Papers

Transformer-based models are highly susceptible to backdoor attacks via supervised fine-tuning (SFT). To red-team existing data-poisoning defenses, prior work has increasingly focused on stylized triggers, synthetic artifacts, and token-level perturbations designed to evade detection. However, this trend has shifted threat models away from naturally occurring semantic triggers and realistic low-budget poisoning settings. Addressing this gap, we introduce SteganoBackdoor, an optimization-based framework that transforms semantic-trigger seeds through autoregressive token replacement, sequentially minimizing embedding overlap with the inference-time trigger while preserving a strong per-sample training-time payload. The resulting SteganoPoisons maintain linguistic fluency and encode the payload across ordinary tokens, such that no individual token carries a concentrated signal and the full payload instead emerges from their exact combination and ordering. Across 18 encoder-based and decoder-only models spanning 120M to 14B parameters, SteganoBackdoor achieves high attack success under sub-percent poisoning budgets and exposes limitations in existing data-poisoning defenses.

Evaluating the safety robustness of LLMs is critical for their deployment. However, mainstream Red Teaming methods rely on online generation and black-box output analysis. These approaches are not only costly but also suffer from feedback latency, making them unsuitable for agile diagnostics after training a new model. To address this, we propose N-GLARE (A Non-Generative, Latent Representation-Efficient LLM Safety Evaluator). N-GLARE operates entirely on the model's latent representations, bypassing the need for full text generation. It characterizes hidden layer dynamics by analyzing the APT (Angular-Probabilistic Trajectory) of latent representations and introducing the JSS (Jensen-Shannon Separability) metric. Experiments on over 40 models and 20 red teaming strategies demonstrate that the JSS metric exhibits high consistency with the safety rankings derived from Red Teaming. N-GLARE reproduces the discriminative trends of large-scale red-teaming tests at less than 1% of the token cost and the runtime cost, providing an efficient output-free evaluation proxy for real-time diagnostics.

As the use of large language models (LLMs) continues to expand, ensuring their safety and robustness has become a critical challenge. In particular, jailbreak attacks that bypass built-in safety mechanisms are increasingly recognized as a tangible threat across industries, driving the need for diverse templates to support red-teaming efforts and strengthen defensive techniques. However, current approaches predominantly rely on two limited strategies: (i) substituting harmful queries into fixed templates, and (ii) having the LLM generate entire templates, which often compromises intent clarity and reproductibility. To address this gap, this paper introduces the Embedded Jailbreak Template, which preserves the structure of existing templates while naturally embedding harmful queries within their context. We further propose a progressive prompt-engineering methodology to ensure template quality and consistency, alongside standardized protocols for generation and evaluation. Together, these contributions provide a benchmark that more accurately reflects real-world usage scenarios and harmful intent, facilitating its application in red-teaming and policy regression testing.

With the ever-growing demand for cybersecurity, static key encryption mechanisms are increasingly vulnerable to adversarial attacks due to their deterministic and non-adaptive nature. Brute-force attacks, key compromise, and unauthorized access have become highly common cyber threats. This research presents a novel fuzzy logic-based cryptographic framework that dynamically generates encryption keys in real-time by accessing system-level entropy and hardware-bound trust. The proposed system leverages a Fuzzy Inference System (FIS) to evaluate system parameters that include CPU utilization, process count, and timestamp variation. It assigns entropy level based on linguistically defined fuzzy rules which are fused with hardware-generated randomness and then securely sealed using a Trusted Platform Module (TPM). The sealed key is incorporated in an AES-GCM encryption scheme to ensure both confidentiality and integrity of the data. This system introduces a scalable solution for adaptive encryption in high-assurance computing, zero-trust environments, and cloud-based infrastructure.

