arXiv subjects

Michael Backes

Publications and source records attributed to Michael Backes.

At least 55 records · Page 3Linked to original sources

arXiv2025

JADES: A Universal Framework for Jailbreak Assessment via Decompositional Scoring

Accurately determining whether a jailbreak attempt has succeeded is a fundamental yet unresolved challenge. Existing evaluation methods rely on misaligned proxy indicators or naive holistic judgments. They frequently misinterpret model responses, leading to inconsistent and subjective assessments that misalign with human perception. To address this gap, we introduce JADES (Jailbreak Assessment via Decompositional Scoring), a universal jailbreak evaluation framework. Its key mechanism is to automatically decompose an input harmful question into a set of weighted sub-questions, score each sub-answer, and weight-aggregate the sub-scores into a final decision. JADES also incorporates an optional fact-checking module to strengthen the detection of hallucinations in jailbreak responses. We validate JADES on JailbreakQR, a newly introduced benchmark proposed in this work, consisting of 400 pairs of jailbreak prompts and responses, each meticulously annotated by humans. In a binary setting (success/failure), JADES achieves 98.5% agreement with human evaluators, outperforming strong baselines by over 9%. Re-evaluating five popular attacks on four LLMs reveals substantial overestimation (e.g., LAA's attack success rate on GPT-3.5-Turbo drops from 93% to 69%). Our results show that JADES could deliver accurate, consistent, and interpretable evaluations, providing a reliable basis for measuring future jailbreak attacks.

cs.CR↗

arXiv2025

Hate in Plain Sight: On the Risks of Moderating AI-Generated Hateful Illusions

Recent advances in text-to-image diffusion models have enabled the creation of a new form of digital art: optical illusions--visual tricks that create different perceptions of reality. However, adversaries may misuse such techniques to generate hateful illusions, which embed specific hate messages into harmless scenes and disseminate them across web communities. In this work, we take the first step toward investigating the risks of scalable hateful illusion generation and the potential for bypassing current content moderation models. Specifically, we generate 1,860 optical illusions using Stable Diffusion and ControlNet, conditioned on 62 hate messages. Of these, 1,571 are hateful illusions that successfully embed hate messages, either overtly or subtly, forming the Hateful Illusion dataset. Using this dataset, we evaluate the performance of six moderation classifiers and nine vision language models (VLMs) in identifying hateful illusions. Experimental results reveal significant vulnerabilities in existing moderation models: the detection accuracy falls below 0.245 for moderation classifiers and below 0.102 for VLMs. We further identify a critical limitation in their vision encoders, which mainly focus on surface-level image details while overlooking the secondary layer of information, i.e., hidden messages. To address this risk, we explore preliminary mitigation measures and identify the most effective approaches from the perspectives of image transformations and training-level strategies.

cs.CR↗

arXiv2025

Bridging the Gap in Vision Language Models in Identifying Unsafe Concepts Across Modalities

Vision-language models (VLMs) are increasingly applied to identify unsafe or inappropriate images due to their internal ethical standards and powerful reasoning abilities. However, it is still unclear whether they can recognize various unsafe concepts when presented in different modalities, such as text and images. To address this, we first compile the UnsafeConcepts dataset, featuring 75 unsafe concepts, i.e., ``Swastika,'' ``Sexual Harassment,'' and ``Assaults,'' along with associated 1.5K images. We then conduct a systematic evaluation of VLMs' perception (concept recognition) and alignment (ethical reasoning) capabilities. We assess eight popular VLMs and find that, although most VLMs accurately perceive unsafe concepts, they sometimes mistakenly classify these concepts as safe. We also identify a consistent modality gap among open-source VLMs in distinguishing between visual and textual unsafe concepts. To bridge this gap, we introduce a simplified reinforcement learning (RL)-based approach using proximal policy optimization (PPO) to strengthen the ability to identify unsafe concepts from images. Our approach uses reward scores based directly on VLM responses, bypassing the need for collecting human-annotated preference data to train a new reward model. Experimental results show that our approach effectively enhances VLM alignment on images while preserving general capabilities. It outperforms baselines such as supervised fine-tuning (SFT) and direct preference optimization (DPO). We hope our dataset, evaluation findings, and proposed alignment solution contribute to the community's efforts in advancing safe VLMs.

