arXiv subjects

Burkhard Stiller

Publications and source records attributed to Burkhard Stiller.

At least 19 recordsLinked to original sources

arXiv2026

CACTUS: Mask-Guided Semantic Clean-Label Backdoors in Decentralized Federated Learning

Semantic triggers in federated learning (FL) can be less conspicuous than synthetic patches, but sample-dependent placement may weaken backdoor implantation across aggregation rounds. This challenge is compounded in decentralized FL (DFL), where topology-dependent peer aggregation repeatedly mixes local models. CACTUS converts label-consistent semantic pairs into target-directed representation shifts. Mask-guided, modality-specific operators isolate trigger effects, couple them across samples, and apply the shifts counterfactually to clean non-target embeddings before peer aggregation. Experiments cover speech, text, tabular, and image tasks under nine aggregation rules. With 30\% malicious nodes, CACTUS reaches a nine-rule mean attack success rate (ASR) of 51.2\% on Speech Commands and the highest nine-rule mean ASR among evaluated attacks on three of four modalities. Sensitivity analyses show that ASR varies with network topology and increases with the malicious-node ratio. These results indicate that CACTUS can propagate backdoors through repeated DFL aggregation.

cs.LG↗

arXiv2026

Towards Interpretable Depression Detection: Linking Acoustic Features to DSM-5 Indicators

Depression affects millions worldwide, yet diagnosis relies on subjective self-reports that may miss authentic behavior. This paper presents an approach linking speech acoustics to DSM-5 depressive-behavior indicators through a transparent Linkage Framework. Unlike black-box models, the framework explicitly maps acoustic features (pitch variability, pauses, speech tempo) to clinical indicators, enabling interpretable, indicator-level outputs. The system runs locally on commodity hardware (HW) to preserve privacy. Preliminary evaluation on DAIC-WOZ shows directionally consistent associations between acoustic features and DSM-5 indicators for psychomotor change and concentration difficulty, supporting the design rationale. Future work will validate on longitudinal datasets and extend multimodal integration while maintaining edge constraints.

cs.CL↗

arXiv2026

Functional Requirements for Decentralized and Self-Sovereign Identities

Centralized identity management systems continuously experience security and privacy challenges, motivating the exploration of Decentralized Identity (DI) and Self-Sovereign Identity (SSI) as alternatives. Despite privacy and security benefits to users, the adoption of DI/SSI systems remains limited. One contributing reason is the lack of reproducible approaches to evaluate system compliance with its promised qualities. Derivation of functional requirements (FR) is the first and necessary step to develop such an evaluation approach. Previous literature on DI/SSI significantly lacks the systematic operationalization of existing non-functional requirements (NFR) or SSI principles. This work addresses this research gap by deriving FR for a generalized DI/SSI use case, which encompasses the fundamental operations of the system. The paper details operationalization methodology, introduces a formalized functional model, and presents a comprehensive set of FR, that can be used for future development and evaluation of DI/SSI systems. As a result, establishing the fundamental step toward a reproducible evaluation framework, rooted in established requirements engineering methods.

cs.SE↗

arXiv2026

Rethinking Self-Sovereign Identity Principles: An Actor-Oriented Categorization of Requirements

Centralized identity management systems continuously experience security and privacy challenges, motivating the exploration of Decentralized Identity (DI) and Self-Sovereign Identity (SSI) as user-focused alternatives. Although prior research has consolidated SSI principles and derived quality requirements for DI/SSI systems, it is significantly limited in integrating the user viewpoint. This work addresses this gap by embedding a user perspective into the requirements engineering process for DI/SSI systems. Building on existing SSI principles, composite requirements were decomposed into 24 simple quality or non-functional requirements (NFR). The resulting NFR are systematically mapped to the key actors, namely data owner, issuer, verifier, and system, based on varying degrees of responsibility and ownership. A dependency model is introduced to formalize relationships between actors. Inspired by trust modeling concepts, the model explicitly describes how actors interact and rely on each other for requirements fulfillment. By integrating user-centered requirements, responsibility allocation, ownership specification, and dependency modeling, this work provides the first structured model for DI/SSI system architectures.

