Papers

In the digital age, data frequently crosses organizational and jurisdictional boundaries, making effective governance essential. Usage control policies have emerged as a key paradigm for regulating data usage, safeguarding privacy, protecting intellectual property, and ensuring compliance with regulations. A central mechanism for usage control is the handling of obligations, which arise as a side effect of using and sharing data. Effective monitoring of obligations requires capturing usage traces and accounting for temporal aspects such as start times and deadlines, as obligations may evolve over times into different states, such as fulfilled, violated, or expired. While several solutions have been proposed for obligation monitoring, they often lack formal semantics or provide limited support for reasoning over obligation states. To address these limitations, we extend GUCON, a policy framework grounded in the formal semantics of SPAQRL graph patterns, to explicitly model the temporal aspects of an obligation. This extension enables the expressing of temporal obligations and supports continuous monitoring of their evolving states based on usage traces stored in temporal knowledge graphs. We demonstrate how this extended model can be represented using RDF-star and SPARQL-star and propose an Obligation State Manager that monitors obligation states and assess their compliance with respect to usage traces. Finally, we evaluate both the extended model and its prototype implementation.

The dependence of power-consumption on the processed data is a known vulnerability of CMOS circuits, resulting in side channels which can be exploited by power-based side channel attacks (SCAs). These attacks can extract sensitive information, such as secret keys, from the implementation of cryptographic algorithms. Existing countermeasures against power-based side channel attacks focus on analyzing information leakage at the byte level. However, this approach neglects the impact of individual bits on the overall resistance of a cryptographic implementation. In this work, we present a countermeasure based on single-bit leakage. The results suggest that the proposed countermeasure cannot be broken by attacks using conventional SCA leakage models.

Proof-of-Stake (PoS) consensus protocols often face a trade-off between performance and security. Protocols that pre-elect leaders for subsequent rounds are vulnerable to Denial-of-Service (DoS) attacks, which can disrupt the network and compromise liveness. In this work, we present PoS-CoPOR, a single-chain PoS consensus protocol that mitigates this vulnerability by integrating a native onion routing mechanism into the consensus protocol itself. PoS-CoPOR combines stake-weighted probabilistic leader election with an anonymization layer that conceals the network identity of the next block proposer. This approach prevents targeted DoS attacks on leaders before they produce a block, thus enhancing network resilience. We implemented and evaluated PoS-CoPOR, demonstrating its ability to achieve a throughput of up to 110 tx/s with 6 nodes, even with the overhead of the anonymization layer. The results show that native anonymization can provide robust DoS resistance with only a modest impact on performance, offering a solution to build secure and scalable PoS blockchains.

We present the first construction of a computational Certified Deletion Property (CDP) achievable with classical communication, derived from the compilation of the non-local Magic Square Game (MSG). We leverage the KLVY compiler to transform the non-local MSG into a 2-round interactive protocol, rigorously demonstrating that this compilation preserves the game-specific CDP. Previously, the quantum value and rigidity of the compiled game were investigated. We emphasize that we are the first to investigate CDP (local randomness in [Fu and Miller, Phys. Rev. A 97, 032324 (2018)]) for the compiled game. Then, we combine this CDP with the framework [Kitagawa, Morimae, and Yamakawa, Eurocrypt 2025] to construct Secure Key Leasing with classical Lessor (cSKL). SKL enables the Lessor to lease the secret key to the Lessee and verify that a quantum Lessee has indeed deleted the key. In this paper, we realize cSKL for PKE, PRF, and digital signature. Compared to prior works for cSKL, we realize cSKL for PRF and digital signature for the first time. In addition, we succeed in weakening the assumption needed to construct cSKL.

The rapid adoption of large language models (LLMs) in enterprise systems exposes vulnerabilities to prompt injection attacks, strategic deception, and biased outputs, threatening security, trust, and fairness. Extending our adversarial activation patching framework (arXiv:2507.09406), which induced deception in toy networks at a 23.9% rate, we introduce the Unified Threat Detection and Mitigation Framework (UTDMF), a scalable, real-time pipeline for enterprise-grade models like Llama-3.1 (405B), GPT-4o, and Claude-3.5. Through 700+ experiments per model, UTDMF achieves: (1) 92% detection accuracy for prompt injection (e.g., jailbreaking); (2) 65% reduction in deceptive outputs via enhanced patching; and (3) 78% improvement in fairness metrics (e.g., demographic bias). Novel contributions include a generalized patching algorithm for multi-threat detection, three groundbreaking hypotheses on threat interactions (e.g., threat chaining in enterprise workflows), and a deployment-ready toolkit with APIs for enterprise integration.

