Papers

Large language models remain vulnerable to jailbreak attacks, and single-layer defenses often trade security for usability. We present TRYLOCK, the first defense-in-depth architecture that combines four heterogeneous mechanisms across the inference stack: weight-level safety alignment via DPO, activation-level control via Representation Engineering (RepE) steering, adaptive steering strength selected by a lightweight sidecar classifier, and input canonicalization to neutralize encoding-based bypasses. On Mistral-7B-Instruct evaluated against a 249-prompt attack set spanning five attack families, TRYLOCK achieves 88.0% relative ASR reduction (46.5% to 5.6%), with each layer contributing unique coverage: RepE blocks 36% of attacks that bypass DPO alone, while canonicalization catches 14% of encoding attacks that evade both. We discover a non-monotonic steering phenomenon – intermediate strength (alpha=1.0) degrades safety below baseline – and provide mechanistic hypotheses explaining RepE-DPO interference. The adaptive sidecar reduces over-refusal from 60% to 48% while maintaining identical attack defense, demonstrating that security and usability need not be mutually exclusive. We release all components – trained adapters, steering vectors, sidecar classifier, preference pairs, and complete evaluation methodology – enabling full reproducibility.

Training data is a critical and often proprietary asset in Large Language Model (LLM) development, motivating the use of data watermarking to embed model-transferable signals for usage verification. We identify low coverage as a vital yet largely overlooked requirement for practicality, as individual data owners typically contribute only a minute fraction of massive training corpora. Prior methods fail to maintain stealthiness, verification feasibility, or robustness when only one or a few sequences can be modified. To address these limitations, we introduce SLIM, a framework enabling per-user data provenance verification under strict black-box access. SLIM leverages intrinsic LLM properties to induce a Latent-Space Confusion Zone by training the model to map semantically similar prefixes to divergent continuations. This manifests as localized generation instability, which can be reliably detected via hypothesis testing. Experiments demonstrate that SLIM achieves ultra-low coverage capability, strong black-box verification performance, and great scalability while preserving both stealthiness and model utility, offering a robust solution for protecting training data in modern LLM pipelines.

We provide new insights into an open problem recently posed by Yuan-Sun [ISIT 2025], concerning the minimum individual key rate required in the vector linear secure aggregation problem. Consider a distributed system with $K$ users, where each user $k\in [K]$ holds a data stream $W_k$ and an individual key $Z_k$. A server aims to compute a linear function $\mathbf{F}[W_1;\ldots;W_K]$ without learning any information about another linear function $\mathbf{G}[W_1;\ldots;W_K]$, where $[W_1;\ldots;W_K]$ denotes the row stack of $W_1,\ldots,W_K$. The open problem is to determine the minimum required length of $Z_k$, denoted as $R_k$, $k\in [K]$. In this paper, we characterize a new achievable region for the rate tuple $(R_1,\ldots,R_K)$. The region is polyhedral, with vertices characterized by a binary rate assignment $(R_1,\ldots,R_K) = (\mathbf{1}(1 \in \mathcal{I}),\ldots,\mathbf{1}(K\in \mathcal{I}))$, where $\mathcal{I}\subseteq [K]$ satisfies the rank-increment condition: $\mathrm{rank}\left(\bigl[\mathbf{F}_{\mathcal{I}};\mathbf{G}_{\mathcal{I}}\bigr]\right) =\mathrm{rank}\bigl(\mathbf{F}_{\mathcal{I}}\bigr)+N$. Here, $\mathbf{F}_\mathcal{I}$ and $\mathbf{G}_\mathcal{I}$ are the submatrices formed by the columns indexed by $\mathcal{I}$. Our results uncover the novel fact that it is not necessary for every user to hold a key, thereby strictly enlarging the best-known achievable region in the literature. Furthermore, we provide a converse analysis to demonstrate its optimality when minimizing the number of users that hold keys.

In this paper, we introduce FlexProofs, a new vector commitment (VC) scheme that achieves two key properties: (1) the prover can generate all individual opening proofs for a vector of size $N$ in optimal time ${\cal O}(N)$, and there is a flexible batch size parameter $b$ that can be increased to further reduce the time to generate all proofs; and (2) the scheme is directly compatible with a family of zkSNARKs that encode their input as a multi-linear polynomial. As a critical building block, we propose the first functional commitment (FC) scheme for multi-exponentiations with batch opening. Compared with HydraProofs, the only existing VC scheme that computes all proofs in optimal time ${\cal O}(N)$ and is directly compatible with zkSNARKs, FlexProofs may speed up the process of generating all proofs, if the parameter $b$ is properly chosen. Our experiments show that for $N=2^{16}$ and $b=\log^2 N$, FlexProofs can be $6\times$ faster than HydraProofs. Moreover, when combined with suitable zkSNARKs, FlexProofs enable practical applications such as verifiable secret sharing and verifiable robust aggregation.