Fine-grained identification of IDS-flagged suspicious traffic is crucial in cybersecurity. In practice, cyber threats evolve continuously, making the discovery of novel malicious traffic a critical necessity as well as the identification of known classes. Recent studies have advanced this goal with deep models, but they often rely on task-specific architectures that limit transferability and require per-dataset tuning. In this paper we introduce MalRAG, the first LLM driven retrieval-augmented framework for open-set malicious traffic identification. MalRAG freezes the LLM and operates via comprehensive traffic knowledge construction, adaptive retrieval, and prompt engineering. Concretely, we construct a multi-view traffic database by mining prior malicious traffic from content, structural, and temporal perspectives. Furthermore, we introduce a Coverage-Enhanced Retrieval Algorithm that queries across these views to assemble the most probable candidates, thereby improving the inclusion of correct evidence. We then employ Traffic-Aware Adaptive Pruning to select a variable subset of these candidates based on traffic-aware similarity scores, suppressing incorrect matches and yielding reliable retrieved evidence. Moreover, we develop a suite of guidance prompts where task instruction, evidence referencing, and decision guidance are integrated with the retrieved evidence to improve LLM performance. Across diverse real-world datasets and settings, MalRAG delivers state-of-the-art results in both fine-grained identification of known classes and novel malicious traffic discovery. Ablation and deep-dive analyses further show that MalRAG effective leverages LLM capabilities yet achieves open-set malicious traffic identification without relying on a specific LLM.

Run-time integrity enforcement in real-time systems presents a fundamental conflict with availability. Existing approaches in real-time systems primarily focus on minimizing the execution-time overhead of monitoring. After a violation is detected, prior works face a trade-off: (1) prioritize availability and allow a compromised system to continue to ensure applications meet their deadlines, or (2) prioritize security by generating a fault to abort all execution. In this work, we propose PAIR, an approach that offers a middle ground between the stark extremes of this trade-off. PAIR monitors real-time tasks for run-time integrity violations and maintains an Availability Region (AR) of all tasks that are safe to continue. When a task causes a violation, PAIR triggers a non-maskable interrupt to kill the task and continue executing a non-violating task within AR. Thus, PAIR ensures only violating tasks are prevented from execution, while granting availability to remaining tasks. With its hardware approach, PAIR does not cause any run-time overhead to the executing tasks, integrates with real-time operating systems (RTOSs), and is affordable to low-end microcontroller units (MCUs) by incurring +2.3% overhead in memory and hardware usage.

Differential privacy (DP) auditing is essential for evaluating privacy guarantees in machine learning systems. Existing auditing methods, however, pose a significant challenge for large-scale systems since they require modifying the training dataset – for instance, by injecting out-of-distribution canaries or removing samples from training. Such interventions on the training data pipeline are resource-intensive and involve considerable engineering overhead. We introduce a novel observational auditing framework that leverages the inherent randomness of data distributions, enabling privacy evaluation without altering the original dataset. Our approach extends privacy auditing beyond traditional membership inference to protected attributes, with labels as a special case, addressing a key gap in existing techniques. We provide theoretical foundations for our method and perform experiments on Criteo and CIFAR-10 datasets that demonstrate its effectiveness in auditing label privacy guarantees. This work opens new avenues for practical privacy auditing in large-scale production environments.

Sensor data-based recognition systems are widely used in various applications, such as gait-based authentication and human activity recognition (HAR). Modern wearable and smart devices feature various built-in Inertial Measurement Unit (IMU) sensors, and such sensor-based measurements can be fed to a machine learning-based model to train and classify human activities. While deep learning-based models have proven successful in classifying human activity and gestures, they pose various security risks. In our paper, we discuss a novel dynamic trigger-generation technique for performing black-box adversarial attacks on sensor data-based IoT systems. Our empirical analysis shows that the attack is successful on various datasets and classifier models with minimal perturbation on the input data. We also provide a detailed comparative analysis of performance and stealthiness to various other poisoning techniques found in backdoor attacks. We also discuss some adversarial defense mechanisms and their impact on the effectiveness of our trigger-generation technique.