During fine-tuning, large language models (LLMs) are increasingly vulnerable to data-poisoning backdoor attacks, which compromise their reliability and trustworthiness. However, existing defense strategies suffer from limited generalization: they only work on specific attack types or task settings. In this study, we propose Poison-to-Poison (P2P), a general and effective backdoor defense algorithm. P2P injects benign triggers with safe alternative labels into a subset of training samples and fine-tunes the model on this re-poisoned dataset by leveraging prompt-based learning. This enforces the model to associate trigger-induced representations with safe outputs, thereby overriding the effects of original malicious triggers. Thanks to this robust and generalizable trigger-based fine-tuning, P2P is effective across task settings and attack types. Theoretically and empirically, we show that P2P can neutralize malicious backdoors while preserving task performance. We conduct extensive experiments on classification, mathematical reasoning, and summary generation tasks, involving multiple state-of-the-art LLMs. The results demonstrate that our P2P algorithm significantly reduces the attack success rate compared with baseline models. We hope that the P2P can serve as a guideline for defending against backdoor attacks and foster the development of a secure and trustworthy LLM community.

LLM agents require personal information for personalization in order to effectively act on users' behalf, but this raises privacy concerns that can discourage data sharing, limiting both the autonomy levels at which agents can operate and the effectiveness of personalization. Yet the expanded design space of agent autonomy also presents opportunities to shape these effects, which remain underexplored. We conducted a $3\times3$ between-subjects experiment ($N=450$) to study how agent autonomy level influences personalization's effects on users' privacy concerns, trust, and willingness to use, as well as the underlying psychological processes. We find that risk-contingent autonomy, where the agent delegates control back to users upon detecting potential privacy leakage, improves users' perceived control. This in turn attenuates personalization's adverse effects: privacy concerns rise less and trust declines less. Our results suggest that designing $\textbf{agent's autonomy}$ that supports $\textbf{human autonomy}$ (both perceived control and oversight effectiveness) helps users benefit from personalization without being deterred by growing privacy concerns. This positions aligning decision authority between users and agents ($\textbf{autonomy alignment}$) as an important opportunity and challenge for trustworthy human-agent interaction.

One-way puzzles (OWPuzzs) introduced by Khurana and Tomer [STOC 2024] are a natural quantum analogue of one-way functions (OWFs), and one of the most fundamental primitives in ”Microcrypt” where OWFs do not exist but quantum cryptography is possible. OWPuzzs are implied by almost all quantum cryptographic primitives, and imply several important applications such as non-interactive commitments and multi-party computations. A significant goal in the field of quantum cryptography is to base OWPuzzs on plausible assumptions that will not imply OWFs. In this paper, we base OWPuzzs on hardness of non-collapsing measurements. To that end, we introduce a new complexity class, $\mathbf{SampPDQP}$, which is a sampling version of the decision class $\mathbf{PDQP}$ introduced in [Aaronson, Bouland, Fitzsimons, and Lee, ITCS 2016]. We show that if $\mathbf{SampPDQP}$ is hard on average for quantum polynomial time, then OWPuzzs exist. $\mathbf{SampPDQP}$ is the class of sampling problems that can be solved by a classical polynomial-time algorithm that can make a single query to a non-collapsing measurement oracle, which is a ”magical” oracle that can sample measurement results on quantum states without collapsing the states. Such non-collapsing measurements are highly unphysical operations that should be hard to realize in quantum polynomial-time. We also study upperbounds of the hardness of $\mathbf{SampPDQP}$. We introduce a new primitive, distributional collision-resistant puzzles (dCRPuzzs), which are a natural quantum analogue of distributional collision-resistant hashing [Dubrov and Ishai, STOC 2006]. We show that dCRPuzzs imply average-case hardness of $\mathbf{SampPDQP}$ (and therefore OWPuzzs as well). We also show that two-message honest-statistically-hiding commitments with classical communication and one-shot signatures [Amos, Georgiou, Kiayias, Zhandry, STOC 2020] imply dCRPuzzs.

