In this research, we address a critical vulnerability of Federated Learning (FL) in practical applications: their susceptibility to adversarial attacks from malicious participants, which can severely degrade a global model’s performance. We propose MADE-PI, a novel defense mechanism designed to identify and exclude malicious clients early in the training process. Our method introduces a pre-aggregation step where the server evaluates each client update using a Proportional-Integral (PI) score. This score is a unique distance-based metric inspired by the PID controller, where a proportional term P measures the immediate deviation of a client’s update, and an integral term I tracks its behavior over time. By analyzing the distribution of these scores, our technique effectively flags and removes malicious contributions. We validate our approach through an empirical study on three datasets representing various applications of FL, including intelligent transportation, medical image analysis, and handwriting recognition. We concentrate on data poisoning as a highly relevant class of attacks. The results demonstrate that MADE-PI outperforms conventional defenses by providing superior malicious client detection and improving both learning efficiency and final model accuracy.
As collaborative robots (cobots) become available in industry, it is crucial to enable non-coders to program cobots easily. For example, in sorting tasks that may require adjustments over time, workers could easily adjust the cobot’s behavior by utilizing learning from demonstrations (LfD) approaches. This paper presents an LfD approach for enabling a cobot to learn sorting tasks by integrating the Gaussian Mixture Model (GMM), validated in Mujoco for policy training. Quantitative analysis shows that the improved BC-GMM achieves the lowest mean squared error (MSE = 0.00207) and highest R-squared score (R2=0.89), outperforming other methods, such as standard BC, diffusion policy, and implicit behavioral cloning (IBC).
In the era of big data, effective feature selection is critical for improving model performance, reducing computational cost, and enhancing interpretability in high-dimensional datasets. This paper introduces a novel filter-based feature selection algorithm, Class-wise Mean Deviation (CMD), which quantifies the discriminative power of features by measuring the absolute deviation of class-wise means from the global mean. Designed for scalability and real-world application, CMD is computationally efficient and particularly effective in handling the severe class imbalance common in large-scale datasets. We evaluate CMD on two large-scale, highly imbalanced benchmark intrusion detection datasets—CICIDS2017 and CICIDS2019—and compare its performance against established filter methods including Variance Threshold, Pearson Correlation, Mutual Information, and Fisher Score. Experimental results demonstrate that CMD consistently achieves competitive or superior classification performance, even with a minimal number of features. These findings highlight CMD’s potential as a robust and interpretable feature selection technique for cybersecurity and other high-dimensional domains.
Due to its low cost and non-invasive nature, electroencephalography (EEG) is widely used in the diagnosis of neurological disorders and in monitoring cognitive states, such as the detection of Alzheimer’s disease (AD). While deep learning methods have shown promise in AD detection from EEG, signal noise and variability across channels and recording durations remain significant challenges. In this study, we propose EDLAD, an ensemble deep learning approach for Alzheimer’s detection that integrates multiple deep learning models, each specialized in different types of EEG-derived features. EDLAD leverages ensemble learning to improve predictive performance by combining the outputs of multiple models. Specifically, three deep learning models are independently trained: a Transformer-based model and two dense network classifiers. The Transformer-based model, called Conformer, which integrates the convolutional and transformer layers, is designed to process the original EEG signals. In parallel, features in the time and frequency domains are extracted and fed into two separate neural networks to complement the learning process and improve the robustness of the final predictions. Experiments conducted on two EEG datasets demonstrate that the proposed ensemble framework, EDLAD, consistently improves the classification accuracy at the segment and subject levels across all window lengths. It outperforms both the single Conformer model and a meta-learning ensemble approach. Furthermore, EDLAD surpasses feature-level fusion methods that simply concatenate feature representations from the three models.
