arXiv subjects

Moshe Sipper

Publications and source records attributed to Moshe Sipper.

At least 19 recordsLinked to original sources

arXiv2026

Through the LENS: Local Geometric Decomposition of Vision-Language Model Representations

Vision-language models (VLMs) process image patches and text tokens in a shared residual stream, but the local geometry through which the two modalities interact remains poorly understood. Most interpretability methods identify global linear directions, which may miss representations that are globally high-dimensional but locally low-dimensional. We introduce LENS (Local Explanation of Neighborhood Subspaces), a method that decomposes VLM activations into local low-rank Gaussian neighborhoods using a Mixture of Factor Analyzers. Applied to LLaVA-1.5-7B and Qwen3-VL-8B, LENS reveals distinct depth-dependent fusion trajectories consistent with each model's fusion mechanism: LLaVA progressively mixes modalities at later layers, whereas Qwen3-VL mixes them early, partially re-segregates them, and recombines them near the output. An automated multimodal labeling pipeline assigns concise semantic descriptions to these neighborhoods. Interpolating activations toward neighborhood centroids causally redirects generation within and across modalities and outperforms difference-in-means and VL-SAE in most evaluated conditions; in one LLaVA vision-to-vision setting, MFA achieves 5.7 times the VL-SAE score. Human evaluation finds MFA steering competitive with prompting and substantially stronger than the other intervention baselines. Finally, the MFA coefficient space improves Qwen3-VL image-to-rendered-text retrieval at the deepest evaluated layer from 14.9% to 48.6% R@1. Ablations show that the reported fusion trajectories are stable across component counts, local ranks, and modality-purity thresholds. These results support local geometric neighborhoods as useful interpretable and causal units for analyzing cross-modal representations in the evaluated VLMs.

cs.AI↗

arXiv2025

Breaking Audio Large Language Models by Attacking Only the Encoder: A Universal Targeted Latent-Space Audio Attack

Audio-language models combine audio encoders with large language models to enable multimodal reasoning, but they also introduce new security vulnerabilities. We propose a universal targeted latent space attack, an encoder-level adversarial attack that manipulates audio latent representations to induce attacker-specified outputs in downstream language generation. Unlike prior waveform-level or input-specific attacks, our approach learns a universal perturbation that generalizes across inputs and speakers and does not require access to the language model. Experiments on Qwen2-Audio-7B-Instruct demonstrate consistently high attack success rates with minimal perceptual distortion, revealing a critical and previously underexplored attack surface at the encoder level of multimodal systems.

cs.SD↗

arXiv2025

Don't Lag, RAG: Training-Free Adversarial Detection Using RAG

Adversarial patch attacks pose a major threat to vision systems by embedding localized perturbations that mislead deep models. Traditional defense methods often require retraining or fine-tuning, making them impractical for real-world deployment. We propose a training-free Visual Retrieval-Augmented Generation (VRAG) framework that integrates Vision-Language Models (VLMs) for adversarial patch detection. By retrieving visually similar patches and images that resemble stored attacks in a continuously expanding database, VRAG performs generative reasoning to identify diverse attack types, all without additional training or fine-tuning. We extensively evaluate open-source large-scale VLMs, including Qwen-VL-Plus, Qwen2.5-VL-72B, and UI-TARS-72B-DPO, alongside Gemini-2.0, a closed-source model. Notably, the open-source UI-TARS-72B-DPO model achieves up to 95 percent classification accuracy, setting a new state-of-the-art for open-source adversarial patch detection. Gemini-2.0 attains the highest overall accuracy, 98 percent, but remains closed-source. Experimental results demonstrate VRAG's effectiveness in identifying a variety of adversarial patches with minimal human annotation, paving the way for robust, practical defenses against evolving adversarial patch attacks.

cs.AI↗

arXiv2025

Pulling Back the Curtain: Unsupervised Adversarial Detection via Contrastive Auxiliary Networks

