

Nowadays, powerful large language models (LLMs) such as ChatGPT have demonstrated revolutionary power in a variety of tasks. Consequently, the detection of machine-generated texts (MGTs) is becoming increasingly crucial as LLMs become more advanced and prevalent. These models have the ability to generate human-like language, making it challenging to discern whether a text is authored by a human or a machine. This raises concerns regarding authenticity, accountability, and potential bias. However, existing methods for detecting MGTs are evaluated using different model architectures, datasets, and experimental settings, resulting in a lack of a comprehensive evaluation framework that encompasses various methodologies. Furthermore, it remains unclear how existing detection methods would perform against powerful LLMs. In this paper, we fill this gap by proposing the first benchmark framework for MGT detection against powerful LLMs, named MGTBench. Extensive evaluations on public datasets with curated texts generated by various powerful LLMs such as ChatGPT-turbo and Claude demonstrate the effectiveness of different detection methods. Our ablation study shows that a larger number of words in general leads to better performance and most detection methods can achieve similar performance with much fewer training samples. Moreover, we delve into a more challenging task: text attribution. Our findings indicate that the model-based detection methods still perform well in the text attribution task. To investigate the robustness of different detection methods, we consider three adversarial attacks, namely paraphrasing, random spacing, and adversarial perturbations. We discover that these attacks can significantly diminish detection effectiveness, underscoring the critical need for the development of more robust detection methods.

cs.CR↗

arXiv2023

FAKEPCD: Fake Point Cloud Detection via Source Attribution

To prevent the mischievous use of synthetic (fake) point clouds produced by generative models, we pioneer the study of detecting point cloud authenticity and attributing them to their sources. We propose an attribution framework, FAKEPCD, to attribute (fake) point clouds to their respective generative models (or real-world collections). The main idea of FAKEPCD is to train an attribution model that learns the point cloud features from different sources and further differentiates these sources using an attribution signal. Depending on the characteristics of the training point clouds, namely, sources and shapes, we formulate four attribution scenarios: close-world, open-world, single-shape, and multiple-shape, and evaluate FAKEPCD's performance in each scenario. Extensive experimental results demonstrate the effectiveness of FAKEPCD on source attribution across different scenarios. Take the open-world attribution as an example, FAKEPCD attributes point clouds to known sources with an accuracy of 0.82-0.98 and to unknown sources with an accuracy of 0.73-1.00. Additionally, we introduce an approach to visualize unique patterns (fingerprints) in point clouds associated with each source. This explains how FAKEPCD recognizes point clouds from various sources by focusing on distinct areas within them. Overall, we hope our study establishes a baseline for the source attribution of (fake) point clouds.

cs.CR↗

arXiv2023

Investigating the Lorentz Invariance Violation effect using different cosmological backgrounds

Familiar concepts in physics, such as Lorentz symmetry, are expected to be broken at energies approaching the Planck energy scale as predicted by several quantum-gravity theories. However, such very large energies are unreachable by current experiments on Earth. Current and future Cherenkov telescope facilities may have the capability to measure the accumulated deformation from Lorentz symmetry for photons traveling over large distances via energy-dependent time delays. One of the best natural laboratories to test Lorentz Invariance Violation~(LIV) signatures are Gamma-ray bursts~(GRBs). The calculation of time delays due to the LIV effect depends on the cosmic expansion history. In most of the previous works calculating time lags due to the LIV effect, the standard $Λ$CDM (or concordance) cosmological model is assumed. In this paper, we investigate whether the LIV signature is significantly different when assuming alternatives to the $Λ$CDM cosmological model. Specifically, we consider cosmological models with a non-trivial dark-energy equation of state ($w \neq -1$), such as the standard Chevallier-Polarski-Linder~(CPL) parameterization, the quadratic parameterization of the dark-energy equation of state, and the Pade parameterizations. We find that the relative difference in the predicted time lags is small, of the order of at most a few percent, and thus likely smaller than the systematic errors of possible measurements currently or in the near future.

astro-ph.HE↗

arXiv2023

UMDFood: Vision-language models boost food composition compilation

Nutrition information is crucial in precision nutrition and the food industry. The current food composition compilation paradigm relies on laborious and experience-dependent methods. However, these methods struggle to keep up with the dynamic consumer market, resulting in delayed and incomplete nutrition data. In addition, earlier machine learning methods overlook the information in food ingredient statements or ignore the features of food images. To this end, we propose a novel vision-language model, UMDFood-VL, using front-of-package labeling and product images to accurately estimate food composition profiles. In order to empower model training, we established UMDFood-90k, the most comprehensive multimodal food database to date, containing 89,533 samples, each labeled with image and text-based ingredient descriptions and 11 nutrient annotations. UMDFood-VL achieves the macro-AUCROC up to 0.921 for fat content estimation, which is significantly higher than existing baseline methods and satisfies the practical requirements of food composition compilation. Meanwhile, up to 82.2% of selected products' estimated error between chemical analysis results and model estimation results are less than 10%. This performance sheds light on generalization towards other food and nutrition-related data compilation and catalyzation for the evolution of generative AI-based technology in other food applications that require personalization.

cs.CV↗

arXiv2023

Comprehensive Assessment of Toxicity in ChatGPT

Moderating offensive, hateful, and toxic language has always been an important but challenging topic in the domain of safe use in NLP. The emerging large language models (LLMs), such as ChatGPT, can potentially further accentuate this threat. Previous works have discovered that ChatGPT can generate toxic responses using carefully crafted inputs. However, limited research has been done to systematically examine when ChatGPT generates toxic responses. In this paper, we comprehensively evaluate the toxicity in ChatGPT by utilizing instruction-tuning datasets that closely align with real-world scenarios. Our results show that ChatGPT's toxicity varies based on different properties and settings of the prompts, including tasks, domains, length, and languages. Notably, prompts in creative writing tasks can be 2x more likely than others to elicit toxic responses. Prompting in German and Portuguese can also double the response toxicity. Additionally, we discover that certain deliberately toxic prompts, designed in earlier studies, no longer yield harmful responses. We hope our discoveries can guide model developers to better regulate these AI systems and the users to avoid undesirable outputs.

cs.CY↗