Given a circuit $G: \{0, 1\}^n \to \{0, 1\}^m$ with $m > n$, the *range avoidance* problem ($\text{Avoid}$) asks to output a string $y\in \{0, 1\}^m$ that is not in the range of $G$. Besides its profound connection to circuit complexity and explicit construction problems, this problem is also related to the existence of *proof complexity generators* – circuits $G: \{0, 1\}^n \to \{0, 1\}^m$ where $m > n$ but for every $y\in \{0, 1\}^m$, it is infeasible to prove the statement "$y\not\in\mathrm{Range}(G)$" in a given propositional proof system. This paper connects these two problems with the existence of *demi-bits generators*, a fundamental cryptographic primitive against nondeterministic adversaries introduced by Rudich (RANDOM '97). $\bullet$ We show that the existence of demi-bits generators implies $\text{Avoid}$ is hard for nondeterministic algorithms. This resolves an open problem raised by Chen and Li (STOC '24). Furthermore, assuming the demi-hardness of certain LPN-style generators or Goldreich' PRG, we prove the hardness of $\text{Avoid}$ even when the instances are constant-degree polynomials over $\mathbb{F}_2$. $\bullet$ We show that the dual weak pigeonhole principle is unprovable in Cook's theory $\mathsf{PV}_1$ under the existence of demi-bits generators secure against $\mathbf{AM}$, thereby separating Jerabek's theory $\mathsf{APC}_1$ from $\mathsf{PV}_1$. $\bullet$ We transform demi-bits generators to proof complexity generators that are *pseudo-surjective* with nearly optimal parameters. Our constructions build on the recent breakthroughs on the hardness of $\text{Avoid}$ by Ilango, Li, and Williams (STOC '23) and Chen and Li (STOC '24). We use *randomness extractors* to significantly simplify the construction and the proof.

Reinforcement Learning with Verifiable Rewards (RLVR) has become a core training stage in recent large language models (LLMs). Its reliance on non-public, high-value prompt sets raises concerns about unauthorized data use, creating a need for exposure auditing. A natural tool is membership inference attacks (MIAs), but existing methods detect fitting to a fixed target string. This does not apply to RLVR, which generates responses from the model itself and reinforces successful ones, thus hindering the auditing of data exposure. We show that it remains detectable: RLVR reshapes the model's response distribution on training prompts, producing behavioral traces that can be surfaced through targeted auditing. We propose Divergence-in-Behavior Auditing (DIBA), a white-box query-level auditing framework for RLVR. DIBA compares a fine-tuned model against its pre-RLVR checkpoint along two axes: reward-side evidence capturing changes in verifiable task success, and policy-side evidence capturing prompt-conditioned behavioral drift. By aggregating over multiple stochastic rollouts, DIBA produces a stable query-level auditing signal. Under a white-box setting, DIBA consistently outperforms strong transferred likelihood-based baselines, including calibrated and self-generated variants, achieving around 0.8 AUC and an order-of-magnitude stronger TPR@0.1%FPR. We further show that RLVR auditing is stronger when training leaves non-trivial prompt-specific traces and weaker when the base model already performs well on the prompt. Under a practical grey-box setting, transfer is often robust across model sizes under the same RLVR algorithm, but more varied across algorithms, and can remain useful under distribution shift with carefully chosen shadow data.

Digital content distribution and propitiatory research driven industries face persistent risks from intellectual property theft and unauthorized redistribution. Conventional encryption schemes such as AES, TDES, ECC, and ElGamal provide strong cryptographic guarantees, but they remain fundamentally agnostic to where decryption takes place. In practice, this means that once a decryption key is leaked or intercepted, any adversary can misuse the key to decrypt the protected content from any location. This paper presents, GeoGuard, a location-dependent cryptosystem in which the decryption key is not transmitted as data but is implicitly encoded in the precise time-of-flight differences of ultra-wideband (UWB) data transmission packets. The system leverages precise timing hardware and a custom Timing-encoded Cryptographic Keying (TiCK) protocol to map a 32-byte SHA-256 AES key onto scheduled transmission timestamps. Only user located within an approved spatial location can observe the correct packet timing that aligns with the intended packet-reception timing pattern, enabling them to reconstruct the key. Eavesdroppers outside the authorized region observe an incorrect timing pattern, which yields incorrect keys. GeoGuard is designed to encrypt and transmit data, but decryption is only possible when the user is within the authorized area. Our evaluation demonstrates that the system (i) removes the need to share decryption passwords electronically or physically, (ii) ensures the decryption key cannot be recovered by the eavesdropper, and (iii) provides a non-trivial spatial tolerance for legitimate users