cs.CR↗

arXiv2025

DivTrackee versus DynTracker: Promoting Diversity in Anti-Facial Recognition against Dynamic FR Strategy

The widespread adoption of facial recognition (FR) models raises serious concerns about their potential misuse, motivating the development of anti-facial recognition (AFR) to protect user facial privacy. In this paper, we argue that the static FR strategy, predominantly adopted in prior literature for evaluating AFR efficacy, cannot faithfully characterize the actual capabilities of determined trackers who aim to track a specific target identity. In particular, we introduce DynTracker, a dynamic FR strategy where the model's gallery database is iteratively updated with newly recognized target identity images. Surprisingly, such a simple approach renders all the existing AFR protections ineffective. To mitigate the privacy threats posed by DynTracker, we advocate for explicitly promoting diversity in the AFR-protected images. We hypothesize that the lack of diversity is the primary cause of the failure of existing AFR methods. Specifically, we develop DivTrackee, a novel method for crafting diverse AFR protections that builds upon a text-guided image generation framework and diversity-promoting adversarial losses. Through comprehensive experiments on various image benchmarks and feature extractors, we demonstrate DynTracker's strength in breaking existing AFR methods and the superiority of DivTrackee in preventing user facial images from being identified by dynamic FR strategies. We believe our work can act as an important initial step towards developing more effective AFR methods for protecting user facial privacy against determined trackers.

cs.CV↗

arXiv2025

Amplifying Machine Learning Attacks Through Strategic Compositions

Machine learning (ML) models are proving to be vulnerable to a variety of attacks that allow the adversary to learn sensitive information, cause mispredictions, and more. While these attacks have been extensively studied, current research predominantly focuses on analyzing each attack type individually. In practice, however, adversaries may employ multiple attack strategies simultaneously rather than relying on a single approach. This prompts a crucial yet underexplored question: When the adversary has multiple attacks at their disposal, are they able to mount or amplify the effect of one attack with another? In this paper, we take the first step in studying the strategic interactions among different attacks, which we define as attack compositions. Specifically, we focus on four well-studied attacks during the model's inference phase: adversarial examples, attribute inference, membership inference, and property inference. To facilitate the study of their interactions, we propose a taxonomy based on three stages of the attack pipeline: preparation, execution, and evaluation. Using this taxonomy, we identify four effective attack compositions, such as property inference assisting attribute inference at its preparation level and adversarial examples assisting property inference at its execution level. We conduct extensive experiments on the attack compositions using three ML model architectures and three benchmark image datasets. Empirical results demonstrate the effectiveness of these four attack compositions. We implement and release a modular reusable toolkit, COAT. Arguably, our work serves as a call for researchers and practitioners to consider advanced adversarial settings involving multiple attack strategies, aiming to strengthen the security and robustness of AI systems.

cs.CR↗

arXiv2025

Efficient and Privacy-Preserving Soft Prompt Transfer for LLMs

Prompting has become a dominant paradigm for adapting large language models (LLMs). While discrete (textual) prompts are widely used for their interpretability, soft (parameter) prompts have recently gained traction in APIs. This is because they can encode information from more training samples while minimizing the user's token usage, leaving more space in the context window for task-specific input. However, soft prompts are tightly coupled to the LLM they are tuned on, limiting their generalization to other LLMs. This constraint is particularly problematic for efficiency and privacy: (1) tuning prompts on each LLM incurs high computational costs, especially as LLMs continue to grow in size. Additionally, (2) when the LLM is hosted externally, soft prompt tuning often requires sharing private data with the LLM provider. For instance, this is the case with the NVIDIA NeMo API. To address these issues, we propose POST (Privacy Of Soft prompt Transfer), a framework that enables private tuning of soft prompts on a small model and subsequently transfers these prompts to a larger LLM. POST uses knowledge distillation to derive a small model directly from the large LLM to improve prompt transferability, tunes the soft prompt locally, optionally with differential privacy guarantees, and transfers it back to the larger LLM using a small public dataset. Our experiments show that POST reduces computational costs, preserves privacy, and effectively transfers high-utility soft prompts.