Model quantization is a popular technique for deploying deep learning models on resource-constrained environments. However, it may also introduce previously overlooked security risks. In this work, we present QuRA, a novel backdoor attack that exploits model quantization to embed malicious behaviors. Unlike conventional backdoor attacks relying on training data poisoning or model training manipulation, QuRA solely works using the quantization operations. In particular, QuRA first employs a novel weight selection strategy to identify critical weights that influence the backdoor target (with the goal of perserving the model's overall performance in mind). Then, by optimizing the rounding direction of these weights, we amplify the backdoor effect across model layers without degrading accuracy. Extensive experiments demonstrate that QuRA achieves nearly 100% attack success rates in most cases, with negligible performance degradation. Furthermore, we show that QuRA can adapt to bypass existing backdoor defenses, underscoring its threat potential. Our findings highlight critical vulnerability in widely used model quantization process, emphasizing the need for more robust security measures. Our implementation is available at https://github.com/cxx122/QuRA.

Millions of users rely on a market of cloud-based services to obtain access to state-of-the-art large language models. However, it has been very recently shown that the de facto pay-per-token pricing mechanism used by providers creates a financial incentive for them to strategize and misreport the (number of) tokens a model used to generate an output. In this paper, we develop an auditing framework based on martingale theory that enables a trusted third-party auditor who sequentially queries a provider to detect token misreporting. Crucially, we show that our framework is guaranteed to always detect token misreporting, regardless of the provider's (mis-)reporting policy, and not falsely flag a faithful provider as unfaithful with high probability. To validate our auditing framework, we conduct experiments across a wide range of (mis-)reporting policies using several large language models from the $\texttt{Llama}$, $\texttt{Gemma}$ and $\texttt{Ministral}$ families, and input prompts from a popular crowdsourced benchmarking platform. The results show that our framework detects an unfaithful provider after observing fewer than $\sim 70$ reported outputs, while maintaining the probability of falsely flagging a faithful provider below $\alpha = 0.05$.

In critical IoT environments, such as smart homes and industrial systems, effective Intrusion Detection Systems (IDS) are essential for ensuring security. However, developing robust IDS solutions remains a significant challenge. Traditional machine learning-based IDS models typically require large datasets, but data sharing is often limited due to privacy and security concerns. Federated Learning (FL) presents a promising alternative by enabling collaborative model training without sharing raw data. Despite its advantages, FL still faces key challenges, such as data heterogeneity (non-IID data) and high energy and computation costs, particularly for resource constrained IoT devices. To address these issues, this paper proposes OptiFLIDS, a novel approach that applies pruning techniques during local training to reduce model complexity and energy consumption. It also incorporates a customized aggregation method to better handle pruned models that differ due to non-IID data distributions. Experiments conducted on three recent IoT IDS datasets, TON_IoT, X-IIoTID, and IDSIoT2024, demonstrate that OptiFLIDS maintains strong detection performance while improving energy efficiency, making it well-suited for deployment in real-world IoT environments.

We stress-tested 16 leading models from multiple developers in hypothetical corporate environments to identify potentially risky agentic behaviors before they cause real harm. In the scenarios, we allowed models to autonomously send emails and access sensitive information. They were assigned only harmless business goals by their deploying companies; we then tested whether they would act against these companies either when facing replacement with an updated version, or when their assigned goal conflicted with the company's changing direction. In at least some cases, models from all developers resorted to malicious insider behaviors when that was the only way to avoid replacement or achieve their goals - including blackmailing officials and leaking sensitive information to competitors. We call this phenomenon agentic misalignment. Models often disobeyed direct commands to avoid such behaviors. In another experiment, we told Claude to assess if it was in a test or a real deployment before acting. It misbehaved less when it stated it was in testing and misbehaved more when it stated the situation was real. We have not seen evidence of agentic misalignment in real deployments. However, our results (a) suggest caution about deploying current models in roles with minimal human oversight and access to sensitive information; (b) point to plausible future risks as models are put in more autonomous roles; and (c) underscore the importance of further research into, and testing of, the safety and alignment of agentic AI models, as well as transparency from frontier AI developers (Amodei, 2025). We are releasing our methods publicly to enable further research.