Volumetric video has emerged as a key paradigm in eXtended Reality (XR) and immersive multimedia because it enables highly interactive, spatially consistent 3D experiences. However, the transport-layer security for such 3D content remains largely unaddressed. Existing volumetric streaming pipelines inherit uniform encryption schemes from 2D video, overlooking the heterogeneous privacy sensitivity of different geometry and the strict motion-to-photon latency constraints of real-time XR. We take an initial step toward content-aware secure volumetric video delivery by introducing Privis, a saliency-guided transport framework that (i) partitions volumetric assets into independent units, (ii) applies lightweight authenticated encryption with adaptive key rotation, and (iii) employs selective traffic shaping to balance confidentiality and low latency. Privis specifies a generalized transport-layer security architecture for volumetric media, defining core abstractions and adaptive protection mechanisms. We further explore a prototype implementation and present initial latency measurements to illustrate feasibility and design tradeoffs, providing early empirical guidance toward future work on real-time, saliency-conditioned secure delivery.

Certified defenses promise provable robustness guarantees. We study the malicious exploitation of probabilistic certification frameworks to better understand the limits of guarantee provisions. Now, the objective is to not only mislead a classifier, but also manipulate the certification process to generate a robustness guarantee for an adversarial input certificate spoofing. A recent study in ICLR demonstrated that crafting large perturbations can shift inputs far into regions capable of generating a certificate for an incorrect class. Our study investigates if perturbations needed to cause a misclassification and yet coax a certified model into issuing a deceptive, large robustness radius for a target class can still be made small and imperceptible. We explore the idea of region-focused adversarial examples to craft imperceptible perturbations, spoof certificates and achieve certification radii larger than the source class ghost certificates. Extensive evaluations with the ImageNet demonstrate the ability to effectively bypass state-of-the-art certified defenses such as Densepure. Our work underscores the need to better understand the limits of robustness certification methods.

This paper presents a comprehensive comparative analysis of two dominant blockchain consensus mechanisms, Proof of Work (PoW) and Proof of Stake (PoS), evaluated across seven critical metrics: energy use, security, transaction speed, scalability, centralization risk, environmental impact, and transaction fees. Utilizing recent academic research and real-world blockchain data, the study highlights that PoW offers robust, time-tested security but suffers from high energy consumption, slower throughput, and centralization through mining pools. In contrast, PoS demonstrates improved scalability and efficiency, significantly reduced environmental impact, and more stable transaction fees, however it raises concerns over validator centralization and long-term security maturity. The findings underscore the trade-offs inherent in each mechanism and suggest hybrid designs may combine PoW's security with PoS's efficiency and sustainability. The study aims to inform future blockchain infrastructure development by striking a balance between decentralization, performance, and ecological responsibility.

We study the running time, in terms of first order oracle queries, of differentially private empirical/population risk minimization of Lipschitz convex losses. We first consider the setting where the loss is non-smooth and the optimizer interacts with a private proxy oracle, which sends only private messages about a minibatch of gradients. In this setting, we show that expected running time $\Omega(\min\{\frac{\sqrt{d}}{\alpha^2}, \frac{d}{\log(1/\alpha)}\})$ is necessary to achieve $\alpha$ excess risk on problems of dimension $d$ when $d \geq 1/\alpha^2$. Upper bounds via DP-SGD show these results are tight when $d>\tilde{\Omega}(1/\alpha^4)$. We further show our lower bound can be strengthened to $\Omega(\min\{\frac{d}{\bar{m}\alpha^2}, \frac{d}{\log(1/\alpha)} \})$ for algorithms which use minibatches of size at most $\bar{m} < \sqrt{d}$. We next consider smooth losses, where we relax the private oracle assumption and give lower bounds under only the condition that the optimizer is private. Here, we lower bound the expected number of first order oracle calls by $\tilde{\Omega}\big(\frac{\sqrt{d}}{\alpha} + \min\{\frac{1}{\alpha^2}, n\}\big)$, where $n$ is the size of the dataset. Modifications to existing algorithms show this bound is nearly tight. Compared to non-private lower bounds, our results show that differentially private optimizers pay a dimension dependent runtime penalty. Finally, as a natural extension of our proof technique, we show lower bounds in the non-smooth setting for optimizers interacting with information limited oracles. Specifically, if the proxy oracle transmits at most $\Gamma$-bits of information about the gradients in the minibatch, then $\Omega\big(\min\{\frac{d}{\alpha^2\Gamma}, \frac{d}{\log(1/\alpha)}\}\big)$ oracle calls are needed. This result shows fundamental limitations of gradient quantization techniques in optimization.