Deep learning models are widely employed in safety-critical applications yet remain susceptible to adversarial attacks -- imperceptible perturbations that can significantly degrade model performance. Conventional defense mechanisms predominantly focus on either enhancing model robustness or detecting adversarial inputs independently. In this work, we propose an Unsupervised adversarial detection via Contrastive Auxiliary Networks (U-CAN) to uncover adversarial behavior within auxiliary feature representations, without the need for adversarial examples. U-CAN is embedded within selected intermediate layers of the target model. These auxiliary networks, comprising projection layers and ArcFace-based linear layers, refine feature representations to more effectively distinguish between benign and adversarial inputs. Comprehensive experiments across multiple datasets (CIFAR-10, Mammals, and a subset of ImageNet) and architectures (ResNet-50, VGG-16, and ViT) demonstrate that our method surpasses existing unsupervised adversarial detection techniques, achieving superior F1 scores against four distinct attack methods. The proposed framework provides a scalable and effective solution for enhancing the security and reliability of deep learning systems.

cs.CV↗

arXiv2024

On the Robustness of Kolmogorov-Arnold Networks: An Adversarial Perspective

Kolmogorov-Arnold Networks (KANs) have recently emerged as a novel approach to function approximation, demonstrating remarkable potential in various domains. Despite their theoretical promise, the robustness of KANs under adversarial conditions has yet to be thoroughly examined. In this paper we explore the adversarial robustness of KANs, with a particular focus on image classification tasks. We assess the performance of KANs against standard white box and black-box adversarial attacks, comparing their resilience to that of established neural network architectures. Our experimental evaluation encompasses a variety of standard image classification benchmark datasets and investigates both fully connected and convolutional neural network architectures, of three sizes: small, medium, and large. We conclude that small- and medium-sized KANs (either fully connected or convolutional) are not consistently more robust than their standard counterparts, but that large-sized KANs are, by and large, more robust. This comprehensive evaluation of KANs in adversarial scenarios offers the first in-depth analysis of KAN security, laying the groundwork for future research in this emerging field.

cs.CV↗

arXiv2024

Deep Learning-Based Operators for Evolutionary Algorithms

We present two novel domain-independent genetic operators that harness the capabilities of deep learning: a crossover operator for genetic algorithms and a mutation operator for genetic programming. Deep Neural Crossover leverages the capabilities of deep reinforcement learning and an encoder-decoder architecture to select offspring genes. BERT mutation masks multiple gp-tree nodes and then tries to replace these masks with nodes that will most likely improve the individual's fitness. We show the efficacy of both operators through experimentation.

cs.NE↗

arXiv2024

Fortify the Guardian, Not the Treasure: Resilient Adversarial Detectors

This paper presents RADAR-Robust Adversarial Detection via Adversarial Retraining-an approach designed to enhance the robustness of adversarial detectors against adaptive attacks, while maintaining classifier performance. An adaptive attack is one where the attacker is aware of the defenses and adapts their strategy accordingly. Our proposed method leverages adversarial training to reinforce the ability to detect attacks, without compromising clean accuracy. During the training phase, we integrate into the dataset adversarial examples, which were optimized to fool both the classifier and the adversarial detector, enabling the adversarial detector to learn and adapt to potential attack scenarios. Experimental evaluations on the CIFAR-10 and SVHN datasets demonstrate that our proposed algorithm significantly improves a detector's ability to accurately identify adaptive adversarial attacks -- without sacrificing clean accuracy.

cs.CV↗

arXiv2024

XAI-Based Detection of Adversarial Attacks on Deepfake Detectors

We introduce a novel methodology for identifying adversarial attacks on deepfake detectors using eXplainable Artificial Intelligence (XAI). In an era characterized by digital advancement, deepfakes have emerged as a potent tool, creating a demand for efficient detection systems. However, these systems are frequently targeted by adversarial attacks that inhibit their performance. We address this gap, developing a defensible deepfake detector by leveraging the power of XAI. The proposed methodology uses XAI to generate interpretability maps for a given method, providing explicit visualizations of decision-making factors within the AI models. We subsequently employ a pretrained feature extractor that processes both the input image and its corresponding XAI image. The feature embeddings extracted from this process are then used for training a simple yet effective classifier. Our approach contributes not only to the detection of deepfakes but also enhances the understanding of possible adversarial attacks, pinpointing potential vulnerabilities. Furthermore, this approach does not change the performance of the deepfake detector. The paper demonstrates promising results suggesting a potential pathway for future deepfake detection mechanisms. We believe this study will serve as a valuable contribution to the community, sparking much-needed discourse on safeguarding deepfake detectors.

