Assessing the robustness of multimodal models against adversarial examples is an important aspect for the safety of its users. We craft L0-norm perturbation attacks on the preprocessed input images. We launch them in a black-box setup against four multimodal models and two unimodal DNNs, considering both targeted and untargeted misclassification. Our attacks target less than 0.04% of perturbed image area and integrate different spatial positioning of perturbed pixels: sparse positioning and pixels arranged in different contiguous shapes (row, column, diagonal, and patch). To the best of our knowledge, we are the first to assess the robustness of three state-of-the-art multimodal models (ALIGN, AltCLIP, GroupViT) against different sparse and contiguous pixel distribution perturbations. The obtained results indicate that unimodal DNNs are more robust than multimodal models. Furthermore, models using CNN-based Image Encoder are more vulnerable than models with ViT - for untargeted attacks, we obtain a 99% success rate by perturbing less than 0.02% of the image area.

Deep learning model design, development, and debugging is a process driven by best practices, guidelines, trial-and-error, and the personal experiences of model developers. At multiple stages of this process, performance and internal model data can be logged and made available. However, due to the sheer complexity and scale of this data and process, model developers often resort to evaluating their model performance based on abstract metrics like accuracy and loss. We argue that a structured analysis of data along the model's architecture and at multiple abstraction levels can considerably streamline the debugging process. Such a systematic analysis can further connect the developer's design choices to their impacts on the model behavior, facilitating the understanding, diagnosis, and refinement of deep learning models. Hence, in this paper, we (1) contribute a conceptual framework structuring the data space of deep learning experiments. Our framework, grounded in literature analysis and requirements interviews, captures design dimensions and proposes mechanisms to make this data explorable and tractable. To operationalize our framework in a ready-to-use application, we (2) present the iNNspector system. iNNspector enables tracking of deep learning experiments and provides interactive visualizations of the data on all levels of abstraction from multiple models to individual neurons. Finally, we (3) evaluate our approach with three real-world use-cases and a user study with deep learning developers and data analysts, proving its effectiveness and usability.

Slot attention aims to decompose an input image into a set of meaningful object files (slots). These latent object representations enable various downstream tasks. Yet, these slots often bind to object parts, not objects themselves, especially for real-world datasets. To address this, we introduce Guided Latent Slot Diffusion - GLASS, an object-centric model that uses generated captions as a guiding signal to better align slots with objects. Our key insight is to learn the slot-attention module in the space of generated images. This allows us to repurpose the pre-trained diffusion decoder model, which reconstructs the images from the slots, as a semantic mask generator based on the generated captions. GLASS learns an object-level representation suitable for multiple tasks simultaneously, e.g., segmentation, image generation, and property prediction, outperforming previous methods. For object discovery, GLASS achieves approx. a +35% and +10% relative improvement for mIoU over the previous state-of-the-art (SOTA) method on the VOC and COCO datasets, respectively, and establishes a new SOTA FID score for conditional image generation amongst slot-attention-based methods. For the segmentation task, GLASS surpasses SOTA weakly-supervised and language-based segmentation models, which were specifically designed for the task.

With Vision-Language Pre-training (VLP) models demonstrating powerful multimodal interaction capabilities, the application scenarios of neural networks are no longer confined to unimodal domains but have expanded to more complex multimodal V+L downstream tasks. The security vulnerabilities of unimodal models have been extensively examined, whereas those of VLP models remain challenging. We note that in CV models, the understanding of images comes from annotated information, while VLP models are designed to learn image representations directly from raw text. Motivated by this discrepancy, we developed the Feature Guidance Attack (FGA), a novel method that uses text representations to direct the perturbation of clean images, resulting in the generation of adversarial images. FGA is orthogonal to many advanced attack strategies in the unimodal domain, facilitating the direct application of rich research findings from the unimodal to the multimodal scenario. By appropriately introducing text attack into FGA, we construct Feature Guidance with Text Attack (FGA-T). Through the interaction of attacking two modalities, FGA-T achieves superior attack effects against VLP models. Moreover, incorporating data augmentation and momentum mechanisms significantly improves the black-box transferability of FGA-T. Our method demonstrates stable and effective attack capabilities across various datasets, downstream tasks, and both black-box and white-box settings, offering a unified baseline for exploring the robustness of VLP models.