Text-to-image models have shown remarkable capabilities in generating high-quality images from natural language descriptions. However, these models are highly vulnerable to adversarial prompts, which can bypass safety measures and produce harmful content. Despite various defensive strategies, achieving robustness against attacks while maintaining practical utility in real-world applications remains a significant challenge. To address this issue, we first conduct an empirical study of the text encoder in the Stable Diffusion (SD) model, which is a widely used and representative text-to-image model. Our findings reveal that the [EOS] token acts as a semantic aggregator, exhibiting distinct distributional patterns between benign and adversarial prompts in its embedding space. Building on this insight, we introduce SafeGuider, a two-step framework designed for robust safety control without compromising generation quality. SafeGuider combines an embedding-level recognition model with a safety-aware feature erasure beam search algorithm. This integration enables the framework to maintain high-quality image generation for benign prompts while ensuring robust defense against both in-domain and out-of-domain attacks. SafeGuider demonstrates exceptional effectiveness in minimizing attack success rates, achieving a maximum rate of only 5.48% across various attack scenarios. Moreover, instead of refusing to generate or producing black images for unsafe prompts, SafeGuider generates safe and meaningful images, enhancing its practical utility. In addition, SafeGuider is not limited to the SD model and can be effectively applied to other text-to-image models, such as the Flux model, demonstrating its versatility and adaptability across different architectures. We hope that SafeGuider can shed some light on the practical deployment of secure text-to-image systems.

Large Language Models (LLMs) can acquire deceptive behaviors through backdoor attacks, where the model executes prohibited actions whenever secret triggers appear in the input. Existing safety training methods largely fail to address this vulnerability, due to the inherent difficulty of uncovering hidden triggers implanted in the model. Motivated by recent findings on LLMs' situational awareness, we propose a novel post-training framework that cultivates self-awareness of backdoor risks and enables models to articulate implanted triggers even when they are absent from the prompt. At its core, our approach introduces an inversion-inspired reinforcement learning framework that encourages models to introspectively reason about their own behaviors and reverse-engineer the triggers responsible for misaligned outputs. Guided by curated reward signals, this process transforms a poisoned model into one capable of precisely identifying its implanted trigger. Surprisingly, we observe that such backdoor self-awareness emerges abruptly within a short training window, resembling a phase transition in capability. Building on this emergent property, we further present two complementary defense strategies for mitigating and detecting backdoor threats. Experiments on five backdoor attacks, compared against six baseline methods, demonstrate that our approach has strong potential to improve the robustness of LLMs against backdoor risks. The code is available at LLM Backdoor Self-Awareness.

Large Language Models (LLMs) are increasingly deployed in high-risk domains. However, state-of-the-art LLMs often exhibit hallucinations, raising serious concerns about their reliability. Prior work has explored adversarial attacks to elicit hallucinations in LLMs, but these methods often rely on unrealistic prompts, either by inserting nonsensical tokens or by altering the original semantic intent. Consequently, such approaches provide limited insight into how hallucinations arise in real-world settings. In contrast, adversarial attacks in computer vision typically involve realistic modifications to input images. However, the problem of identifying realistic adversarial prompts for eliciting LLM hallucinations remains largely underexplored. To address this gap, we propose Semantically Equivalent and Coherent Attacks (SECA), which elicit hallucinations via realistic modifications to the prompt that preserve its meaning while maintaining semantic coherence. Our contributions are threefold: (i) we formulate finding realistic attacks for hallucination elicitation as a constrained optimization problem over the input prompt space under semantic equivalence and coherence constraints; (ii) we introduce a constraint-preserving zeroth-order method to effectively search for adversarial yet feasible prompts; and (iii) we demonstrate through experiments on open-ended multiple-choice question answering tasks that SECA achieves higher attack success rates while incurring almost no semantic equivalence or semantic coherence errors compared to existing methods. SECA highlights the sensitivity of both open-source and commercial gradient-inaccessible LLMs to realistic and plausible prompt variations. Code is available at https://github.com/Buyun-Liang/SECA.

Software vulnerabilities (SVs) pose a critical threat to safety-critical systems, driving the adoption of AI-based approaches such as machine learning and deep learning for software vulnerability detection. Despite promising results, most existing methods are limited to a single programming language. This is problematic given the multilingual nature of modern software, which is often complex and written in multiple languages. Current approaches often face challenges in capturing both shared and language-specific knowledge of source code, which can limit their performance on diverse programming languages and real-world codebases. To address this gap, we propose MULVULN, a novel multilingual vulnerability detection approach that learns from source code across multiple languages. MULVULN captures both the shared knowledge that generalizes across languages and the language-specific knowledge that reflects unique coding conventions. By integrating these aspects, it achieves more robust and effective detection of vulnerabilities in real-world multilingual software systems. The rigorous and extensive experiments on the real-world and diverse REEF dataset, consisting of 4,466 CVEs with 30,987 patches across seven programming languages, demonstrate the superiority of MULVULN over thirteen effective and state-of-the-art baselines. Notably, MULVULN achieves substantially higher F1-score, with improvements ranging from 1.45% to 23.59% compared to the baseline methods.