cs.SE↗

arXiv2026

A Crowdsensing Intrusion Detection Dataset For Decentralized Federated Learning Models

This paper introduces a dataset and an experimental study on Decentralized Federated Learning (DFL) for Internet of Things (IoT) crowdsensing malware detection. The dataset comprises behavioral records from benign and eight malware attacks. A total of 21,582,484 original records were collected from system calls, file system activities, resource usage, kernel events, input/output events, and network records. These records were aggregated into 30-second windows, resulting in 342,106 data records used for model training and evaluation. Experiments on the DFL platform compare traditional Machine Learning (ML), Centralized Federated Learning (CFL), and DFL across different node counts, topologies, and data distributions. Results show that DFL maintains competitive performance while preserving data locality, outperforming CFL in most settings. This dataset provides a solid foundation for studying the security of IoT crowdsensing environments.

cs.CR↗

arXiv2026

UnlinkableDFL: A Framework for Network-Layer Unlinkability in Decentralized Federated Learning

Decentralized Federated Learning (DFL) removes the central aggregator of conventional Federated Learning, but peer-to-peer model exchange still exposes network traces: who communicates, when fragments move, and which packets correlate across rounds. This paper studies network-layer sender--message linkability for DFL model sharing and presents UnlinkableDFL, a framework in which every participant acts as both a learner and a peer-based mix relay. Shareable model states are split into uniform, onion-encrypted fragment packets and carried over a peer-run mixnet with cover traffic, randomized delays, and independently sampled multi-hop paths. Nodes then perform fragmented aggregation over local and received fragments without sender identities. The analysis bounds sender-linking probability through route uncertainty and relay shuffles, and characterizes when fragment-level aggregation preserves FedAvg-style behavior. A prototype implements QUIC transport, Sphinx-style packets, and Single-Use Reply Block (SURB) acknowledgments. Experiments show that the design sustains learning under sparse deployment while exposing a privacy--cost trade-off: path diversity and relay mixing raise network-layer uncertainty, whereas delay and forwarding dominate overhead. Stress tests confirm robustness to churn and Byzantine updates. A curious-recipient attack marks the boundary of the network-layer guarantee, where payload-level fingerprints survive network-layer anonymization and need complementary defenses, although partial updates and more IID data weaken this attack surface.

cs.NI↗

arXiv2025

IHearYou: Linking Acoustic Features to DSM-5 Depressive Behavior Indicators

Depression affects over millions people worldwide, yet diagnosis still relies on subjective self-reports and interviews that may not capture authentic behavior. We present IHearYou, an approach to automated depression detection focused on speech acoustics. Using passive sensing in household environments, IHearYou extracts voice features and links them to DSM-5 (Diagnostic and Statistical Manual of Mental Disorders) indicators through a structured Linkage Framework instantiated for Major Depressive Disorder. The system runs locally to preserve privacy and includes a persistence schema and dashboard, presenting real-time throughput on a commodity laptop. To ensure reproducibility, we define a configuration-driven protocol with False Discovery Rate (FDR) correction and gender-stratified testing. Applied to the DAIC-WOZ dataset, this protocol reveals directionally consistent feature-indicator associations, while a TESS-based audio streaming experiment validates end-to-end feasibility. Our results show how passive voice sensing can be turned into explainable DSM-5 indicator scores, bridging the gap between black-box detection and clinically interpretable, on-device analysis.