We present an overview of the FIGNEWS shared task, organized as part of the ArabicNLP 2024 conference co-located with ACL 2024. The shared task addresses bias and propaganda annotation in multilingual news posts. We focus on the early days of the Israel War on Gaza as a case study. The task aims to foster collaboration in developing annotation guidelines for subjective tasks by creating frameworks for analyzing diverse narratives highlighting potential bias and propaganda. In a spirit of fostering and encouraging diversity, we address the problem from a multilingual perspective, namely within five languages: English, French, Arabic, Hebrew, and Hindi. A total of 17 teams participated in two annotation subtasks: bias (16 teams) and propaganda (6 teams). The teams competed in four evaluation tracks: guidelines development, annotation quality, annotation quantity, and consistency. Collectively, the teams produced 129,800 data points. Key findings and implications for the field are discussed.

Limited Voting (LV) is an approval-based method for multi-winner elections where all ballots are required to have a same fixed size. While it appears to be used as voting method in corporate governance and has some political applications, to the best of our knowledge, no formal analysis of the rule exists to date. We provide such an analysis here, prompted by a request for advice about this voting rule by a health insurance company in the Netherlands, which uses it to elect its work council. We study conditions under which LV would improve representation over standard approval voting and when it would not. We establish the extent of such an improvement, or lack thereof, both in terms of diversity and proportionality notions. These results help us understand if, and how, LV may be used as a low-effort fix of approval voting in order to enhance representation.

In the era of generative AI, the widespread adoption of Neural Text Generators (NTGs) presents new cybersecurity challenges, particularly within the realms of Digital Forensics and Incident Response (DFIR). These challenges primarily involve the detection and attribution of sources behind advanced attacks like spearphishing and disinformation campaigns. As NTGs evolve, the task of distinguishing between human and NTG-authored texts becomes critically complex. This paper rigorously evaluates the DFIR pipeline tailored for text-based security systems, specifically focusing on the challenges of detecting and attributing authorship of NTG-authored texts. By introducing a novel human-NTG co-authorship text attack, termed CS-ACT, our study uncovers significant vulnerabilities in traditional DFIR methodologies, highlighting discrepancies between ideal scenarios and real-world conditions. Utilizing 14 diverse datasets and 43 unique NTGs, up to the latest GPT-4, our research identifies substantial vulnerabilities in the forensic profiling phase, particularly in attributing authorship to NTGs. Our comprehensive evaluation points to factors such as model sophistication and the lack of distinctive style within NTGs as significant contributors for these vulnerabilities. Our findings underscore the necessity for more sophisticated and adaptable strategies, such as incorporating adversarial learning, stylizing NTGs, and implementing hierarchical attribution through the mapping of NTG lineages to enhance source attribution. This sets the stage for future research and the development of more resilient text-based security systems.

The black box sits on the army truck dashboard like a talisman, its tiny screen lighting up with warnings when Russian drones are above us. We are driving fast along a country road in the darkness near the front lines outside Kharkiv. Like many in this war, the soldiers inside have come to revere the little cube they call "sugar"; it warns of the unseen dangers above. On the vehicle's roof are three mushroom-shaped antennas that make up separate drone-jamming equipment. The car emits an invisible aura of protection that will thwart some, but not all, of the Russian attack drones patrolling the skies above this battlefield.

The US Senate unanimously passed a bill on Tuesday designed to hold accountable those who make or share deepfake porn. The Disrupt Explicit Forged Images and Non-Consensual Edits Act (DEFIANCE Act) would allow victims to sue those who create, share or possess AI-generated sexual images or videos using their likeness. The issue took root in the public consciousness after the infamous Taylor Swift deepfake that circulated among online lowlifes early this year. The bill would let victims sue for up to 150,000 in damages. That number grows to 250,000 if it's related to attempted sexual assault, stalking or harassment.

The Lebanese armed group Hezbollah has broadcast a drone video that it said showed air defence facilities, planes and fuel storage units at Israel's Ramat David airbase, nearly 50km (30 miles) into Israeli territory. The footage included labels pointing out apparent military infrastructure, including the short-range Iron Dome air defence system which is designed to destroy rockets. The video, more than eight minutes long, was mostly filmed on Tuesday, the Iran-aligned group said. The video also included nighttime shots that Hezbollah said were captured "earlier" and other images the group said were taken earlier in July. The caption said it was only "some" of what the drone had captured.