cs.CR↗

arXiv2024

Genetic Programming Theory and Practice: A Fifteen-Year Trajectory

The GPTP workshop series, which began in 2003, has served over the years as a focal meeting for genetic programming (GP) researchers. As such, we think it provides an excellent source for studying the development of GP over the past fifteen years. We thus present herein a trajectory of the thematic developments in the field of GP.

cs.NE↗

arXiv2024

Coevolving Artistic Images Using OMNIREP

We have recently developed OMNIREP, a coevolutionary algorithm to discover both a representation and an interpreter that solve a particular problem of interest. Herein, we demonstrate that the OMNIREP framework can be successfully applied within the field of evolutionary art. Specifically, we coevolve representations that encode image position, alongside interpreters that transform these positions into one of three pre-defined shapes (chunks, polygons, or circles) of varying size, shape, and color. We showcase a sampling of the unique image variations produced by this approach.

cs.NE↗

arXiv2024

New Pathways in Coevolutionary Computation

The simultaneous evolution of two or more species with coupled fitness -- coevolution -- has been put to good use in the field of evolutionary computation. Herein, we present two new forms of coevolutionary algorithms, which we have recently designed and applied with success. OMNIREP is a cooperative coevolutionary algorithm that discovers both a representation and an encoding for solving a particular problem of interest. SAFE is a commensalistic coevolutionary algorithm that maintains two coevolving populations: a population of candidate solutions and a population of candidate objective functions needed to measure solution quality during evolution.

cs.NE↗

arXiv2024

Task and Explanation Network

Explainability in deep networks has gained increased importance in recent years. We argue herein that an AI must be tasked not just with a task but also with an explanation of why said task was accomplished as such. We present a basic framework -- Task and Explanation Network (TENet) -- which fully integrates task completion and its explanation. We believe that the field of AI as a whole should insist -- quite emphatically -- on explainability.

cs.LG↗

arXiv2023

Fitness Approximation through Machine Learning

We present a novel approach to performing fitness approximation in genetic algorithms (GAs) using machine-learning (ML) models, through dynamic adaptation to the evolutionary state. Maintaining a dataset of sampled individuals along with their actual fitness scores, we continually update a fitness-approximation ML model throughout an evolutionary run. We compare different methods for: 1) switching between actual and approximate fitness, 2) sampling the population, and 3) weighting the samples. Experimental findings demonstrate significant improvement in evolutionary runtimes, with fitness scores that are either identical or slightly lower than that of the fully run GA -- depending on the ratio of approximate-to-actual-fitness computation. Although we focus on evolutionary agents in Gymnasium (game) simulators -- where fitness computation is costly -- our approach is generic and can be easily applied to many different domains.

cs.NE↗

arXiv2023

Open Sesame! Universal Black Box Jailbreaking of Large Language Models

Large language models (LLMs), designed to provide helpful and safe responses, often rely on alignment techniques to align with user intent and social guidelines. Unfortunately, this alignment can be exploited by malicious actors seeking to manipulate an LLM's outputs for unintended purposes. In this paper we introduce a novel approach that employs a genetic algorithm (GA) to manipulate LLMs when model architecture and parameters are inaccessible. The GA attack works by optimizing a universal adversarial prompt that -- when combined with a user's query -- disrupts the attacked model's alignment, resulting in unintended and potentially harmful outputs. Our novel approach systematically reveals a model's limitations and vulnerabilities by uncovering instances where its responses deviate from expected behavior. Through extensive experiments we demonstrate the efficacy of our technique, thus contributing to the ongoing discussion on responsible AI development by providing a diagnostic tool for evaluating and enhancing alignment of LLMs with human intent. To our knowledge this is the first automated universal black box jailbreak attack.