Security operation centers (SOCs) often produce analysis reports on security incidents, and large language models (LLMs) will likely be used for this task in the near future. We postulate that a better understanding of how veteran analysts evaluate reports, including their feedback, can help produce analysis reports in SOCs. In this paper, we aim to leverage LLMs for analysis reports. To this end, we first construct a Analyst-wise checklist to reflect SOC practitioners' opinions for analysis report evaluation through literature review and user study with SOC practitioners. Next, we design a novel LLM-based conceptual framework, named MESSALA, by further introducing two new techniques, granularization guideline and multi-perspective evaluation. MESSALA can maximize report evaluation and provide feedback on veteran SOC practitioners' perceptions. When we conduct extensive experiments with MESSALA, the evaluation results by MESSALA are the closest to those of veteran SOC practitioners compared with the existing LLM-based methods. We then show two key insights. We also conduct qualitative analysis with MESSALA, and then identify that MESSALA can provide actionable items that are necessary for improving analysis reports.

Despite extensive safety alignment, Large Language Models (LLMs) often fail against jailbreak attacks. While machine unlearning has emerged as a promising defense by erasing specific harmful parameters, current methods remain vulnerable to diverse jailbreaks. We first conduct an empirical study and discover that this failure mechanism is caused by jailbreaks primarily activating non-erased parameters in the intermediate layers. Further, by probing the underlying mechanism through which these circumvented parameters reassemble into the prohibited output, we verify the persistent existence of dynamic $\textbf{jailbreak paths}$ and show that the inability to rectify them constitutes the fundamental gap in existing unlearning defenses. To bridge this gap, we propose $\textbf{J}$ailbreak $\textbf{P}$ath $\textbf{U}$nlearning (JPU), which is the first to rectify dynamic jailbreak paths towards safety anchors by dynamically mining on-policy adversarial samples to expose vulnerabilities and identify jailbreak paths. Extensive experiments demonstrate that JPU significantly enhances jailbreak resistance against dynamic attacks while preserving the model's utility.

The aim of this work is to enhance blockchain security by deepening the understanding of selfish mining attacks in various consensus protocols, especially the ones that have the potential to mitigate selfish mining. Previous research was mainly focused on a particular protocol with a single selfish miner, while only limited studies have been conducted on two or more attackers. To address this gap, we proposed a stochastic simulation framework that enables analysis of selfish mining with multiple attackers across various consensus protocols. We created the model of Proof-of-Work (PoW) Nakamoto consensus (serving as the baseline) as well as models of two additional consensus protocols designed to mitigate selfish mining: Fruitchain and Strongchain. Using our framework, thresholds reported in the literature were verified, and several novel thresholds were discovered for 2 and more attackers. We made the source code of our framework available, enabling researchers to evaluate any newly added protocol with one or more selfish miners and cross-compare it with already modeled protocols.

The high rate of development of Internet of Things (IoT) devices has brought to attention new challenges in the area of data security, especially within the resource-limited realm of RFID tags, sensors, and embedded systems. Traditional cryptographic implementations can be of inappropriate computational complexity and energy usage and hence are not suitable on these platforms. This paper examines the design, implementation, and testing of lightweight cryptographic algorithms that have been specifically designed to be used in secure embedded systems. A comparison of some of the state-of-the-art lightweight encryption algorithms, that is PRESENT, SPECK, and SIMON, focuses on the main performance indicators, i.e., throughput, use of memory, and energy utilization. The study presents novel lightweight algorithms that are founded upon the Feistel-network architecture and their safety under cryptanalytic attacks, e.g., differential and linear cryptanalysis. The proposed solutions are proven through hardware implementation on the FPGA platform. The results have shown that lightweight cryptography is an effective strategy that could be used to establish security and maintain performance in the IoT and other resource-limited settings.

