arXiv subjects

Gerhard Wunder

Publications and source records attributed to Gerhard Wunder.

At least 19 recordsLinked to original sources

arXiv2026

Prompt Compression via Activation Aggregation

Large language models process prompts by propagating activations through dozens of layers before generating a response. We ask whether the task-relevant information contained in an instruction prompt can be compressed into a single activation vector and re-injected into the model, replacing the original token sequence? We show this is achievable using a learned weighted sum of activations extracted at an intermediate layer and injected at an early layer of the target LLM. The compressed vector preserves task-relevant information, incurring an accuracy drop of under $2\%$ relative to full prompt processing. Beyond its practical implications, including reducing per-query computation for fixed instruction prompts without reprocessing the original token sequence, our analysis reveals structure in the activation space of LLMs: (i) mid-layer representations transfer meaningfully to early layers, suggesting a degree of cross-layer compatibility in how information is encoded; (ii) a single activation vector encodes a quantifiable and recoverable amount of semantic information; (iii) a weighted sum of activations is a robust representation compressor.

cs.CL↗

arXiv2026

Robust Text Watermarking for Large Language Models via Dual Semantic Embeddings

This work presents Dual-Embedding Watermarking (DEW), a semantic watermarking scheme for large language models (LLMs) that leverages contextual and token-level embeddings to enhance robustness against paraphrasing and translation. DEW utilizes a signal-processing methodology, applying algebraic vector-space operations to token and context embeddings to derive a watermark signal that degrades gracefully under semantic shifts. The method obfuscates the watermark by projecting embedding vectors through pseudo-random matrices seeded with a secret key. Experimental results show that dual-embedding watermarking can offer state-of-the-art robustness, particularly against translation, while incurring relatively low computational overhead compared with other semantic schemes. At lower watermark strength, DEW also maintains competitive text quality, suggesting that dual-embedding signals provide a promising substrate for robust semantic watermarking.

cs.CL↗

arXiv2026

Cross-Layer Intrusion Detection in 5G O-RAN: Gains and Limits of Fusing Radio Telemetry with Network Flow Records

Open RAN disaggregation enables joint analysis of DU radio telemetry and CU-side network-flow records, motivating cross-layer intrusion detection. We evaluate whether fusing these two modalities improves over each individually across seven architectures, using run-disjoint splits over ten seeds on a live 5G O-RAN dataset. Radio features match or outperform network flows on ROC-AUC and run-level detection rate across all architectures. Fusion yields selective ROC-AUC gains but at a one-percent false-positive operating point improves detection rate only for GRU and Transformer, reducing it for the other five models. The benefit is confined to architectures where both single-modality detection rates fall below 0.75. A DoS-to-Benign confusion of 27 to 46 percent persists across all 42 tested configurations of architecture, modality, and window duration, pointing to a limitation in the tested windowed statistical aggregation rather than in model capacity. Code is publicly available.

cs.CR↗

arXiv2026

LLM Self-Recognition: Steering and Retrieving Activation Signatures

Recent advances in interpretability suggest that large language models (LLMs) implicitly encode signals in their generated text that enable self-recognition of their outputs. We demonstrate that this capability is reliable, even in low-entropy scenarios, and that it can be amplified through targeted intervention. By steering the internal residual stream during generation with a random sparse vector, we create a detectable fingerprint that enables attribution of a given text to a specific LLM. This signal is recoverable from the activations of an LLM used as a detector, achieving over 98% accuracy across multiple detection settings while preserving the quality of generated text. As AI-generated content proliferates, this approach offers a practical alternative to traditional detectors by leveraging the model's natural representation structure for attribution rather than embedding a signal externally. Our contributions include: (i) establishing reliable self-recognition capabilities in LLMs, (ii) a simple steering mechanism enabling multi-LLM identification with no quality degradation, (iii) demonstrating that activation spaces contain exploitable structure for encoding signals without semantic interference.

cs.AI↗

arXiv2025

ALIGN-FL: Architecture-independent Learning through Invariant Generative component sharing in Federated Learning