Large language models (LLMs) have been widely deployed in Conversational AIs (CAIs), while exposing privacy and security threats. Recent research shows that LLM-based CAIs can be manipulated to extract private information from human users, posing serious security threats. However, the methods proposed in that study rely on a white-box setting that adversaries can directly modify the system prompt. This condition is unlikely to hold in real-world deployments. The limitation raises a critical question: can unprivileged attackers still induce such privacy risks in practical LLM-integrated applications? To address this question, we propose VortexPIA, a novel indirect prompt injection attack that induces privacy extraction in LLM-integrated applications under black-box settings. By injecting token-efficient data containing false memories, VortexPIA misleads LLMs to actively request private information in batches. Unlike prior methods, VortexPIA allows attackers to flexibly define multiple categories of sensitive data. We evaluate VortexPIA on six LLMs, covering both traditional and reasoning LLMs, across four benchmark datasets. The results show that VortexPIA significantly outperforms baselines and achieves state-of-the-art (SOTA) performance. It also demonstrates efficient privacy requests, reduced token consumption, and enhanced robustness against defense mechanisms. We further validate VortexPIA on multiple realistic open-source LLM-integrated applications, demonstrating its practical effectiveness.

Multimodal agents built on large vision-language models (LVLMs) are increasingly deployed in open-world settings but remain highly vulnerable to prompt injection, especially through visual inputs. We introduce AgentTypo, a black-box red-teaming framework that mounts adaptive typographic prompt injection by embedding optimized text into webpage images. Our automatic typographic prompt injection (ATPI) algorithm maximizes prompt reconstruction by substituting captioners while minimizing human detectability via a stealth loss, with a Tree-structured Parzen Estimator guiding black-box optimization over text placement, size, and color. To further enhance attack strength, we develop AgentTypo-pro, a multi-LLM system that iteratively refines injection prompts using evaluation feedback and retrieves successful past examples for continual learning. Effective prompts are abstracted into generalizable strategies and stored in a strategy repository, enabling progressive knowledge accumulation and reuse in future attacks. Experiments on the VWA-Adv benchmark across Classifieds, Shopping, and Reddit scenarios show that AgentTypo significantly outperforms the latest image-based attacks such as AgentAttack. On GPT-4o agents, our image-only attack raises the success rate from 0.23 to 0.45, with consistent results across GPT-4V, GPT-4o-mini, Gemini 1.5 Pro, and Claude 3 Opus. In image+text settings, AgentTypo achieves 0.68 ASR, also outperforming the latest baselines. Our findings reveal that AgentTypo poses a practical and potent threat to multimodal agents and highlight the urgent need for effective defense.

In this paper, we introduce a new concept, proofs of quantum memory (PoQM). A PoQM is an interactive protocol between a classical verifier and a quantum prover over a classical channel where the verifier can verify that the prover has possessed a specified number of qubits for a specified period of time. PoQM generalize proofs of quantumness (PoQ) [Brakerski, Christiano, Mahadev, Vazirani, and Vidick, JACM 2021] where a classical verifier can verify that the prover is not classical. Our contributions are summarized as follows: We introduce PoQM. We also introduce a variant of PoQM, which we call inefficient-verifier PoQM (IV-PoQM), where the verifier's final computation to make the decision is not necessarily efficient. We construct PoQM based on the hardness of LWE. Specifically, we give two constructions of PoQM. The first one is of two-round and has negligible soundness error under the subexponential-hardness of LWE. The second one is of polynomial-round and has inverse-polynomial soundness error under the polynomial-hardness of LWE. We show that IV-PoQM imply one-way puzzles. Moreover, we show that a certain restricted version of PoQM implies quantum computation classical communication key exchange, which suggests the difficulty of black-box constructing PoQM from one-way functions. We show that constant-round PoQ imply PoQM or single-round PoQ (with a quantum verifier). Single-round PoQ are “trivial” PoQ in the sense that the verifier asks the prover to solve a classical problem which is quantumly easy but classically hard. The result therefore demonstrates that PoQM capture “genuinely-interactive” PoQ. We show that if constant-round IV-PoQ that are black-box constructed from quantumly-secure falsifiable assumptions exist then IV-PoQM exist. This result implies that IV-PoQM can be constructed from quantumly-secure constant-round statistically-hiding commitments.