Quantization plays a central role in digital communication by mapping continuous-valued signals to a finite set of levels with minimal distortion. Beyond mean-square error, mutual information between the channel input and the quantizer output provides a powerful metric for signal recovery. However, finding the quantizer that maximizes the mutual information is NP-complete for non-binary inputs. To that end, while not optimal, thresholding schemes, whether single-threshold or multi-threshold, are widely adopted. In this work, we study the parallel stochastic single-threshold quantizer architecture, provide some information-theoretic insights, and introduce a momentum-accelerated gradient ascent algorithm that efficiently tunes a single decision threshold to maximize the mutual information. We demonstrate convergence improvements over exhaustive search and quantify mutual information gains across binary and non-binary input distributions. We also validate our theoretical framework with simulations on the MNIST dataset, demonstrating that increasing the number of parallel quantization branches, i.e., mutual information, significantly improves classification accuracy, especially when quantization thresholds are learned and training data is limited.
The solar wind, a continuous stream of charged particles from the Sun's corona, shapes the heliosphere and impacts space systems near Earth. Variations such as high-speed streams and coronal mass ejections can disrupt satellites, power grids, and communications, making accurate modeling essential for space weather forecasting. While 3D magnetohydrodynamic (MHD) models are used to simulate and investigate these variations in the solar wind, they tend to be computationally expensive, limiting their usefulness in investigating the impacts of boundary condition uncertainty. In this work, we develop a surrogate for steady state solar wind modeling, using a Spherical Fourier Neural Operator (SFNO). We compare our model to a previously developed numerical surrogate for this task called HUX, and we show that the SFNO achieves comparable or better performance across several metrics. Though HUX retains advantages in physical smoothness, this underscores the need for improved evaluation criteria rather than a flaw in SFNO. As a flexible and trainable approach, SFNO enables efficient real-time forecasting and can improve with more data. The source code and more visual results are available at https://github.com/rezmansouri/solarwind-sfno-velocity.
Causal modeling offers a principled foundation for uncovering stable, invariant relationships in time-series data, thereby improving robustness and generalization under distribution shifts. Yet its potential is underutilized in spatiotemporal Earth observation, where models often depend on purely correlational features that fail to transfer across heterogeneous domains. We propose RIC-TSC, a regionally-informed causal time-series classification framework that embeds lag-aware causal discovery directly into sequence modeling, enabling both predictive accuracy and scientific interpretability. Using multi-modal satellite and reanalysis data-including Sentinel-1 microwave backscatter, Sentinel-2 and Landsat-8 optical reflectance, and CARRA meteorological variables-we leverage Joint PCMCI+ (J-PCMCI+) to identify region-specific and invariant predictors of supraglacial lake evolution in Greenland. Causal graphs are estimated globally and per basin, with validated predictors and their time lags supplied to lightweight classifiers. On a balanced benchmark of 1000 manually labeled lakes from two contrasting melt seasons (2018-2019), causal models achieve up to 12.59
We propose SAM2LoRA, a parameter-efficient fine-tuning strategy that adapts the Segment Anything Model 2 (SAM2) for fundus image segmentation. SAM2 employs a masked autoencoder-pretrained Hierarchical Vision Transformer for multi-scale feature decoding, enabling rapid inference in low-resource settings; however, fine-tuning remains challenging. To address this, SAM2LoRA integrates a low-rank adapter into both the image encoder and mask decoder, requiring fewer than 5% of the original trainable parameters. Our analysis indicates that for cross-dataset fundus segmentation tasks, a composite loss function combining segmentationBCE, SoftDice, and FocalTversky losses is essential for optimal network tuning. Evaluated on 11 challenging fundus segmentation datasets, SAM2LoRA demonstrates high performance in both blood vessel and optic disc segmentation under cross-dataset training conditions. It achieves Dice scores of up to 0.86 and 0.93 for blood vessel and optic disc segmentation, respectively, and AUC values of up to 0.98 and 0.99, achieving state-of-the-art performance while substantially reducing training overhead.