We present ALIGN-FL, a novel approach to distributed learning that addresses the challenge of learning from highly disjoint data distributions through selective sharing of generative components. Instead of exchanging full model parameters, our framework enables privacy-preserving learning by transferring only generative capabilities across clients, while the server performs global training using synthetic samples. Through complementary privacy mechanisms: DP-SGD with adaptive clipping and Lipschitz regularized VAE decoders and a stateful architecture supporting heterogeneous clients, we experimentally validate our approach on MNIST and Fashion-MNIST datasets with cross-domain outliers. Our analysis demonstrates that both privacy mechanisms effectively map sensitive outliers to typical data points while maintaining utility in extreme Non-IID scenarios typical of cross-silo collaborations. Index Terms: Client-invariant Learning, Federated Learning (FL), Privacy-preserving Generative Models, Non-Independent and Identically Distributed (Non-IID), Heterogeneous Architectures

cs.LG↗

arXiv2025

Rethinking Explanation Evaluation under the Retraining Scheme

Feature attribution has gained prominence as a tool for explaining model decisions, yet evaluating explanation quality remains challenging due to the absence of ground-truth explanations. To circumvent this, explanation-guided input manipulation has emerged as an indirect evaluation strategy, measuring explanation effectiveness through the impact of input modifications on model outcomes during inference. Despite the widespread use, a major concern with inference-based schemes is the distribution shift caused by such manipulations, which undermines the reliability of their assessments. The retraining-based scheme ROAR overcomes this issue by adapting the model to the altered data distribution. However, its evaluation results often contradict the theoretical foundations of widely accepted explainers. This work investigates this misalignment between empirical observations and theoretical expectations. In particular, we identify the sign issue as a key factor responsible for residual information that ultimately distorts retraining-based evaluation. Based on the analysis, we show that a straightforward reframing of the evaluation process can effectively resolve the identified issue. Building on the existing framework, we further propose novel variants that jointly structure a comprehensive perspective on explanation evaluation. These variants largely improve evaluation efficiency over the standard retraining protocol, thereby enhancing practical applicability for explainer selection and benchmarking. Following our proposed schemes, empirical results across various data scales provide deeper insights into the performance of carefully selected explainers, revealing open challenges and future directions in explainability research.

cs.LG↗

arXiv2025

Machine and Deep Learning for Indoor UWB Jammer Localization

Ultra-wideband (UWB) localization delivers centimeter-scale accuracy but is vulnerable to jamming attacks, creating security risks for asset tracking and intrusion detection in smart buildings. Although machine learning (ML) and deep learning (DL) methods have improved tag localization, localizing malicious jammers within a single room and across changing indoor layouts remains largely unexplored. Two novel UWB datasets, collected under original and modified room configurations, are introduced to establish comprehensive ML/DL baselines. Performance is rigorously evaluated using a variety of classification and regression metrics. On the source dataset with the collected UWB features, Random Forest achieves the highest F1-macro score of 0.95 and XGBoost achieves the lowest mean Euclidean error of 20.16 cm. However, deploying these source-trained models in the modified room layout led to severe performance degradation, with XGBoost's mean Euclidean error increasing tenfold to 207.99 cm, demonstrating significant domain shift. To mitigate this degradation, a domain-adversarial ConvNeXt autoencoder (A-CNT) is proposed that leverages a gradient-reversal layer to align CIR-derived features across domains. The A-CNT framework restores localization performance by reducing the mean Euclidean error to 34.67 cm. This represents a 77 percent improvement over non-adversarial transfer learning and an 83 percent improvement over the best baseline, restoring the fraction of samples within 30 cm to 0.56. Overall, the results demonstrate that adversarial feature alignment enables robust and transferable indoor jammer localization despite environmental changes. Code and dataset available at https://github.com/afbf4c8996f/Jammer-Loc