cs.SD↗

arXiv2025

Smarter, not Bigger: Fine-Tuned RAG-Enhanced LLMs for Automotive HIL Testing

Hardware-in-the-Loop (HIL) testing is essential for automotive validation but suffers from fragmented and underutilized test artifacts. This paper presents HIL-GPT, a retrieval-augmented generation (RAG) system integrating domain-adapted large language models (LLMs) with semantic retrieval. HIL-GPT leverages embedding fine-tuning using a domain-specific dataset constructed via heuristic mining and LLM-assisted synthesis, combined with vector indexing for scalable, traceable test case and requirement retrieval. Experiments show that fine-tuned compact models, such as \texttt{bge-base-en-v1.5}, achieve a superior trade-off between accuracy, latency, and cost compared to larger models, challenging the notion that bigger is always better. An A/B user study further confirms that RAG-enhanced assistants improve perceived helpfulness, truthfulness, and satisfaction over general-purpose LLMs. These findings provide insights for deploying efficient, domain-aligned LLM-based assistants in industrial HIL environments.

cs.CL↗

arXiv2025

Distributed Pulse-Wave Simulator for DDoS Dataset Generation

Pulse-wave Distributed Denial-of-Service (DDoS) attacks generate short, synchronized bursts of traffic that circumvent pattern-based detection and quickly exhaust traditional defense systems. This transient and spatially distributed behavior makes analysis extremely challenging, as no public datasets capture how such attacks evolve across multiple network domains. Since each domain observes only a partial viewpoint of the attack, a correlated, multi-vantage view is essential for comprehensive analysis, early detection, and attribution. This paper presents DPWS, an open-source simulator for generating distributed pulse-wave DDoS datasets. DPWS models multi-AS topologies and produces synchronized packet captures at multiple autonomous systems, showing the distributed structure of coordinated bursts. It enables fine-grained control of traffic parameters through a lightweight YAML interface. DPWS reproduces pulse-wave dynamics across multiple vantage points, exhibits natural fingerprint variability at equal aggregate rates, and scales with MPI in ns-3, providing a reproducible basis for studying pulse-wave behaviour and benchmarking distributed DDoS defenses, while sharing practical insights on ns-3 scalability and synchronization gained during development.

cs.NI↗

arXiv2025

ColNet: Collaborative Optimization in Decentralized Federated Multi-task Learning Systems

The integration of Federated Learning (FL) and Multi-Task Learning (MTL) has been explored to address client heterogeneity, with Federated Multi-Task Learning (FMTL) treating each client as a distinct task. However, most existing research focuses on data heterogeneity (e.g., addressing non-IID data) rather than task heterogeneity, where clients solve fundamentally different tasks. Additionally, much of the work relies on centralized settings with a server managing the federation, leaving the more challenging domain of decentralized FMTL largely unexplored. Thus, this work bridges this gap by proposing ColNet, a framework designed for heterogeneous tasks in decentralized federated environments. ColNet partitions models into a backbone and task-specific heads, and uses adaptive clustering based on model and data sensitivity to form task-coherent client groups. Backbones are averaged within groups, and group leaders perform hyper-conflict-averse cross-group aggregation. Across datasets and federations, ColNet outperforms competing schemes under label and task heterogeneity and shows robustness to poisoning attacks.

cs.LG↗

arXiv2025

Bridging Technical Capability and User Accessibility: Off-grid Civilian Emergency Communication

During large-scale crises disrupting cellular and Internet infrastructure, civilians lack reliable methods for communication, aid coordination, and access to trustworthy information. This paper presents a unified emergency communication system integrating a low-power, long-range network with a crisis-oriented smartphone application, enabling decentralized and off-grid civilian communication. Unlike previous solutions separating physical layer resilience from user layer usability, our design merges these aspects into a cohesive crisis-tailored framework. The system is evaluated in two dimensions: communication performance and application functionality. Field experiments in urban Zürich demonstrate that the 868 MHz band, using the LongFast configuration, achieves a communication range of up to 1.2 km with 92% Packet Delivery Ratio, validating network robustness under real-world infrastructure degraded conditions. In parallel, a purpose-built mobile application featuring peer-to-peer messaging, identity verification, and community moderation was evaluated through a requirements-based analysis.

cs.NI↗

arXiv2025

GreenDFL: a Framework for Assessing the Sustainability of Decentralized Federated Learning Systems