A large language model (LLM) can map a feedback causal fuzzy cognitive map (FCM) into text and then reconstruct the FCM from the text. This explainable AI system approximates an identity map from the FCM to itself and resembles the operation of an autoencoder (AE). Both the encoder and the decoder explain their decisions in contrast to black-box AEs. Humans can read and interpret the encoded text in contrast to the hidden variables and synaptic webs in AEs. The LLM agent approximates the identity map through a sequence of system instructions that does not compare the output to the input. The reconstruction is lossy because it removes weak causal edges or rules while it preserves strong causal edges. The encoder preserves the strong causal edges even when it trades off some details about the FCM to make the text sound more natural.
Vision-Language Models (VLMs) have remarkable abilities in generating multimodal reasoning tasks. However, potential misuse or safety alignment concerns of VLMs have increased significantly due to different categories of attack vectors. Among various attack vectors, recent studies have demonstrated that image-based perturbations are particularly effective in generating harmful outputs. In the literature, many existing techniques have been proposed to jailbreak VLMs, leading to unstable performance and visible perturbations. In this study, we propose Jailbreaking with Loss-guided Image Perturbation (JaiLIP), a jailbreaking attack in the image space that minimizes a joint objective combining the mean squared error (MSE) loss between clean and adversarial image with the models harmful-output loss. We evaluate our proposed method on VLMs using standard toxicity metrics from Perspective API and Detoxify. Experimental results demonstrate that our method generates highly effective and imperceptible adversarial images, outperforming existing methods in producing toxicity. Moreover, we have evaluated our method in the transportation domain to demonstrate the attacks practicality beyond toxic text generation in specific domain. Our findings emphasize the practical challenges of image-based jailbreak attacks and the need for efficient defense mechanisms for VLMs.
Compared to traditional models, agentic AI represents a highly valuable target for potential attackers as they possess privileged access to data sources and API tools, which are traditionally not incorporated into classical agents. Unlike a typical software application residing in a Demilitarized Zone (DMZ), agentic LLMs consciously rely on nondeterministic behavior of the AI (only defining a final goal, leaving the path selection to LLM). This characteristic introduces substantial security risk to both operational security and information security. Most common existing defense mechanism rely on detection of malicious intent and preventing it from reaching the LLM agent, thus protecting against jailbreak attacks such as prompt injection. In this paper, we present an alternative approach, LLMZ+, which moves beyond traditional detection-based approaches by implementing prompt whitelisting. Through this method, only contextually appropriate and safe messages are permitted to interact with the agentic LLM. By leveraging the specificity of context, LLMZ+ guarantees that all exchanges between external users and the LLM conform to predefined use cases and operational boundaries. Our approach streamlines the security framework, enhances its long-term resilience, and reduces the resources required for sustaining LLM information security. Our empirical evaluation demonstrates that LLMZ+ provides strong resilience against the most common jailbreak prompts. At the same time, legitimate business communications are not disrupted, and authorized traffic flows seamlessly between users and the agentic LLM. We measure the effectiveness of approach using false positive and false negative rates, both of which can be reduced to 0 in our experimental setting.
Utilizing offline reinforcement learning (RL) with real-world clinical data is getting increasing attention in AI for healthcare. However, implementation poses significant challenges. Defining direct rewards is difficult, and inverse RL struggles to infer accurate reward functions from expert behavior in complex environments. Offline RL also encounters distributional discrepancies between learned policies and observed human behavior, a critical issue in healthcare applications. To address challenges in applying offline RL to physical activity promotion for older adults at high risk of falls, based on wearable sensor activity monitoring, we introduce Kolmogorov-Arnold Networks and Diffusion Policies for Offline Inverse Reinforcement Learning (KANDI). Specifically, by leveraging the flexible function approximation in Kolmogorov-Arnold Networks, we estimate the reward function by learning free-living environment behavior from low-fall-risk older adults (experts). Additionally, diffusion-based policies within an Actor-Critic framework provide a generative approach for action refinement, enabling controlled exploration and mitigating distributional shift issues in offline RL. We evaluate KANDI using wearable activity monitoring data in a two-arm clinical trial from our Physio-feedback Exercise Program (PEER) study, emphasizing its practical application in a fall-risk intervention program to promote physical activity among older adults. Our analysis identifies the optimal timing for anti-sedentariness interventions tailored to varying levels of fall risk, thereby maximizing daily physical activity. Additionally, we evaluate KANDI on the D4RL benchmark, outperforming the state-of-the-art methods in each domain. These results underscore KANDI's potential to address key challenges in offline RL for healthcare applications, offering an effective solution for the optimal timing and policy for activity promotion intervention strategies.