Trust between entities in any scenario without a trusted third party is very difficult, and trust is exactly what blockchain aims to bring into the digital world with its basic features. Many applications are moving to blockchain adoption, enabling users to work in a trustworthy manner. The early generations of blockchain have a problem; they cannot share information with other blockchains. As more and more entities move their applications to the blockchain, they generate large volumes of data, and as applications have become more complex, sharing information between different blockchains has become a necessity. This has led to the research and development of interoperable solutions allowing blockchains to connect together. This paper discusses a few blockchain platforms that provide interoperable solutions, emphasising their ability to connect heterogeneous blockchains. It also discusses a case study scenario to illustrate the importance and benefits of using interoperable solutions. We also present a few topics that need to be solved in the realm of interoperability.

Synthetic Data Generation (SDG) can be used to facilitate privacy-preserving data sharing. However, most existing research focuses on privacy attacks where the adversary is the recipient of the released synthetic data and attempts to infer sensitive information from it. This study investigates quality degradation attacks initiated by adversaries who possess access to the real dataset or control over the generation process, such as the data owner, the synthetic data provider, or potential intruders. We formalize a corresponding threat model and empirically evaluate the effectiveness of targeted manipulations of real data (e.g., label flipping and feature-importance-based interventions) on the quality of generated synthetic data. The results show that even small perturbations can substantially reduce downstream predictive performance and increase statistical divergence, exposing vulnerabilities within SDG pipelines. This study highlights the need to integrate integrity verification and robustness mechanisms, alongside privacy protection, to ensure the reliability and trustworthiness of synthetic data sharing frameworks.

SAST (Static Application Security Testing) tools are among the most widely used techniques in defensive cybersecurity, employed by commercial and non-commercial organizations to identify potential vulnerabilities in software. Despite their great utility, they generate numerous false positives, requiring costly manual filtering (aka triage). While LLM-powered agents show promise for automating cybersecurity tasks, existing benchmarks fail to emulate real-world SAST finding distributions. We introduce SastBench, a benchmark for evaluating SAST triage agents that combines real CVEs as true positives with filtered SAST tool findings as approximate false positives. SastBench features an agent-agnostic design. We evaluate different agents on the benchmark and present a comparative analysis of their performance, provide a detailed analysis of the dataset, and discuss the implications for future development.

As audio deepfakes transition from research artifacts to widely available commercial tools, robust biometric authentication faces pressing security threats in high-stakes industries. This paper presents a systematic empirical evaluation of state-of-the-art speaker authentication systems based on a large-scale speech synthesis dataset, revealing two major security vulnerabilities: 1) modern voice cloning models trained on very small samples can easily bypass commercial speaker verification systems; and 2) anti-spoofing detectors struggle to generalize across different methods of audio synthesis, leading to a significant gap between in-domain performance and real-world robustness. These findings call for a reconsideration of security measures and stress the need for architectural innovations, adaptive defenses, and the transition towards multi-factor authentication.

Linear queries, as the basis of broad analysis tasks, are often released through privacy mechanisms based on differential privacy (DP), the most popular framework for privacy protection. However, DP adopts a context-free definition that operates independently of the data-generating distribution. In this paper, we revisit the privacy analysis of the Laplace mechanism through the lens of pointwise maximal leakage (PML). We demonstrate that the distribution-agnostic definition of the DP framework often mandates excessive noise. To address this, we incorporate an assumption about the prior distribution by lower-bounding the probability of any single record belonging to any specific class. With this assumption, we derive a tight, context-aware leakage bound for general linear queries, and prove that our derived bound is strictly tighter than the standard DP guarantee and converges to the DP guarantee as this probability lower bound approaches zero. Numerical evaluations demonstrate that by exploiting this prior knowledge, the required noise scale can be reduced while maintaining privacy guarantees.

Most membership inference attacks (MIAs) against Large Language Models (LLMs) rely on global signals, like average loss, to identify training data. This approach, however, dilutes the subtle, localized signals of memorization, reducing attack effectiveness. We challenge this global-averaging paradigm, positing that membership signals are more pronounced within localized contexts. We introduce WBC (Window-Based Comparison), which exploits this insight through a sliding window approach with sign-based aggregation. Our method slides windows of varying sizes across text sequences, with each window casting a binary vote on membership based on loss comparisons between target and reference models. By ensembling votes across geometrically spaced window sizes, we capture memorization patterns from token-level artifacts to phrase-level structures. Extensive experiments across eleven datasets demonstrate that WBC substantially outperforms established baselines, achieving higher AUC scores and 2-3 times improvements in detection rates at low false positive thresholds. Our findings reveal that aggregating localized evidence is fundamentally more effective than global averaging, exposing critical privacy vulnerabilities in fine-tuned LLMs.