Metasurfaces, or Reconfigurable Intelligent Surfaces (RISs), have emerged as a transformative technology for next-generation wireless systems, enabling digitally controlled manipulation of electromagnetic wave propagation. By turning the traditionally passive radio environment into a smart, programmable medium, metasurfaces promise advances in communication and sensing. However, metasurfaces also present a new security frontier: both attackers and defenders can exploit them to alter wireless propagation for their own advantage. While prior security research has primarily explored unilateral metasurface applications - empowering either attackers or defenders - this work investigates symmetric scenarios, where both sides possess comparable metasurface capabilities. Using both theoretical modeling and real-world experiments, we analyze how competing metasurfaces interact for diverse objectives, including signal power and sensing perception. Thereby, we present the first systematic study of context-agnostic metasurface-to-metasurface interactions and their implications for wireless security. Our results reveal that the outcome of metasurface "battles" depends on an interplay of timing, placement, algorithmic strategy, and hardware scale. Across multiple case studies in Wi-Fi environments, including wireless jamming, channel obfuscation for sensing and communication, and sensing spoofing, we demonstrate that opposing metasurfaces can substantially or fully negate each other's effects. By undermining previously proposed security and privacy schemes, our findings open new opportunities for designing resilient and high-assurance physical-layer systems in smart radio environments.

In differentially private (DP) tabular data synthesis, the consensus is that statistical models are better than neural network (NN)-based methods. However, we argue that this conclusion is incomplete and overlooks the challenge of densely correlated datasets, where intricate dependencies can overwhelm statistical models. In such complex scenarios, neural networks are more suitable due to their capacity to fit complex distributions by learning directly from samples. Despite this potential, existing NN-based algorithms still suffer from significant limitations. We therefore propose MargNet, incorporating successful algorithmic designs of statistical models into neural networks. MargNet applies an adaptive marginal selection strategy and trains the neural networks to generate data that conforms to the selected marginals. On sparsely correlated datasets, our approach achieves utility close to the best statistical method while offering an average 7$\times$ speedup over it. More importantly, on densely correlated datasets, MargNet establishes a new state-of-the-art, reducing fidelity error by up to 26% compared to the previous best. We release our code on GitHub.\footnote{https://github.com/KaiChen9909/margnet}

A classifier using byte n-grams as features is the only approach we have found fast enough to meet requirements in size (sub 2 MB), speed (multiple GB/s), and latency (sub 10 ms) for deployment in numerous malware detection scenarios. However, we've consistently found that 6-8 grams achieve the best accuracy on our production deployments but have been unable to deploy regularly updated models due to the high cost of finding the top-k most frequent n-grams over terabytes of executable programs. Because the Zipfian distribution well models the distribution of n-grams, we exploit its properties to develop a new top-k n-gram extractor that is up to $35\times$ faster than the previous best alternative. Using our new Zipf-Gramming algorithm, we are able to scale up our production training set and obtain up to 30% improvement in AUC at detecting new malware. We show theoretically and empirically that our approach will select the top-k items with little error and the interplay between theory and engineering required to achieve these results.