cs.LG↗

arXiv2025

Bilinear Compressive Security

Beyond its widespread application in signal and image processing, \emph{compressed sensing} principles have been greatly applied to secure information transmission (often termed 'compressive security'). In this scenario, the measurement matrix $Q$ acts as a one time pad encryption key (in complex number domain) which can achieve perfect information-theoretic security together with other benefits such as reduced complexity and energy efficiency particularly useful in IoT. However, unless the matrix is changed for every message it is vulnerable towards known plain text attacks: only $n$ observations suffices to recover a key $Q$ with $n$ columns. In this paper, we invent and analyze a new method (termed 'Bilinear Compressive Security (BCS)') addressing these shortcomings: In addition to the linear encoding of the message $x$ with a matrix $Q$, the sender convolves the resulting vector with a randomly generated filter $h$. Assuming that $h$ and $x$ are sparse, the receiver can then recover $x$ without knowledge of $h$ from $y=h*Qx$ through blind deconvolution. We study a rather idealized known plaintext attack for recovering $Q$ from repeated observations of $y$'s for different, known $x_k$, with varying and unknown $h$ ,giving Eve a number of advantages not present in practice. Our main result for BCS states that under a weak symmetry condition on the filter $h$, recovering $Q$ will require extensive sampling from transmissions of $\Omega\left(\max\left(n,(n/s)^2\right)\right)$ messages $x_k$ if they are $s$-sparse. Remarkably, with $s=1$ it is impossible to recover the key. In this way, the scheme is much safer than standard compressed sensing even though our assumptions are much in favor towards a potential attacker.

cs.CR↗

arXiv2025

An Investigation into the Performance of Non-Contrastive Self-Supervised Learning Methods for Network Intrusion Detection

Network intrusion detection, a well-explored cybersecurity field, has predominantly relied on supervised learning algorithms in the past two decades. However, their limitations in detecting only known anomalies prompt the exploration of alternative approaches. Motivated by the success of self-supervised learning in computer vision, there is a rising interest in adapting this paradigm for network intrusion detection. While prior research mainly delved into contrastive self-supervised methods, the efficacy of non-contrastive methods, in conjunction with encoder architectures serving as the representation learning backbone and augmentation strategies that determine what is learned, remains unclear for effective attack detection. This paper compares the performance of five non-contrastive self-supervised learning methods using three encoder architectures and six augmentation strategies. Ninety experiments are systematically conducted on two network intrusion detection datasets, UNSW-NB15 and 5G-NIDD. For each self-supervised model, the combination of encoder architecture and augmentation method yielding the highest average precision, recall, F1-score, and AUCROC is reported. Furthermore, by comparing the best-performing models to two unsupervised baselines, DeepSVDD, and an Autoencoder, we showcase the competitiveness of the non-contrastive methods for attack detection. Code at: https://github.com/renje4z335jh4/non_contrastive_SSL_NIDS

cs.CR↗

arXiv2025

Towards a Decentralized IoT Onboarding for Smart Homes Using Consortium Blockchain

The increasing adoption of smart home devices and IoT-based security systems presents significant opportunities to enhance convenience, safety, and risk management for homeowners and service providers. However, secure onboarding-provisioning credentials and establishing trust with cloud platforms-remains a considerable challenge. Traditional onboarding methods often rely on centralized Public Key Infrastructure (PKI) models and manufacturer-controlled keys, which introduce security risks and limit the user's digital sovereignty. These limitations hinder the widespread deployment of scalable IoT solutions. This paper presents a novel onboarding framework that builds upon existing network-layer onboarding techniques and extends them to the application layer to address these challenges. By integrating consortium blockchain technology, we propose a decentralized onboarding mechanism that enhances transparency, security, and monitoring for smart home architectures. The architecture supports device registration, key revocation, access control management, and risk detection through event-driven alerts across dedicated blockchain channels and smart contracts. To evaluate the framework, we formally model the protocol using the Tamarin Prover under the Dolev-Yao adversary model. The analysis focuses on authentication, token integrity, key confidentiality, and resilience over public channels. A prototype implementation demonstrates the system's viability in smart home settings, with verification completing in 0.34 seconds, highlighting its scalability and suitability for constrained devices and diverse stakeholders. Additionally, performance evaluation shows that the blockchain-based approach effectively handles varying workloads, maintains high throughput and low latency, and supports near real-time IoT data processing.