Learning and Employment Record (LER) systems are emerging as critical infrastructure for securely compiling and sharing educational and work achievements. Existing blockchain-based platforms leverage verifiable credentials but typically lack automated skill-credential generation and the ability to incorporate unstructured evidence of learning. In this paper,a privacy-preserving, AI-enabled decentralized LER system is proposed to address these gaps. Digitally signed transcripts from educational institutions are accepted, and verifiable self-issued skill credentials are derived inside a trusted execution environment (TEE) by a natural language processing pipeline that analyzes formal records (e.g., transcripts, syllabi) and informal artifacts. All verification and job-skill matching are performed inside the enclave with selective disclosure, so raw credentials and private keys remain enclave-confined. Job matching relies solely on attested skill vectors and is invariant to non-skill resume fields, thereby reducing opportunities for screening bias.The NLP component was evaluated on sample learner data; the mapping follows the validated Syllabus-to-O*NET methodology,and a stability test across repeated runs observed <5% variance in top-ranked skills. Formal security statements and proof sketches are provided showing that derived credentials are unforgeable and that sensitive information remains confidential. The proposed system thus supports secure education and employment credentialing, robust transcript verification,and automated, privacy-preserving skill extraction within a decentralized framework.

AI-assisted developer services are increasingly embedded in modern IDEs, yet enterprises must ensure these tools operate within existing identity, access control, and governance requirements. The Model Context Protocol (MCP) enables AI assistants to retrieve structured internal context, but its specification provides only a minimal authorization model and lacks guidance on integrating enterprise SSO. This article presents a practical architecture that incorporates OAuth 2.0 and OpenID Connect (OIDC) into MCP-enabled developer environments. It describes how IDE extensions obtain and present tokens, how MCP servers validate them through an identity provider, and how scopes and claims can enforce least-privilege access. A prototype implementation using Visual Studio Code, a Python-based MCP server, and an OIDC-compliant IdP demonstrates feasibility. A case study evaluates authentication latency, token-validation overhead, operational considerations, and AI-specific risks. The approach provides a deployable pattern for organizations adopting AI-assisted developer tools while maintaining identity assurance and auditability.

Model quantization is critical for deploying large language models (LLMs) on resource-constrained hardware, yet recent work has revealed severe security risks that benign LLMs in full precision may exhibit malicious behaviors after quantization. In this paper, we propose Adversarial Contrastive Learning (ACL), a novel gradient-based quantization attack that achieves superior attack effectiveness by explicitly maximizing the gap between benign and harmful responses probabilities. ACL formulates the attack objective as a triplet-based contrastive loss, and integrates it with a projected gradient descent two-stage distributed fine-tuning strategy to ensure stable and efficient optimization. Extensive experiments demonstrate ACL's remarkable effectiveness, achieving attack success rates of 86.00% for over-refusal, 97.69% for jailbreak, and 92.40% for advertisement injection, substantially outperforming state-of-the-art methods by up to 44.67%, 18.84%, and 50.80%, respectively.

The growing complexity of modern system-on-chip (SoC) and IP designs is making security assurance difficult day by day. One of the fundamental steps in the pre-silicon security verification of a hardware design is the identification of security assets, as it substantially influences downstream security verification tasks, such as threat modeling, security property generation, and vulnerability detection. Traditionally, assets are determined manually by security experts, requiring significant time and expertise. To address this challenge, we present LAsset, a novel automated framework that leverages large language models (LLMs) to identify security assets from both hardware design specifications and register-transfer level (RTL) descriptions. The framework performs structural and semantic analysis to identify intra-module primary and secondary assets and derives inter-module relationships to systematically characterize security dependencies at the design level. Experimental results show that the proposed framework achieves high classification accuracy, reaching up to 90% recall rate in SoC design, and 93% recall rate in IP designs. This automation in asset identification significantly reduces manual overhead and supports a scalable path forward for secure hardware development.