cs.LG↗

arXiv2025

Excessive Reasoning Attack on Reasoning LLMs

Recent reasoning large language models (LLMs), such as OpenAI o1 and DeepSeek-R1, exhibit strong performance on complex tasks through test-time inference scaling. However, prior studies have shown that these models often incur significant computational costs due to excessive reasoning, such as frequent switching between reasoning trajectories (e.g., underthinking) or redundant reasoning on simple questions (e.g., overthinking). In this work, we expose a novel threat: adversarial inputs can be crafted to exploit excessive reasoning behaviors and substantially increase computational overhead without compromising model utility. Therefore, we propose a novel loss framework consisting of three components: (1) Priority Cross-Entropy Loss, a modification of the standard cross-entropy objective that emphasizes key tokens by leveraging the autoregressive nature of LMs; (2) Excessive Reasoning Loss, which encourages the model to initiate additional reasoning paths during inference; and (3) Delayed Termination Loss, which is designed to extend the reasoning process and defer the generation of final outputs. We optimize and evaluate our attack for the GSM8K and ORCA datasets on DeepSeek-R1-Distill-LLaMA and DeepSeek-R1-Distill-Qwen. Empirical results demonstrate a 3x to 9x increase in reasoning length with comparable utility performance. Furthermore, our crafted adversarial inputs exhibit transferability, inducing computational overhead in o3-mini, o1-mini, DeepSeek-R1, and QWQ models.

cs.CR↗

arXiv2025

Watermarking LLM-Generated Datasets in Downstream Tasks

Large Language Models (LLMs) have experienced rapid advancements, with applications spanning a wide range of fields, including sentiment classification, review generation, and question answering. Due to their efficiency and versatility, researchers and companies increasingly employ LLM-generated data to train their models. However, the inability to track content produced by LLMs poses a significant challenge, potentially leading to copyright infringement for the LLM owners. In this paper, we propose a method for injecting watermarks into LLM-generated datasets, enabling the tracking of downstream tasks to detect whether these datasets were produced using the original LLM. These downstream tasks can be divided into two categories. The first involves using the generated datasets at the input level, commonly for training classification tasks. The other is the output level, where model trainers use LLM-generated content as output for downstream tasks, such as question-answering tasks. We design a comprehensive set of experiments to evaluate both watermark methods. Our results indicate the high effectiveness of our watermark approach. Additionally, regarding model utility, we find that classifiers trained on the generated datasets achieve a test accuracy exceeding 0.900 in many cases, suggesting that the utility of such models remains robust. For the output-level watermark, we observe that the quality of the generated text is comparable to that produced using real-world datasets. Through our research, we aim to advance the protection of LLM copyrights, taking a significant step forward in safeguarding intellectual property in this domain.

cs.CR↗

arXiv2025

Provably Cost-Sensitive Adversarial Defense via Randomized Smoothing

As ML models are increasingly deployed in critical applications, robustness against adversarial perturbations is crucial. While numerous defenses have been proposed to counter such attacks, they typically assume that all adversarial transformations are equally important, an assumption that rarely aligns with real-world applications. To address this, we study the problem of robust learning against adversarial perturbations under cost-sensitive scenarios, where the potential harm of different types of misclassifications is encoded in a cost matrix. Our solution introduces a provably robust learning algorithm to certify and optimize for cost-sensitive robustness, building on the scalable certification framework of randomized smoothing. Specifically, we formalize the definition of cost-sensitive certified radius and propose our novel adaptation of the standard certification algorithm to generate tight robustness certificates tailored to any cost matrix. In addition, we design a robust training method that improves certified cost-sensitive robustness without compromising model accuracy. Extensive experiments on benchmark datasets, including challenging ones unsolvable by existing methods, demonstrate the effectiveness of our certification algorithm and training method across various cost-sensitive scenarios.