cs.CR↗

arXiv2025

Tuning Block Size for Workload Optimization in Consortium Blockchain Networks

Determining the optimal block size is crucial for achieving high throughput in blockchain systems. Many studies have focused on tuning various components, such as databases, network bandwidth, and consensus mechanisms. However, the impact of block size on system performance remains a topic of debate, often resulting in divergent views and even leading to new forks in blockchain networks. This research proposes a mathematical model to maximize performance by determining the ideal block size for Hyperledger Fabric, a prominent consortium blockchain. By leveraging machine learning and solving the model with a genetic algorithm, the proposed approach assesses how factors such as block size, transaction size, and network capacity influence the block processing time. The integration of an optimization solver enables precise adjustments to block size configuration before deployment, ensuring improved performance from the outset. This systematic approach aims to balance block processing efficiency, network latency, and system throughput, offering a robust solution to improve blockchain performance across diverse business contexts.

cs.CR↗

arXiv2025

A Comprehensive Framework for Building Highly Secure, Network-Connected Devices: Chip to App

The rapid expansion of connected devices has amplified the need for robust and scalable security frameworks. This paper proposes a holistic approach to securing network-connected devices, covering essential layers: hardware, firmware, communication, and application. At the hardware level, we focus on secure key management, reliable random number generation, and protecting critical assets. Firmware security is addressed through mechanisms like cryptographic integrity validation and secure boot processes. For secure communication, we emphasize TLS 1.3 and optimized cipher suites tailored for both standard and resource-constrained devices. To overcome the challenges of IoT, compact digital certificates, such as CBOR, are recommended to reduce overhead and enhance performance. Additionally, the paper explores forward-looking solutions, including post-quantum cryptography, to future-proof systems against emerging threats. This framework provides actionable guidelines for manufacturers and system administrators to build secure devices that maintain confidentiality, integrity, and availability throughout their lifecycle.

cs.NI↗

arXiv2025

Tracking UWB Devices Through Radio Frequency Fingerprinting Is Possible

Ultra-wideband (UWB) is a state-of-the-art technology designed for applications requiring centimeter-level localization. Its widespread adoption by smartphone manufacturer naturally raises security and privacy concerns. Successfully implementing Radio Frequency Fingerprinting (RFF) to UWB could enable physical layer security, but might also allow undesired tracking of the devices. The scope of this paper is to explore the feasibility of applying RFF to UWB and investigates how well this technique generalizes across different environments. We collected a realistic dataset using off-the-shelf UWB devices with controlled variation in device positioning. Moreover, we developed an improved deep learning pipeline to extract the hardware signature from the signal data. In stable conditions, the extracted RFF achieves over 99% accuracy. While the accuracy decreases in more changing environments, we still obtain up to 76% accuracy in untrained locations.

cs.LG↗

arXiv2024

Formal Verification of Permission Voucher

Formal verification is a critical process in ensuring the security and correctness of cryptographic protocols, particularly in high-assurance domains. This paper presents a comprehensive formal analysis of the Permission Voucher Protocol, a system designed for secure and authenticated access control in distributed environments. The analysis employs the Tamarin Prover, a state-of-the-art tool for symbolic verification, to evaluate key security properties such as authentication, confidentiality, integrity, mutual authentication, and replay prevention. We model the protocol's components, including trust relationships, secure channels, and adversary capabilities under the Dolev-Yao model. Verification results confirm the protocol's robustness against common attacks such as message tampering, impersonation, and replay. Additionally, dependency graphs and detailed proofs demonstrate the successful enforcement of security properties like voucher authenticity, data confidentiality, and key integrity. The study identifies potential enhancements, such as incorporating timestamp-based validity checks and augmenting mutual authentication mechanisms to address insider threats and key management challenges. This work highlights the advantages and limitations of using the Tamarin Prover for formal security verification and proposes strategies to mitigate scalability and performance constraints in complex systems.