LLM-based agents are increasingly deployed to autonomously solve complex tasks, raising urgent needs for IP protection and regulatory provenance. While content watermarking effectively attributes LLM-generated outputs, it fails to directly identify the high-level planning behaviors (e.g., tool and subgoal choices) that govern multi-step execution. Critically, watermarking at the planning-behavior layer faces unique challenges: minor distributional deviations in decision-making can compound during long-term agent operation, degrading utility, and many agents operate as black boxes that are difficult to intervene in directly. To bridge this gap, we propose AgentMark, a behavioral watermarking framework that embeds multi-bit identifiers into planning decisions while preserving utility. It operates by eliciting an explicit behavior distribution from the agent and applying distribution-preserving conditional sampling, enabling deployment under black-box APIs while remaining compatible with action-layer content watermarking. Experiments across embodied, tool-use, and social environments demonstrate practical multi-bit capacity, robust recovery from partial logs, and utility preservation. The code is available at https://github.com/Tooooa/AgentMark.

We present SWaRL, a robust and fidelity-preserving watermarking framework designed to protect the intellectual property of code LLMs by embedding unique and verifiable signatures in the generated program. Existing watermarking approaches either rely on handcrafted code transformations or manipulate token generation probabilities at inference time, making them vulnerable to removal attacks or prone to breaking functional correctness. To address these challenges, SWaRL employs a reinforcement learning-based co-training framework that uses compiler feedback for functional correctness and a jointly trained confidential verifier as a reward signal to maintain watermark detectability. Furthermore, SWaRL employs low-rank adaptation (LoRA) during fine-tuning, enabling efficient integration of watermarking behavior and transferability across model updates. Extensive experiments show that SWaRL achieves strong watermark detection accuracy compared to prior methods while fully maintaining watermarked code functionality. Moreover, SWaRL exhibits strong resilience against refactoring and adversarial transformation attacks, which maintains reliable attribution without substantial computational overhead.

This paper explores coordinated deception strategies by synchronizing defenses across coupled cyber and physical systems to mislead attackers and strengthen defense mechanisms. We introduce a Stackelberg game framework to model the strategic interaction between defenders and attackers, where the defender leverages CVSS-based exploit probabilities and real-world vulnerability data from the National Vulnerability Database (NVD) to guide the deployment of deception. Cyber and physical replicas are used to disrupt attacker reconnaissance and enhance defensive effectiveness. We propose a CVE-based utility function to identify the most critical vulnerabilities and demonstrate that coordinated multilayer deception outperforms single-layer and baseline strategies in improving defender utility across both CVSS versions.

We study a hybrid computational model for integer factorization in which the only non-classical resource is access to an iterated diffusion process on a finite graph. Concretely, a diffusion step is defined to be one application of a symmetric stochastic matrix (the half-lazy walk operator) to an $\ell^{1}$–normalized state vector, followed by an optional readout of selected coordinates. Let $N\ge 3$ be an odd integer which is neither prime nor a prime power, and let $b\in(\mathbb{Z}/N\mathbb{Z})^\ast$ have odd multiplicative order $r={\rm ord}_N(b)$. We construct, without knowing $r$ in advance, a weighted Cayley graph whose vertex set is the cyclic subgroup $\langle b\rangle$ and whose edges correspond to the powers $b^{\pm 2^t}$ for $t\le \lfloor \log_2 N\rfloor+1$. Using an explicit spectral decomposition together with an elementary doubling lemma, we show that $r$ can be recovered from a single heat-kernel value after at most $O((\log_2 N)^2)$ diffusion steps, with an effective bound. We then combine this order-finding model with the standard reduction from factoring to order finding (in the spirit of Shor's framework) to obtain a randomized factorization procedure whose success probability depends only on the number $m$ of distinct prime factors of $N$. Our comparison with Shor's algorithm is conceptual and model-based. We replace unitary $\ell^2$ evolution by Markovian $\ell^1$ evolution, and we report complexity in two cost measures: digital steps and diffusion steps. Finally, we include illustrative examples and discussion of practical implementations.

We introduce Auditable Proof-of-Work (APoW), a novel proof-of-work (PoW) construction inspired by Hashcash-style nonce searching, which enables the auditing of other miners' work through accountable re-scanning of the nonce space. The proposed scheme allows a miner to probabilistically attest to having searched specified regions of the nonce space in earlier mining rounds, while concurrently earning rewards for performing productive work for a new block or pool share. This capability enables miners belonging to a mining pools to audit another miner's claimed effort retroactively, thereby allowing the probabilistic detection of block withholding attacks (BWAs) without requiring trusted hardware or trusted third parties. As a consequence, the construction supports the design of decentralized mining pools in which work attribution is verifiable and withholding incentives are substantially reduced. The scheme preserves the fundamental properties of conventional PoW, including public verifiability and difficulty adjustment, while adding an orthogonal auditability layer tailored to pool-based mining. Finally, while a full deployment of APoW in Bitcoin would require a consensus rule change and minor modifications to mining ASICs, the construction remains practically useful even without consensus changes, for instance, as a pool-level auditing mechanism that enables verifiable pay-for-auditing using existing pool reserves.