cs.LG↗

arXiv2025

SoK: Data Reconstruction Attacks Against Machine Learning Models: Definition, Metrics, and Benchmark

Data reconstruction attacks, which aim to recover the training dataset of a target model with limited access, have gained increasing attention in recent years. However, there is currently no consensus on a formal definition of data reconstruction attacks or appropriate evaluation metrics for measuring their quality. This lack of rigorous definitions and universal metrics has hindered further advancement in this field. In this paper, we address this issue in the vision domain by proposing a unified attack taxonomy and formal definitions of data reconstruction attacks. We first propose a set of quantitative evaluation metrics that consider important criteria such as quantifiability, consistency, precision, and diversity. Additionally, we leverage large language models (LLMs) as a substitute for human judgment, enabling visual evaluation with an emphasis on high-quality reconstructions. Using our proposed taxonomy and metrics, we present a unified framework for systematically evaluating the strengths and limitations of existing attacks and establishing a benchmark for future research. Empirical results, primarily from a memorization perspective, not only validate the effectiveness of our metrics but also offer valuable insights for designing new attacks.

cs.CR↗

arXiv2025

Are We in the AI-Generated Text World Already? Quantifying and Monitoring AIGT on Social Media

Social media platforms are experiencing a growing presence of AI-Generated Texts (AIGTs). However, the misuse of AIGTs could have profound implications for public opinion, such as spreading misinformation and manipulating narratives. Despite its importance, it remains unclear how prevalent AIGTs are on social media. To address this gap, this paper aims to quantify and monitor the AIGTs on online social media platforms. We first collect a dataset (SM-D) with around 2.4M posts from 3 major social media platforms: Medium, Quora, and Reddit. Then, we construct a diverse dataset (AIGTBench) to train and evaluate AIGT detectors. AIGTBench combines popular open-source datasets and our AIGT datasets generated from social media texts by 12 LLMs, serving as a benchmark for evaluating mainstream detectors. With this setup, we identify the best-performing detector (OSM-Det). We then apply OSM-Det to SM-D to track AIGTs across social media platforms from January 2022 to October 2024, using the AI Attribution Rate (AAR) as the metric. Specifically, Medium and Quora exhibit marked increases in AAR, rising from 1.77% to 37.03% and 2.06% to 38.95%, respectively. In contrast, Reddit shows slower growth, with AAR increasing from 1.31% to 2.45% over the same period. Our further analysis indicates that AIGTs on social media differ from human-written texts across several dimensions, including linguistic patterns, topic distributions, engagement levels, and the follower distribution of authors. We envision our analysis and findings on AIGTs in social media can shed light on future research in this domain.

cs.AI↗

arXiv2025

PSGraph: Differentially Private Streaming Graph Synthesis by Considering Temporal Dynamics

Streaming graphs are ubiquitous in daily life, such as evolving social networks and dynamic communication systems. Due to the sensitive information contained in the graph, directly sharing the streaming graphs poses significant privacy risks. Differential privacy, offering strict theoretical guarantees, has emerged as a standard approach for private graph data synthesis. However, existing methods predominantly focus on static graph publishing, neglecting the intrinsic relationship between adjacent graphs, thereby resulting in limited performance in streaming data publishing scenarios. To address this gap, we propose PSGraph, the first differentially private streaming graph synthesis framework that integrates temporal dynamics. PSGraph adaptively adjusts the privacy budget allocation mechanism by analyzing the variations in the current graph compared to the previous one for conserving the privacy budget. Moreover, PSGraph aggregates information across various timestamps and adopts crucial post-processing techniques to enhance the synthetic streaming graphs. We conduct extensive experiments on four real-world datasets under five commonly used metrics. The experimental results demonstrate the superiority of our proposed PSGraph.