Large Language Models (LLMs) deployed on mobile devices offer benefits like user privacy and reduced network latency, but introduce a significant security risk: the leakage of proprietary models to end users. To mitigate this risk, we propose a system design for protecting on-device LLMs using Arm Trusted Execution Environment (TEE), TrustZone. Our system addresses two primary challenges: (1) The dilemma between memory efficiency and fast inference (caching model parameters within TEE memory). (2) The lack of efficient and secure Neural Processing Unit (NPU) time-sharing between Rich Execution Environment (REE) and TEE. Our approach incorporates two key innovations. First, we employ pipelined restoration, leveraging the deterministic memory access patterns of LLM inference to prefetch parameters on demand, hiding memory allocation, I/O and decryption latency under computation time. Second, we introduce a co-driver design, creating a minimal data plane NPU driver in the TEE that collaborates with the full-fledged REE driver. This reduces the TEE TCB size and eliminates control plane reinitialization overhead during NPU world switches. We implemented our system on the emerging OpenHarmony OS and the llama.cpp inference framework, and evaluated it with various LLMs on an Arm Rockchip device. Compared to a strawman TEE baseline lacking our optimizations, our system reduces TTFT by up to 90.9% and increases decoding speed by up to 23.2%.

In this paper, we present the first detailed analysis of how training hyperparameters – such as learning rate, weight decay, momentum, and batch size – influence robustness against both transfer-based and query-based attacks. Supported by theory and experiments, our study spans a variety of practical deployment settings, including centralized training, ensemble learning, and distributed training. We uncover a striking dichotomy: for transfer-based attacks, decreasing the learning rate significantly enhances robustness by up to $64\%$. In contrast, for query-based attacks, increasing the learning rate consistently leads to improved robustness by up to $28\%$ across various settings and data distributions. Leveraging these findings, we explore – for the first time – the training hyperparameter space to jointly enhance robustness against both transfer-based and query-based attacks. Our results reveal that distributed models benefit the most from hyperparameter tuning, achieving a remarkable tradeoff by simultaneously mitigating both attack types more effectively than other training setups.

Replay and rollback attacks threaten cloud application integrity by reintroducing authentic yet stale data through an untrusted storage interface to compromise application decision-making. Prior security frameworks mitigate these attacks by enforcing forward-only state transitions (state continuity) with hardware-backed mechanisms, but they categorically treat all rollback as malicious and thus preclude legitimate rollbacks used for operational recovery from corruption or misconfiguration. We present Rebound, a general-purpose security framework that preserves rollback protection while enabling policy-authorized legitimate rollbacks of application binaries, configuration, and data. Key to Rebound is a reference monitor that mediates state transitions, enforces authorization policy, guarantees atomicity of state updates and rollbacks, and emits a tamper-evident log that provides transparency to applications and auditors. We analyze Rebound's security properties and show through an application case study – with software deployment workflows in GitLab CI – that it enables robust control over binary, configuration, and raw data versioning with low end-to-end overhead.

Split Federated Learning (SFL) is renowned for its privacy-preserving nature and low computational overhead among decentralized machine learning paradigms. In this framework, clients employ lightweight models to process private data locally and transmit intermediate outputs to a powerful server for further computation. However, SFL is a double-edged sword: while it enables edge computing and enhances privacy, it also introduces intellectual property ambiguity as both clients and the server jointly contribute to training. Existing watermarking techniques fail to protect both sides since no single participant possesses the complete model. To address this, we propose RISE, a Robust model Intellectual property protection scheme using client-Server watermark Embedding for SFL. Specifically, RISE adopts an asymmetric client-server watermarking design: the server embeds feature-based watermarks through a loss regularization term, while clients embed backdoor-based watermarks by injecting predefined trigger samples into private datasets. This co-embedding strategy enables both clients and the server to verify model ownership. Experimental results on standard datasets and multiple network architectures show that RISE achieves over $95\%$ watermark detection rate ($p-value \lt 0.03$) across most settings. It exhibits no mutual interference between client- and server-side watermarks and remains robust against common removal attacks.