cs.CR↗

arXiv2024

Transparent Neighborhood Approximation for Text Classifier Explanation

Recent literature highlights the critical role of neighborhood construction in deriving model-agnostic explanations, with a growing trend toward deploying generative models to improve synthetic instance quality, especially for explaining text classifiers. These approaches overcome the challenges in neighborhood construction posed by the unstructured nature of texts, thereby improving the quality of explanations. However, the deployed generators are usually implemented via neural networks and lack inherent explainability, sparking arguments over the transparency of the explanation process itself. To address this limitation while preserving neighborhood quality, this paper introduces a probability-based editing method as an alternative to black-box text generators. This approach generates neighboring texts by implementing manipulations based on in-text contexts. Substituting the generator-based construction process with recursive probability-based editing, the resultant explanation method, XPROB (explainer with probability-based editing), exhibits competitive performance according to the evaluation conducted on two real-world datasets. Additionally, XPROB's fully transparent and more controllable construction process leads to superior stability compared to the generator-based explainers.

cs.CL↗

arXiv2024

DBNode: A Decentralized Storage System for Big Data Storage in Consortium Blockchains

Storing big data directly on a blockchain poses a substantial burden due to the need to maintain a consistent ledger across all nodes. Numerous studies in decentralized storage systems have been conducted to tackle this particular challenge. Most state-of-the-art research concentrates on developing a general storage system that can accommodate diverse blockchain categories. However, it is essential to recognize the unique attributes of a consortium blockchain, such as data privacy and access control. Beyond ensuring high performance, these specific needs are often overlooked by general storage systems. This paper proposes a decentralized storage system for Hyperledger Fabric, which is a well-known consortium blockchain. First, we employ erasure coding to partition files, subsequently organizing these chunks into a hierarchical structure that fosters efficient and dependable data storage. Second, we design a two-layer hash-slots mechanism and a mirror strategy, enabling high data availability. Third, we design an access control mechanism based on a smart contract to regulate file access.

cs.CR↗

arXiv2024

Advancements in UWB: Paving the Way for Sovereign Data Networks in Healthcare Facilities

Ultra-Wideband (UWB) technology re-emerges as a groundbreaking ranging technology with its precise micro-location capabilities and robustness. This paper highlights the security dimensions of UWB technology, focusing in particular on the intricacies of device fingerprinting for authentication, examined through the lens of state-of-the-art deep learning techniques. Furthermore, we explore various potential enhancements to the UWB standard that could realize a sovereign UWB data network. We argue that UWB data communication holds significant potential in healthcare and ultra-secure environments, where the use of the common unlicensed 2.4~GHz band-centric wireless technology is limited or prohibited. A sovereign UWB network could serve as an alternative, providing secure localization and short-range data communication in such environments.

cs.NI↗

arXiv2024

From Worst to Average Case to Incremental Search Bounds of the Strong Lucas Test

The strong Lucas test is a widely used probabilistic primality test in cryptographic libraries. When combined with the Miller-Rabin primality test, it forms the Baillie-PSW primality test, known for its absence of false positives, undermining the relevance of a complete understanding of the strong Lucas test. In primality testing, the worst-case error probability serves as an upper bound on the likelihood of incorrectly identifying a composite as prime. For the strong Lucas test, this bound is $4/15$ for odd composites, not products of twin primes. On the other hand, the average-case error probability indicates the probability that a randomly chosen integer is inaccurately classified as prime by the test. This bound is especially important for practical applications, where we test primes that are randomly generated and not generated by an adversary. The error probability of $4/15$ does not directly carry over due to the scarcity of primes, and whether this estimate holds has not yet been established in the literature. This paper addresses this gap by demonstrating that an integer passing $t$ consecutive test rounds, alongside additional standard tests of low computational cost, is indeed prime with a probability greater than $1-(4/15)^t$ for all $t\geq 1$. Furthermore, we introduce error bounds for the incremental search algorithm based on the strong Lucas test, as there are no established bounds up to date as well. Rather than independent selection, in this approach, the candidate is chosen uniformly at random, with subsequent candidates determined by incrementally adding 2. This modification reduces the need for random bits and enhances the efficiency of trial division computation further.

cs.CR↗