The rapid advancement of speech synthesis technologies, including text-to-speech (TTS) and voice conversion (VC), has intensified security and privacy concerns related to voice cloning. Recent defenses attempt to prevent unauthorized cloning by embedding protective perturbations into speech to obscure speaker identity while maintaining intelligibility. However, adversaries can apply advanced purification techniques to remove these perturbations, recover authentic acoustic characteristics, and regenerate cloneable voices. Despite the growing realism of such attacks, the robustness of existing defenses under adaptive purification remains insufficiently studied. Most existing purification methods are designed to counter adversarial noise in automatic speech recognition (ASR) systems rather than speaker verification or voice cloning pipelines. As a result, they fail to suppress the fine-grained acoustic cues that define speaker identity and are often ineffective against speaker verification attacks (SVA). To address these limitations, we propose Diffusion-Bridge (VocalBridge), a purification framework that learns a latent mapping from perturbed to clean speech in the EnCodec latent space. Using a time-conditioned 1D U-Net with a cosine noise schedule, the model enables efficient, transcript-free purification while preserving speaker-discriminative structure. We further introduce a Whisper-guided phoneme variant that incorporates lightweight temporal guidance without requiring ground-truth transcripts. Experimental results show that our approach consistently outperforms existing purification methods in recovering cloneable voices from protected speech. Our findings demonstrate the fragility of current perturbation-based defenses and highlight the need for more robust protection mechanisms against evolving voice-cloning and speaker verification threats.

Software vulnerability detection can be formulated as a binary classification problem that determines whether a given code snippet contains security defects. Existing multimodal methods typically fuse Natural Code Sequence (NCS) representations extracted by pretrained models with Code Property Graph (CPG) representations extracted by graph neural networks, under the implicit assumption that introducing an additional modality necessarily yields information gain. Through empirical analysis, we demonstrate the limitations of this assumption: pretrained models already encode substantial structural information implicitly, leading to strong overlap between the two modalities; moreover, graph encoders are generally less effective than pretrained language models in feature extraction. As a result, naive fusion not only struggles to obtain complementary signals but can also dilute effective discriminative cues due to noise propagation. To address these challenges, we propose a task-conditioned complementary fusion strategy that uses Fisher information to quantify task relevance, transforming cross-modal interaction from full-spectrum matching into selective fusion within a task-sensitive subspace. Our theoretical analysis shows that, under an isotropic perturbation assumption, this strategy significantly tightens the upper bound on the output error. Based on this insight, we design the TaCCS-DFA framework, which combines online low-rank Fisher subspace estimation with an adaptive gating mechanism to enable efficient task-oriented fusion. Experiments on the BigVul, Devign, and ReVeal benchmarks demonstrate that TaCCS-DFA delivers up to a 6.3-point gain in F1 score with only a 3.4% increase in inference latency, while maintaining low calibration error.

We study differentially-private statistics in the fully dynamic continual observation model, where many updates can arrive at each time step and updates to a stream can involve both insertions and deletions of an item. Earlier work (e.g., Jain et al., NeurIPS 2023 for counting distinct elements; Raskhodnikova & Steiner, PODS 2025 for triangle counting with edge updates) reduced the respective cardinality estimation problem to continual counting on the difference stream associated with the true function values on the input stream. In such reductions, a change in the original stream can cause many changes in the difference stream, this poses a challenge for applying private continual counting algorithms to obtain optimal error bounds. We improve the accuracy of several such reductions by studying the associated $\ell_p$-sensitivity vectors of the resulting difference streams and isolating their properties. We demonstrate that our framework gives improved bounds for counting distinct elements, estimating degree histograms, and estimating triangle counts (under a slightly relaxed privacy model), thus offering a general approach to private continual cardinality estimation in streaming settings. Our improved accuracy stems from tight analysis of known factorization mechanisms for the counting matrix in this setting; the key technical challenge is arguing that one can use state-of-the-art factorizations for sensitivity vector sets with the properties we isolate. Empirically and analytically, we demonstrate that our improved error bounds offer a substantial improvement in accuracy for cardinality estimation problems over a large range of parameters.