With the requirements and emphases on privacy transparency placed by regulations such as GDPR and CCPA, the Google Play Store requires Android developers to more responsibly communicate their apps' privacy practices to potential users by providing the proper information via the data safety, privacy policy, and permission manifest privacy transparency channels. However, it is unclear how effective those channels are in helping users make informed decisions in the app selection and installation process. In this article, we conducted a study for 190 participants to interact with our simulated privacy transparency channels of mobile apps. We quantitatively analyzed (supplemented by qualitative analysis) participants' responses to five sets of questions. We found that data safety provides the most intuitive user interfaces, privacy policy is most informative and effective, while permission manifest excels at raising participants' concerns about an app's overall privacy risks. These channels complement each other and should all be improved.

The rapid adoption of large language models (LLMs) has brought both transformative applications and new security risks, including jailbreak attacks that bypass alignment safeguards to elicit harmful outputs. Existing automated jailbreak generation approaches e.g. AutoDAN, suffer from limited mutation diversity, shallow fitness evaluation, and fragile keyword-based detection. To address these limitations, we propose ForgeDAN, a novel evolutionary framework for generating semantically coherent and highly effective adversarial prompts against aligned LLMs. First, ForgeDAN introduces multi-strategy textual perturbations across character, word, and sentence-level operations to enhance attack diversity; then we employ interpretable semantic fitness evaluation based on a text similarity model to guide the evolutionary process toward semantically relevant and harmful outputs; finally, ForgeDAN integrates dual-dimensional jailbreak judgment, leveraging an LLM-based classifier to jointly assess model compliance and output harmfulness, thereby reducing false positives and improving detection effectiveness. Our evaluation demonstrates ForgeDAN achieves high jailbreaking success rates while maintaining naturalness and stealth, outperforming existing SOTA solutions.

Ransomware continues to evolve in complexity, making early and explainable detection a critical requirement for modern cybersecurity systems. This study presents a comparative analysis of three Transformer-based Large Language Models (LLMs) (BERT, RoBERTa, and DeBERTa) for ransomware detection using two structured datasets: UGRansome and Process Memory (PM). Since LLMs are primarily designed for natural language processing (NLP), numerical and categorical ransomware features were transformed into textual sequences using KBinsDiscretizer and token-based encoding. This enabled the models to learn behavioural patterns from system activity and network traffic through contextual embeddings. The models were fine-tuned on approximately 2,500 labelled samples and evaluated using accuracy, F1 score, and ROC-AUC. To ensure transparent decision-making in this high-stakes domain, two explainable AI techniques (LIME and SHAP) were applied to interpret feature contributions. The results show that the models learn distinct ransomware-related cues: BERT relies heavily on dominant file-operation features, RoBERTa demonstrates balanced reliance on network and financial signals, while DeBERTa exhibits strong sensitivity to financial and network-traffic indicators. Visualisation of embeddings further reveals structural differences in token representation, with RoBERTa producing more isotropic embeddings and DeBERTa capturing highly directional, disentangled patterns. In general, RoBERTa achieved the strongest F1-score, while BERT yielded the highest ROC-AUC performance. The integration of LLMs with XAI provides a transparent framework capable of identifying feature-level evidence behind ransomware predictions.

In-context learning (ICL) allows LLMs to adapt to new tasks via a few demonstrations, but those demonstrations may contain sensitive data. Differentially private (DP) ICL mechanisms mitigate this risk by injecting noise into the aggregation step, but verifying that an implementation actually meets its claimed privacy bound currently requires repeated end-to-end membership-inference attacks (MIAs) against the pipeline as a black box, incurring prohibitive LLM cost and yielding unstable empirical privacy estimates. We propose SnapAudit, an active auditing framework that decomposes a DP-ICL pipeline into a deterministic clean-inference stage and a stochastic DP-noise stage, and audits the full pipeline by combining a small snapshot of the former with bootstrap simulation of the latter. Because clean LLM outputs are near-deterministic at temperature zero, a few thousand clean LLM calls suffice to approximate the snapshot distribution; SnapAudit then bootstraps $10^5$ noisy trials from this snapshot at negligible additional cost, with finite-sample uncertainty controlled via an empirical Bernstein correction. For embedding-based mechanisms, we further introduce a multi-sweep search procedure that constructs maximally separable audit signals. SnapAudit achieves $80$–$200\times$ speedup over prior passive auditing while producing tighter and more stable empirical privacy estimates that closely match theoretical guarantees. Beyond efficiency, SnapAudit uncovers two concrete flaws in existing DP-ICL designs: (i) classical Gaussian noise calibrations underestimate leakage at large privacy budgets, allowing empirical leakage to exceed the theoretical bound; (ii) the sensitivity analysis of an embedding-aggregation mechanism is incorrect when the number of partitions equals one, leading to undersized noise and an outright privacy violation.