cs.CR↗

arXiv2025

When GPT Spills the Tea: Comprehensive Assessment of Knowledge File Leakage in GPTs

Knowledge files have been widely used in large language model (LLM) agents, such as GPTs, to improve response quality. However, concerns about the potential leakage of knowledge files have grown significantly. Existing studies demonstrate that adversarial prompts can induce GPTs to leak knowledge file content. Yet, it remains uncertain whether additional leakage vectors exist, particularly given the complex data flows across clients, servers, and databases in GPTs. In this paper, we present a comprehensive risk assessment of knowledge file leakage, leveraging a novel workflow inspired by Data Security Posture Management (DSPM). Through the analysis of 651,022 GPT metadata, 11,820 flows, and 1,466 responses, we identify five leakage vectors: metadata, GPT initialization, retrieval, sandboxed execution environments, and prompts. These vectors enable adversaries to extract sensitive knowledge file data such as titles, content, types, and sizes. Notably, the activation of the built-in tool Code Interpreter leads to a privilege escalation vulnerability, enabling adversaries to directly download original knowledge files with a 95.95% success rate. Further analysis reveals that 28.80% of leaked files are copyrighted, including digital copies from major publishers and internal materials from a listed company. In the end, we provide actionable solutions for GPT builders and platform providers to secure the GPT data supply chain.

cs.CR↗

arXiv2025

JailbreakRadar: Comprehensive Assessment of Jailbreak Attacks Against LLMs

Jailbreak attacks aim to bypass the LLMs' safeguards. While researchers have proposed different jailbreak attacks in depth, they have done so in isolation -- either with unaligned settings or comparing a limited range of methods. To fill this gap, we present a large-scale evaluation of various jailbreak attacks. We collect 17 representative jailbreak attacks, summarize their features, and establish a novel jailbreak attack taxonomy. Then we conduct comprehensive measurement and ablation studies across nine aligned LLMs on 160 forbidden questions from 16 violation categories. Also, we test jailbreak attacks under eight advanced defenses. Based on our taxonomy and experiments, we identify some important patterns, such as heuristic-based attacks could achieve high attack success rates but are easy to mitigate by defenses, causing low practicality. Our study offers valuable insights for future research on jailbreak attacks and defenses. We hope our work could help the community avoid incremental work and serve as an effective benchmark tool for practitioners.

cs.CR↗

arXiv2025

Captured by Captions: On Memorization and its Mitigation in CLIP Models

Multi-modal models, such as CLIP, have demonstrated strong performance in aligning visual and textual representations, excelling in tasks like image retrieval and zero-shot classification. Despite this success, the mechanisms by which these models utilize training data, particularly the role of memorization, remain unclear. In uni-modal models, both supervised and self-supervised, memorization has been shown to be essential for generalization. However, it is not well understood how these findings would apply to CLIP, which incorporates elements from both supervised learning via captions that provide a supervisory signal similar to labels, and from self-supervised learning via the contrastive objective. To bridge this gap in understanding, we propose a formal definition of memorization in CLIP (CLIPMem) and use it to quantify memorization in CLIP models. Our results indicate that CLIP's memorization behavior falls between the supervised and self-supervised paradigms, with "mis-captioned" samples exhibiting highest levels of memorization. Additionally, we find that the text encoder contributes more to memorization than the image encoder, suggesting that mitigation strategies should focus on the text domain. Building on these insights, we propose multiple strategies to reduce memorization while at the same time improving utility--something that had not been shown before for traditional learning paradigms where reducing memorization typically results in utility decrease.

cs.CV↗

arXiv2025

A comparative analysis of GNSS-inferred precipitable water vapour at the potential sites for the Africa Millimetre Telescope