Decentralized Federated Learning (DFL) is an emerging paradigm that enables collaborative model training without centralized data and model aggregation, enhancing privacy and resilience. However, its sustainability remains underexplored, as energy consumption and carbon emissions vary across different system configurations. Understanding the environmental impact of DFL is crucial for optimizing its design and deployment. This work aims to develop a comprehensive and operational framework for assessing the sustainability of DFL systems. To address it, this work provides a systematic method for quantifying energy consumption and carbon emissions, offering insights into improving the sustainability of DFL. This work proposes GreenDFL, a fully implementable framework that has been integrated into a real-world DFL platform. GreenDFL systematically analyzes the impact of various factors, including hardware accelerators, model architecture, communication medium, data distribution, network topology, and federation size, on the sustainability of DFL systems. Besides, a sustainability-aware aggregation algorithm (GreenDFL-SA) and a node selection algorithm (GreenDFL-SN) are developed to optimize energy efficiency and reduce carbon emissions in DFL training. Empirical experiments are conducted on multiple datasets, measuring energy consumption and carbon emissions at different phases of the DFL lifecycle. The proposed GreenDFL provides a comprehensive and practical approach for assessing the sustainability of DFL systems. Furthermore, it offers best practices for improving environmental efficiency in DFL, making sustainability considerations more actionable in real-world deployments.

cs.CY↗

arXiv2025

Assessing the Sustainability and Trustworthiness of Federated Learning Models

Artificial intelligence (AI) increasingly influences critical decision-making across sectors. Federated Learning (FL), as a privacy-preserving collaborative AI paradigm, not only enhances data protection but also holds significant promise for intelligent network management, including distributed monitoring, adaptive control, and edge intelligence. Although the trustworthiness of FL systems has received growing attention, the sustainability dimension remains insufficiently explored, despite its importance for scalable real-world deployment. To address this gap, this work introduces sustainability as a distinct pillar within a comprehensive trustworthy FL taxonomy, consistent with AI-HLEG guidelines. This pillar includes three key aspects: hardware efficiency, federation complexity, and the carbon intensity of energy sources. Experiments using the FederatedScope framework under diverse scenarios, including varying participants, system complexity, hardware, and energy configurations, validate the practicality of the approach. Results show that incorporating sustainability into FL evaluation supports environmentally responsible deployment, enabling more efficient, adaptive, and trustworthy network services and management AI models.

cs.CY↗

arXiv2025

From Models to Network Topologies: A Topology Inference Attack in Decentralized Federated Learning

Federated Learning (FL) is widely recognized as a privacy-preserving Machine Learning paradigm due to its model-sharing mechanism that avoids direct data exchange. Nevertheless, model training leaves exploitable traces that can be used to infer sensitive information. In Decentralized FL (DFL), the topology, defining how participants are connected, plays a crucial role in shaping the model's privacy, robustness, and convergence. However, the topology introduces an unexplored vulnerability: attackers can exploit it to infer participant relationships and launch targeted attacks. This work uncovers the hidden risks of DFL topologies by proposing a novel Topology Inference Attack that infers the topology solely from model behavior. A taxonomy of topology inference attacks is introduced, categorizing them by the attacker's capabilities and knowledge. Practical attack strategies are designed for various scenarios, and experiments are conducted to identify key factors influencing attack success. The results demonstrate that analyzing only the model of each node can accurately infer the DFL topology, highlighting a critical privacy risk in DFL systems. These findings offer insights for improving privacy preservation in DFL environments.

cs.LG↗

arXiv2025

FEST: A Unified Framework for Evaluating Synthetic Tabular Data

Synthetic data generation, leveraging generative machine learning techniques, offers a promising approach to mitigating privacy concerns associated with real-world data usage. Synthetic data closely resembles real-world data while maintaining strong privacy guarantees. However, a comprehensive assessment framework is still missing in the evaluation of synthetic data generation, especially when considering the balance between privacy preservation and data utility in synthetic data. This research bridges this gap by proposing FEST, a systematic framework for evaluating synthetic tabular data. FEST integrates diverse privacy metrics (attack-based and distance-based), along with similarity and machine learning utility metrics, to provide a holistic assessment. We develop FEST as an open-source Python-based library and validate it on multiple datasets, demonstrating its effectiveness in analyzing the privacy-utility trade-off of different synthetic data generation models. The source code of FEST is available on Github.