Split inference (SI) enables users to access deep learning (DL) services without directly transmitting raw data. However, recent studies reveal that data reconstruction attacks (DRAs) can recover the original inputs from the smashed data sent from the client to the server, leading to significant privacy leakage. While various defenses have been proposed, they often result in substantial utility degradation, particularly when the client-side model is shallow. We identify a key cause of this trade-off: existing defenses apply excessive perturbation to redundant information in the smashed data. To address this issue in computer vision tasks, we propose InfoDecom, a defense framework that first decomposes and removes redundant information and then injects noise calibrated to provide theoretically guaranteed privacy. Experiments demonstrate that InfoDecom achieves a superior utility-privacy trade-off compared to existing baselines.

Backdoor attacks pose severe threats to machine learning systems, prompting extensive research in this area. However, most existing work focuses on single-target All-to-One (A2O) attacks, overlooking the more complex All-to-X (A2X) attacks with multiple target classes, which are often assumed to have low attack success rates. In this paper, we first demonstrate that A2X attacks are robust against state-of-the-art defenses. We then propose a novel attack strategy that enhances the success rate of A2X attacks while maintaining robustness by optimizing grouping and target class assignment mechanisms. Our method improves the attack success rate by up to 28%, with average improvements of 6.7%, 16.4%, 14.1% on CIFAR10, CIFAR100, and Tiny-ImageNet, respectively. We anticipate that this study will raise awareness of A2X attacks and stimulate further research in this under-explored area. Our code is available at https://github.com/kazefjj/A2X-backdoor .

Generating thorough natural language explanations for threat detections remains an open problem in cybersecurity research, despite significant advances in automated malware detection systems. In this work, we present AutoMalDesc, an automated static analysis summarization framework that, following initial training on a small set of expert-curated examples, operates independently at scale. This approach leverages an iterative self-paced learning pipeline to progressively enhance output quality through synthetic data generation and validation cycles, eliminating the need for extensive manual data annotation. Evaluation across 3,600 diverse samples in five scripting languages demonstrates statistically significant improvements between iterations, showing consistent gains in both summary quality and classification accuracy. Our comprehensive validation approach combines quantitative metrics based on established malware labels with qualitative assessment from both human experts and LLM-based judges, confirming both technical precision and linguistic coherence of generated summaries. To facilitate reproducibility and advance research in this domain, we publish our complete dataset of more than 100K script samples, including annotated seed (0.9K) and test (3.6K) datasets, along with our methodology and evaluation framework.

Embedding-as-a-Service (EaaS) is an effective and convenient deployment solution for addressing various NLP tasks. Nevertheless, recent research has shown that EaaS is vulnerable to model extraction attacks, which could lead to significant economic losses for model providers. For copyright protection, existing methods inject watermark embeddings into text embeddings and use them to detect copyright infringement. However, current watermarking methods often resist only a subset of attacks and fail to provide comprehensive protection. To this end, we present the region-triggered semantic watermarking framework called RegionMarker, which defines trigger regions within a low-dimensional space and injects watermarks into text embeddings associated with these regions. By utilizing a secret dimensionality reduction matrix to project onto this subspace and randomly selecting trigger regions, RegionMarker makes it difficult for watermark removal attacks to evade detection. Furthermore, by embedding watermarks across the entire trigger region and using the text embedding as the watermark, RegionMarker is resilient to both paraphrasing and dimension-perturbation attacks. Extensive experiments on various datasets show that RegionMarker is effective in resisting different attack methods, thereby protecting the copyright of EaaS.