The Event Horizon Telescope (EHT) is a network of antennas across the globe currently used to image super-massive black holes (SMBHs) at a frequency of 230 GHz. Since the release of the image of M87$^\ast$ in 2019 and, subsequently, that of Sgr A$^\ast$ in 2022 by the EHT collaboration, the focus has shifted to dynamically imaging SMBHs. This has led to a search for potential sites to extend and fill in the gaps within the EHT network. The Gamsberg Mountain and the H.E.S.S. site are both located within the Khomas highlands and have been identified as potential sites for the Africa Millimetre Telescope (AMT). Precipitable water vapour (PWV) in the atmosphere is the main source of opacity and noise from atmospheric emissions when observing at millimetre to sub-millimetre wavelengths. This study aims to establish the PWV content and the atmospheric transmission at 86, 230, and 345 GHz at the AMT potential sites using Global Navigation Satellite System (GNSS) derived PWV data. Results show both sites have potential for observations at 86 and 230 GHz, with 345 GHz possible at the Gamsberg Mountain during winter. The overall median PWV of 14.27 mm and 9.25 mm was calculated at the H.E.S.S. site and the Gamsberg Mountain, respectively. The EHT window had PWV medians of 16.62 mm and 11.20 mm at the H.E.S.S. site and Gamsberg Mountain, respectively. Among the two sites, the Gamsberg Mountain had the lowest PWV conditions, therefore making it the most suitable site for the AMT.

astro-ph.IM↗

arXiv2025

Analysis of the accuracy of GNSS inferred precipitable water vapour against that from a 210 GHz WVR at the H.E.S.S. site

The High Energy Stereoscopic System (H.E.S.S.) site and the Gamsberg Mountain have been identified as potential sites for the Africa Millimetre Telescope (AMT). The AMT is poised to observe at millimetre and possibly at submillimetre wavelengths. At these wavelengths, precipitable water vapour (PWV) in the atmosphere is the main source of opacity during observations and therefore needs to be accurately assessed at the potential sites for the AMT. In order to investigate the PWV conditions for the AMT, identical Global Navigation Satellite System (GNSS) stations were installed and used to assess the PWV at the two potential sites. In this study, the accuracy of those PWV measurements by the GNSS stations was assessed by comparing the H.E.S.S. installed GNSS station PWV measurements to that from a 210 GHz Water Vapour Radiometer (WVR) also installed at the H.E.S.S. site. A correlation of 98% and an offset of 0.34 mm was found between the GNSS station and the 210 GHz WVR PWV data when on-site pressure and the Nevada Geodetic Laboratory (NGL) weighted mean temperature ($\mathrm{T_m}$) were used calculate the GNSS station PWV data. In comparison, the offset reduces to 0.15 mm when on-site derived $\mathrm{T_m}$ and pressure were used to calculate the GNSS station PWV. The results show that the GNSS station with on-site meteorological data can be used with high accuracy to reliably determine the PWV conditions at the H.E.S.S. site.

astro-ph.IM↗

arXiv2025

Do You Even Lift? Strengthening Compiler Security Guarantees Against Spectre Attacks

Mainstream compilers implement different countermeasures to prevent specific classes of speculative execution attacks. Unfortunately, these countermeasures either lack formal guarantees or come with proofs restricted to speculative semantics capturing only a subset of the speculation mechanisms supported by modern CPUs, thereby limiting their practical applicability. Ideally, these security proofs should target a speculative semantics capturing the effects of all speculation mechanisms implemented in modern CPUs. However, this is impractical and requires new secure compilation proofs to support additional speculation mechanisms. In this paper, we address this problem by proposing a novel secure compilation framework that allows lifting the security guarantees provided by Spectre countermeasures from weaker speculative semantics (ignoring some speculation mechanisms) to stronger ones (accounting for the omitted mechanisms) without requiring new secure compilation proofs. Using our lifting framework, we performed the most comprehensive security analysis of Spectre countermeasures implemented in mainstream compilers to date. Our analysis spans 9 different countermeasures against 5 classes of Spectre attacks, which we proved secure against a speculative semantics accounting for five different speculation mechanisms.

cs.PL↗