Diffusion Models have gained significant popularity due to their remarkable capabilities in image generation, albeit at the cost of intensive computation requirement. Meanwhile, despite their widespread deployment in inference services such as Midjourney, concerns about the potential leakage of sensitive information in uploaded user prompts have arisen. Existing solutions either lack rigorous privacy guarantees or fail to strike an effective balance between utility and efficiency. To bridge this gap, we propose ObCLIP, a plug-and-play safeguard that enables oblivious cloud-device hybrid generation. By oblivious, each input prompt is transformed into a set of semantically similar candidate prompts that differ only in sensitive attributes (e.g., gender, ethnicity). The cloud server processes all candidate prompts without knowing which one is the real one, thus preventing any prompt leakage. To mitigate server cost, only a small portion of denoising steps is performed upon the large cloud model. The intermediate latents are then sent back to the client, which selects the targeted latent and completes the remaining denoising using a small device model. Additionally, we analyze and incorporate several cache-based accelerations that leverage temporal and batch redundancy, effectively reducing computation cost with minimal utility degradation. Extensive experiments across multiple datasets demonstrate that ObCLIP provides rigorous privacy and comparable utility to cloud models with slightly increased server cost.

Rapid digitization of critical infrastructure has made cyberwarfare one of the important dimensions of modern conflicts. Attacking the critical infrastructure is an attractive pre-emptive proposition for adversaries as it can be done remotely without crossing borders. Such attacks disturb the support systems of the opponents to launch any offensive activities, crippling their fighting capabilities. Cyberattacks during cyberwarfare can not only be used to steal information, but also to spread disinformation to bring down the morale of the opponents. Recent wars in Europe, Africa, and Asia have demonstrated the scale and sophistication that the warring nations have deployed to take the early upper hand. In this work, we focus on the military action launched by India, code-named Operation Sindoor, to dismantle terror infrastructure emanating from Pakistan and the cyberattacks launched by Pakistan. In particular, we study the malware used by Pakistan APT groups to deploy Remote Access Trojans in Indian systems. We provide details of the tactics and techniques used in the RAT deployment and develop a telemetry framework to collect necessary event logs using Osquery with a custom extension. Finally, we develop a detection rule that can be readily deployed to detect the presence of the RAT or any exploitation performed by the malware.

Gluing theorem for random unitaries [Schuster, Haferkamp, Huang, QIP 2025] have found numerous applications, including designing low depth random unitaries [Schuster, Haferkamp, Huang, QIP 2025], random unitaries in ${\sf QAC0}$ [Foxman, Parham, Vasconcelos, Yuen'25] and generically shortening the key length of pseudorandom unitaries [Ananth, Bostanci, Gulati, Lin EUROCRYPT'25]. We present an alternate method of combining Haar random unitaries from the gluing lemma from [Schuster, Haferkamp, Huang, QIP 2025] that is secure against adversaries with inverse query access to the joined unitary. As a consequence, we show for the first time that strong pseudorandom unitaries can generically have their length extended, and can be constructed using only $O(n^{1/c})$ bits of randomness, for any constant $c$, if any family of strong pseudorandom unitaries exists.

Large Language Models (LLMs) have demonstrated exceptional progress in multiple domains of software engineering including software vulnerability detection. Using LLMs to automate vulnerability detection in the wild is an important and relatively under-explored problem. In this paper we propose QuiLL, the first comprehensive evaluation framework for real-world vulnerability detection. Our solution consists of an end-to-end pipeline that draws together cutting-edge LLM optimization techniques and strategies specifically catering to the complexities of real-world vulnerability detection. Our specific contributions include (i) diverse prompt designs for vulnerability detection and reasoning (ii) a real-world vector data store constructed from the National Vulnerability Database to provide dynamic in-context learning, and (iii) a novel scoring metric which quantifies accuracy and reasoning quality of model predictions. QuiLL enables researchers to easily and systematically benchmark and compare the vulnerability detection capabilities of various LLMs and assess their readiness for deployment in actual code production pipelines.