Continuous human motion understanding remains a core challenge in computer vision due to its high dimensionality and inherent redundancy. Efficient compression and representation are crucial for analyzing complex motion dynamics. In this work, we introduce an adversarially-refined VQ-GAN framework with dense motion tokenization for compressing spatio-temporal heatmaps while preserving the fine-grained traces of human motion. Our approach combines dense motion tokenization with adversarial refinement, which eliminates reconstruction artifacts like motion smearing and temporal misalignment observed in non-adversarial baselines. Our experiments on the CMU Panoptic dataset provide conclusive evidence of our method's superiority, outperforming the dVAE baseline by 9.31
Modern software development pipelines face growing challenges in securing large codebases with extensive dependencies. Static analysis tools like Bandit are effective at vulnerability detection but suffer from high false positives and lack repair capabilities. Large Language Models (LLMs), in contrast, can suggest fixes but often hallucinate changes and lack self-validation. We present SecureFixAgent, a hybrid repair framework integrating Bandit with lightweight local LLMs (<8B parameters) in an iterative detect-repair-validate loop. To improve precision, we apply parameter-efficient LoRA-based fine-tuning on a diverse, curated dataset spanning multiple Python project domains, mitigating dataset bias and reducing unnecessary edits. SecureFixAgent uses Bandit for detection, the LLM for candidate fixes with explanations, and Bandit re-validation for verification, all executed locally to preserve privacy and reduce cloud reliance. Experiments show SecureFixAgent reduces false positives by 10.8
Large Language Models (LLMs) have recently demonstrated strong potential for cybersecurity question answering (QA), supporting decision-making in real-time threat detection and response workflows. However, their substantial computational demands pose significant challenges for deployment on resource-constrained edge devices. Quantization, a widely adopted model compression technique, can alleviate these constraints. Nevertheless, quantization may degrade model accuracy and increase susceptibility to adversarial attacks. Fine-tuning offers a potential means to mitigate these limitations, but its effectiveness when combined with quantization remains insufficiently explored. Hence, it is essential to understand the trade-offs among accuracy, efficiency, and robustness. We propose AQUA-LLM, an evaluation framework designed to benchmark several state-of-the-art small LLMs under four distinct configurations: base, quantized-only, fine-tuned, and fine-tuned combined with quantization, specifically for cybersecurity QA. Our results demonstrate that quantization alone yields the lowest accuracy and robustness despite improving efficiency. In contrast, combining quantization with fine-tuning enhances both LLM robustness and predictive performance, achieving an optimal balance of accuracy, robustness, and efficiency. These findings highlight the critical need for quantization-aware, robustness-preserving fine-tuning methodologies to enable the robust and efficient deployment of LLMs for cybersecurity QA.
3D semantic scene graphs (3DSSG) provide compact structured representations of environments by explicitly modeling objects, attributes, and relationships. While 3DSSGs have shown promise in robotics and embodied AI, many existing methods rely mainly on sensor data, not integrating further information from semantically rich environments. Additionally, most methods assume access to complete scene reconstructions, limiting their applicability in real-world, incremental settings. This paper introduces a novel heterogeneous graph model for incremental 3DSSG prediction that integrates additional, multi-modal information, such as prior observations, directly into the message-passing process. Utilizing multiple layers, the model flexibly incorporates global and local scene representations without requiring specialized modules or full scene reconstructions. We evaluate our approach on the 3DSSG dataset, showing that GNNs enriched with multi-modal information such as semantic embeddings (e.g., CLIP) and prior observations offer a scalable and generalizable solution for complex, real-world environments. The full source code of the presented architecture will be made available at https://github.com/m4renz/incremental-scene-graph-prediction.