cs.CL↗

arXiv2023

I See Dead People: Gray-Box Adversarial Attack on Image-To-Text Models

Modern image-to-text systems typically adopt the encoder-decoder framework, which comprises two main components: an image encoder, responsible for extracting image features, and a transformer-based decoder, used for generating captions. Taking inspiration from the analysis of neural networks' robustness against adversarial perturbations, we propose a novel gray-box algorithm for creating adversarial examples in image-to-text models. Unlike image classification tasks that have a finite set of class labels, finding visually similar adversarial examples in an image-to-text task poses greater challenges because the captioning system allows for a virtually infinite space of possible captions. In this paper, we present a gray-box adversarial attack on image-to-text, both untargeted and targeted. We formulate the process of discovering adversarial perturbations as an optimization problem that uses only the image-encoder component, meaning the proposed attack is language-model agnostic. Through experiments conducted on the ViT-GPT2 model, which is the most-used image-to-text model in Hugging Face, and the Flickr30k dataset, we demonstrate that our proposed attack successfully generates visually similar adversarial examples, both with untargeted and targeted captions. Notably, our attack operates in a gray-box manner, requiring no knowledge about the decoder module. We also show that our attacks fool the popular open-source platform Hugging Face.

cs.CV↗

arXiv2023

A Melting Pot of Evolution and Learning

We survey eight recent works by our group, involving the successful blending of evolutionary algorithms with machine learning and deep learning: 1. Binary and Multinomial Classification through Evolutionary Symbolic Regression, 2. Classy Ensemble: A Novel Ensemble Algorithm for Classification, 3. EC-KitY: Evolutionary Computation Tool Kit in Python, 4. Evolution of Activation Functions for Deep Learning-Based Image Classification, 5. Adaptive Combination of a Genetic Algorithm and Novelty Search for Deep Neuroevolution, 6. An Evolutionary, Gradient-Free, Query-Efficient, Black-Box Algorithm for Generating Adversarial Instances in Deep Networks, 7. Foiling Explanations in Deep Neural Networks, 8. Patch of Invisibility: Naturalistic Black-Box Adversarial Attacks on Object Detectors.

cs.NE↗

arXiv2023

Patch of Invisibility: Naturalistic Physical Black-Box Adversarial Attacks on Object Detectors

Adversarial attacks on deep learning models have received increased attention in recent years. Work in this area has mostly focused on gradient-based techniques, so-called 'white-box' attacks, where the attacker has access to the targeted model's internal parameters; such an assumption is usually untenable in the real world. Additionally, some attacks use the entire pixel space to fool a given model, which is neither practical nor physical. To accommodate these problems we propose the BBNP algorithm (Black-Box Naturalistic Patch): a direct, black-box, naturalistic, gradient-free method that uses the learned image manifold of a pretrained, generative adversarial network (GAN) to generate naturalistic adversarial patches for object detectors. This method performs model-agnostic black-box naturalistic attacks on object detection models by relying solely on the outputs of the model. Comparing our approach against five models, five black-box and two white-box attacks, we show that our proposed method achieves state-of-the-art results, outperforming all other tested black-box approaches.

cs.CV↗

arXiv2023

Classy Ensemble: A Novel Ensemble Algorithm for Classification

We present Classy Ensemble, a novel ensemble-generation algorithm for classification tasks, which aggregates models through a weighted combination of per-class accuracy. Tested over 153 machine learning datasets we demonstrate that Classy Ensemble outperforms two other well-known aggregation algorithms -- order-based pruning and clustering-based pruning -- as well as the recently introduced lexigarden ensemble generator. We then present three enhancements: 1) Classy Cluster Ensemble, which combines Classy Ensemble and cluster-based pruning; 2) Deep Learning experiments, showing the merits of Classy Ensemble over four image datasets: Fashion MNIST, CIFAR10, CIFAR100, and ImageNet; and 3) Classy Evolutionary Ensemble, wherein an evolutionary algorithm is used to select the set of models which Classy Ensemble picks from. This latter, combining learning and evolution, resulted in improved performance on the hardest dataset.

cs.LG↗