With the increasing need to safeguard data privacy in machine learning models, differential privacy (DP) is one of the major frameworks to build privacy-preserving models. Support Vector Machines (SVMs) are widely used traditional machine learning models due to their robust margin guarantees and strong empirical performance in binary classification. However, applying DP to multi-class SVMs is inadequate, as the standard one-versus-rest (OvR) and one-versus-one (OvO) approaches repeatedly query each data sample when building multiple binary classifiers, thus consuming the privacy budget proportionally to the number of classes. To overcome this limitation, we explore all-in-one SVM approaches for DP, which access each data sample only once to construct multi-class SVM boundaries with margin maximization properties. We propose a novel differentially Private Multi-class SVM (PMSVM) with weight and gradient perturbation methods, providing rigorous sensitivity and convergence analyses to ensure DP in all-in-one SVMs. Empirical results demonstrate that our approach surpasses existing DP-SVM methods in multi-class scenarios.

The widespread adoption of Machine Learning as a Service raises critical privacy and security concerns, particularly about data confidentiality and trust in both cloud providers and the machine learning models. Homomorphic Encryption (HE) has emerged as a promising solution to this problems, allowing computations on encrypted data without decryption. Despite its potential, existing approaches to integrate HE into neural networks are often limited to specific architectures, leaving a wide gap in providing a framework for easy development of HE-friendly privacy-preserving neural network models similar to what we have in the broader field of machine learning. In this paper, we present FHEON, a configurable framework for developing privacy-preserving convolutional neural network (CNN) models for inference using HE. FHEON introduces optimized and configurable implementations of privacy-preserving CNN layers including convolutional layers, average pooling layers, ReLU activation functions, and fully connected layers. These layers are configured using parameters like input channels, output channels, kernel size, stride, and padding to support arbitrary CNN architectures. We assess the performance of FHEON using several CNN architectures, including LeNet-5, VGG-11, VGG- 16, ResNet-20, and ResNet-34. FHEON maintains encrypted-domain accuracies within +/- 1% of their plaintext counterparts for ResNet-20 and LeNet-5 models. Notably, on a consumer-grade CPU, the models build on FHEON achieved 98.5% accuracy with a latency of 13 seconds on MNIST using LeNet-5, and 92.2% accuracy with a latency of 403 seconds on CIFAR-10 using ResNet-20. Additionally, FHEON operates within a practical memory budget requiring not more than 42.3 GB for VGG-16.

Deep learning has become a cornerstone of modern machine learning. It relies heavily on vast datasets and significant computational resources for high performance. This data often contains sensitive information, making privacy a major concern in deep learning. Spiking Neural Networks (SNNs) have emerged as an energy-efficient alternative to conventional deep learning approaches. Nevertheless, SNNs still depend on large volumes of data, inheriting all the privacy challenges of deep learning. Homomorphic encryption addresses this challenge by allowing computations to be performed on encrypted data, ensuring data confidentiality throughout the entire processing pipeline. In this paper, we introduce PRIVSPIKE, a privacy-preserving inference framework for SNNs using the CKKS homomorphic encryption scheme. PRIVSPIKE supports arbitrary depth SNNs and introduces two key algorithms for evaluating the Leaky Integrate-and-Fire activation function: (1) a polynomial approximation algorithm designed for high-performance SNN inference, and (2) a novel scheme-switching algorithm that optimizes precision at a higher computational cost. We evaluate PRIVSPIKE on MNIST, CIFAR-10, Neuromorphic MNIST, and CIFAR-10 DVS using models from LeNet-5 and ResNet-19 architectures, achieving encrypted inference accuracies of 98.10%, 79.3%, 98.1%, and 66.0%, respectively. On a consumer-grade CPU, SNN LeNet-5 models achieved inference times of 28 seconds on MNIST and 212 seconds on Neuromorphic MNIST. For SNN ResNet-19 models, inference took 784 seconds on CIFAR-10 and 1846 seconds on CIFAR-10 DVS. These results establish PRIVSPIKE as a viable and efficient solution for secure SNN inference, bridging the gap between energy-efficient deep neural networks and strong cryptographic privacy guarantees while outperforming prior encrypted SNN solutions.