This paper introduces EdgeProfiler, a fast profiling framework designed for evaluating lightweight Large Language Models (LLMs) on edge systems. While LLMs offer remarkable capabilities in natural language understanding and generation, their high computational, memory, and power requirements often confine them to cloud environments. EdgeProfiler addresses these challenges by providing a systematic methodology for assessing LLM performance in resource-constrained edge settings. The framework profiles compact LLMs, including TinyLLaMA, Gemma3.1B, Llama3.2-1B, and DeepSeek-r1-1.5B, using aggressive quantization techniques and strict memory constraints. Analytical modeling is used to estimate latency, FLOPs, and energy consumption. The profiling reveals that 4-bit quantization reduces model memory usage by approximately 60-70
Multilayer Perceptrons (MLPs) have long been a cornerstone in deep learning, known for their capacity to model complex relationships. Recently, Kolmogorov-Arnold Networks (KANs) have emerged as a compelling alternative, utilizing highly flexible learnable activation functions directly on network edges, a departure from the neuron-centric approach of MLPs. However, KANs significantly increase the number of learnable parameters, raising concerns about their effectiveness in data-scarce environments. This paper presents a comprehensive comparative study of MLPs and KANs from both algorithmic and experimental perspectives, with a focus on low-data regimes. We introduce an effective technique for designing MLPs with unique, parameterized activation functions for each neuron, enabling a more balanced comparison with KANs. Using empirical evaluations on simulated data and two real-world data sets from medicine and engineering, we explore the trade-offs between model complexity and accuracy, with particular attention to the role of network depth. Our findings show that MLPs with individualized activation functions achieve significantly higher predictive accuracy with only a modest increase in parameters, especially when the sample size is limited to around one hundred. For example, in a three-class classification problem within additive manufacturing, MLPs achieve a median accuracy of 0.91, significantly outperforming KANs, which only reach a median accuracy of 0.53 with default hyperparameters. These results offer valuable insights into the impact of activation function selection in neural networks.
Machine learning systems often require updates for various reasons, such as the availability of new data or models and the need to optimize different technical or ethical metrics. Typically, these metrics reflect an average performance rather than sample-wise behavior. Indeed, improvements in metrics like accuracy can introduce negative flips, where the updated model makes errors that the previous model did not make. In certain applications, these negative flips can be perceived by developers or users as a regression in performance, contributing to the hidden technical debt of machine learning systems. Moreover, if the distribution of negative flips is biased with respect to some sensitive attribute (e.g., gender or race), it may be perceived as discrimination, termed unfair regression. In this paper we show, for the first time, the existence of the phenomenon of unfair regression and propose different ethical metrics to measure it. Additionally, we offer two mitigation strategies - one focused on modifying the learning algorithm and one focused on modifying the tuning phase - to address this issue. Our results on real-world datasets confirm the existence of the unfair regression phenomenon and demonstrate the effectiveness of the proposed mitigation strategies.
Palm vein pattern recognition offers a unique personal identification feature. Unfortunately, these techniques typically require a Near Infrared (NIR) camera sensor to extract the individual's venous pattern, challenging their wide deployment. This paper proposes a new feasible palm vein verification scheme using a Deep Autoencoder and a Siamese Network, implemented threefold. First, we capture the individual's palm using a traditional visible spectrum camera sensor and perform preprocessing tasks to correct imprecise positioning, easing palm support accessories requirements. Second, we eliminate NIR sensor requirement by fine-tuning a Deep Autoencoder model to convert images from the visible spectrum to their infrared counterparts. Third, generated images are processed by a lightweight Siamese network using a contrastive loss function for individual verification. Experiments conducted on a publicly available dataset with over a hundred individuals confirmed the feasibility of our proposal. Our scheme reaches up to 0.97 of true-negative rate, with only 0.01 decrease compared to traditional NIR-based approaches. In addition, individual identification can be conducted in less than 6 seconds in a resource-constrained environment thanks to our lightweight model's implementation.