cs.LG↗

arXiv2025

Demo: A Practical Testbed for Decentralized Federated Learning on Physical Edge Devices

Federated Learning (FL) enables collaborative model training without sharing raw data, preserving participant privacy. Decentralized FL (DFL) eliminates reliance on a central server, mitigating the single point of failure inherent in the traditional FL paradigm, while introducing deployment challenges on resource-constrained devices. To evaluate real-world applicability, this work designs and deploys a physical testbed using edge devices such as Raspberry Pi and Jetson Nano. The testbed is built upon a DFL training platform, NEBULA, and extends it with a power monitoring module to measure energy consumption during training. Experiments across multiple datasets show that model performance is influenced by the communication topology, with denser topologies leading to better outcomes in DFL settings.

cs.LG↗

arXiv2025

AugMixCloak: A Defense against Membership Inference Attacks via Image Transformation

Traditional machine learning (ML) raises serious privacy concerns, while federated learning (FL) mitigates the risk of data leakage by keeping data on local devices. However, the training process of FL can still leak sensitive information, which adversaries may exploit to infer private data. One of the most prominent threats is the membership inference attack (MIA), where the adversary aims to determine whether a particular data record was part of the training set. This paper addresses this problem through a two-stage defense called AugMixCloak. The core idea is to apply data augmentation and principal component analysis (PCA)-based information fusion to query images, which are detected by perceptual hashing (pHash) as either identical to or highly similar to images in the training set. Experimental results show that AugMixCloak successfully defends against both binary classifier-based MIA and metric-based MIA across five datasets and various decentralized FL (DFL) topologies. Compared with regularization-based defenses, AugMixCloak demonstrates stronger protection. Compared with confidence score masking, AugMixCloak exhibits better generalization.

cs.LG↗

arXiv2025

QUIC-Exfil: Exploiting QUIC's Server Preferred Address Feature to Perform Data Exfiltration Attacks

The QUIC protocol is now widely adopted by major tech companies and accounts for a significant fraction of today's Internet traffic. QUIC's multiplexing capabilities, encrypted headers, dynamic IP address changes, and encrypted parameter negotiations make the protocol not only more efficient, secure, and censorship-resistant, but also practically unmanageable by firewalls. This opens doors for attackers who may exploit certain traits of the QUIC protocol to perform targeted attacks, such as data exfiltration attacks. Whereas existing data exfiltration techniques, such as TLS and DNS-based exfiltration, can be detected on a firewall level, QUIC-based data exfiltration is more difficult to detect, since changes in IP addresses and ports are inherent to the protocol's normal behavior. To show the feasibility of a QUIC-based data exfiltration attack, we introduce a novel method leveraging the server preferred address feature of the QUIC protocol and, thus, allows an attacker to exfiltrate sensitive data from an infected machine to a malicious server, disguised as a server-side connection migration. The attack is implemented as a proof of concept tool in Rust. We evaluated the performance of five anomaly detection classifiers - Random Forest, Multi-Layer Perceptron, Support Vector Machine, Autoencoder, and Isolation Forest - trained on datasets collected from three network traffic scenarios. The classifiers were trained on over 700K benign and malicious QUIC packets and 786 connection migration events, but were unable to detect the data exfiltration attempts. Furthermore, post-analysis of the traffic captures did not reveal any identifiable fingerprint. As part of our evaluation, we also interviewed five leading firewall vendors and found that, as of today, no major firewall vendor implements functionality capable of distinguishing between benign and malicious QUIC connection migrations.

cs.CR↗