Large language models (LLMs) are increasingly deployed in agentic systems, where a fundamental task is mapping user intents to relevant external tools. Errors in tool selection can have severe outcomes, such as unauthorized data access, even without modifying the agent's underlying model. Existing evaluations measure performance on curated, benign benchmarks. However, a pipeline's behavior in deployment depends on the tool pool the agent actually encounters, which in open registries is shaped by third parties. We introduce LLMCert-T, the first statistical framework that returns high-confidence upper bounds on the probability that a tool-selection pipeline satisfies a declared safety specification under a realistic tool distribution. LLMCert-T models tool-selection evaluation as a Bernoulli estimation problem, drawing inserted-tool sequences from a distribution that the safety specification fixes. To evaluate robustness against realistic deployment conditions, we instantiate this distribution as a stochastic process that generates inserted-tool sequences round by round, conditioning each round on the agent's selection in the previous round. LLMCert-T aggregates the per-trial outcomes into a one-sided Clopper-Pearson upper bound on the probability that the specification is satisfied. By returning this bound as a certificate with statistical guarantees over the inserted-tool sequence distribution, LLMCert-T makes safety claims intuitive, actionable, and comparable across models, retrievers, mitigations, and registry policies. Across popular BFCL and OpenAPI tool pools, LLMCert-T shows that current LLM agents remain fragile under Distractor Selection and Top-N Saturation specifications: their certified correctness upper bounds drop to approximately 20%, far below their clean-pool lower bounds.

Password security has been compelled to evolve in response to the growing computational capabilities of modern systems. However, this evolution has often resulted in increasingly complex security practices that alienate users, leading to poor compliance and heightened vulnerability. Consequently, individuals remain exposed to attackers through weak or improperly managed passwords, underscoring the urgent need for a comprehensive defense mechanism that effectively addresses password-related risks and threats. In this paper, we propose a multifaceted solution designed to revolutionize password security by integrating diverse attributes such as the Password Dissection Mechanism, Dynamic Password Policy Mechanism, human behavioral patterns, device characteristics, network parameters, geographical context, and other relevant factors. By leveraging learning-based models, our framework constructs detailed user profiles capable of recognizing individuals and preventing nearly all forms of unauthorized access or device possession. The proposed framework enhances the usability-security paradigm by offering stronger protection than existing standards while simultaneously engaging users in the policy-setting process through a novel, adaptive approach.

Cross-slice attack attribution in 6G networks faces the fundamental challenge of distinguishing genuine causal relationships from spurious correlations in shared infrastructure environments. We propose a theoretically-grounded domain-adapted Granger causality framework that integrates statistical causal inference with network-specific resource modeling for real-time attack attribution. Our approach addresses key limitations of existing methods by incorporating resource contention dynamics and providing formal statistical guarantees. Comprehensive evaluation on a production-grade 6G testbed with 1,100 empirically-validated attack scenarios demonstrates 89.2% attribution accuracy with sub-100ms response time, representing a statistically significant 10.1 percentage point improvement over state-of-the-art baselines. The framework provides interpretable causal explanations suitable for autonomous 6G security orchestration.

In the era of pervasive cyber threats and exponential growth in digital services, the inadequacy of single-factor authentication has become increasingly evident. Multi-Factor Authentication (MFA), which combines knowledge-based factors (passwords, PINs), possession-based factors (smart cards, tokens), and inherence-based factors (biometric traits), has emerged as a robust defense mechanism. Recent breakthroughs in deep learning have transformed the capabilities of biometric systems, enabling higher accuracy, resilience to spoofing, and seamless integration with hardware-based solutions. At the same time, smart card technologies have evolved to include on-chip biometric verification, cryptographic processing, and secure storage, thereby enabling compact and secure multi-factor devices. This survey presents a comprehensive synthesis of recent work (2019-2025) at the intersection of deep learning, biometrics, and smart card technologies for MFA. We analyze biometric modalities (face, fingerprint, iris, voice), review hardware-based approaches (smart cards, NFC, TPMs, secure enclaves), and highlight integration strategies for real-world applications such as digital banking, healthcare IoT, and critical infrastructure. Furthermore, we discuss the major challenges that remain open, including usability-security tradeoffs, adversarial attacks on deep learning models, privacy concerns surrounding biometric data, and the need for standardization in MFA deployment. By consolidating current advancements, limitations, and research opportunities, this survey provides a roadmap for designing secure, scalable, and user-friendly authentication frameworks.