Modern identity and trust systems collapse in the environments where they are needed most: disaster zones, disconnected or damaged networks, and adversarial conditions such as censorship or infrastructure interference. These systems depend on functioning networks to reach online authorities, resolvers, directories, and revocation services, leaving trust unverifiable whenever communication is unavailable or untrusted. This work demonstrates that secure identity and trust are possible without such infrastructure. We introduce the Zero-Infrastructure Capability Graph (ZI-CG), a model showing that identity, delegation, and revocation can be represented as self-contained, signed statements whose validity is determined entirely by local, deterministic evaluation. We further present Vouchsafe, a complete working instantiation of this model built using widely deployed primitives including Ed25519, SHA-256, and structured JSON Web Tokens, requiring no new cryptography or online services. The results show that a practical, offline-verifiable trust substrate can be constructed today using only the cryptographic data presented at evaluation time.

The rapid increase of Internet of Things (IoT) systems across several domains has led to the generation of vast volumes of sensitive data, presenting significant challenges in terms of storage and data analytics. Cloud-assisted IoT solutions offer storage, scalability, and computational resources, but introduce new security and privacy risks that conventional trust-based approaches fail to adequately mitigate. To address these challenges, this paper presents MOZAIK, a novel end-to-end privacy-preserving confidential data storage and distributed processing architecture tailored for IoT-to-cloud scenarios. MOZAIK ensures that data remains encrypted throughout its lifecycle, including during transmission, storage, and processing. This is achieved by employing a cryptographic privacy-enhancing technology known as computing on encrypted data (COED). Two distinct COED techniques are explored, specifically secure multi-party computation (MPC) and fully homomorphic encryption (FHE). The paper includes a comprehensive analysis of the MOZAIK architecture, including a proof-of-concept implementation and performance evaluations. The evaluation results demonstrate the feasibility of the MOZAIK system and indicate the cost of an end-to-end privacy-preserving system compared to regular plaintext alternatives. All components of the MOZAIK platform are released as open-source software alongside this publication, with the aim of advancing secure and privacy-preserving data processing practices.

Modern cyberattacks are increasingly complex, posing significant challenges to classical machine learning methods, particularly when labeled data is limited and feature interactions are highly non-linear. In this study we investigates the potential of hybrid quantum-classical learning to enhance feature representations for intrusion detection and explore possible quantum advantages in cybersecurity analytics. Using the UNSW-NB15 dataset, network traffic is transformed into structured feature vectors through classical preprocessing and normalization. Classical models, including Logistic Regression and Support Vector Machines with linear and RBF kernels, are evaluated on the full dataset to establish baseline performance under large-sample conditions. Simultaneously, a quantum-enhanced pipeline maps classical features into variational quantum circuits via angle encoding and entangling layers, executed on a CPU-based quantum simulator, with resulting quantum embeddings classified using a classical SVM. Experiments show that while classical models achieve higher overall accuracy with large datasets, quantum-enhanced representations demonstrate superior attack recall and improved class separability when data is scarce, suggesting that quantum feature spaces capture complex correlations inaccessible to shallow classical models. These results highlight the potential of quantum embeddings to improve generalization and representation quality in cybersecurity tasks and provide a reproducible framework for evaluating quantum advantages as quantum hardware and simulators continue to advance.

This paper explores how a recent European Union proposal, the so-called Chat Control, which creates regulatory incentives for providers to implement content detection and communication scanning, could transform the foundations of human-robot interaction (HRI). As robots increasingly act as interpersonal communication channels in care, education, and telepresence, they convey not only speech but also gesture, emotion, and contextual cues. We argue that extending digital surveillance laws to such embodied systems would entail continuous monitoring, embedding observation into the very design of everyday robots. This regulation blurs the line between protection and control, turning companions into potential informants. At the same time, monitoring mechanisms that undermine end-to-end encryption function as de facto backdoors, expanding the attack surface and allowing adversaries to exploit legally induced monitoring infrastructures. This creates a paradox of safety through insecurity: systems introduced to protect users may instead compromise their privacy, autonomy, and trust. This work does not aim to predict the future, but to raise awareness and help prevent certain futures from materialising.