Papers

As large language models (LLMs) are increasingly deployed, ensuring their safe use is paramount. Jailbreaking, adversarial prompts that bypass model alignment to trigger harmful outputs, present significant risks, with existing studies reporting high success rates in evading common LLMs. However, previous evaluations have focused solely on the models, neglecting the full deployment pipeline, which typically incorporates additional safety mechanisms like content moderation filters. To address this gap, we present the first systematic evaluation of jailbreak attacks targeting LLM safety alignment, assessing their success across the full inference pipeline, including both input and output filtering stages. Our findings yield two key insights: first, nearly all evaluated jailbreak techniques can be detected by at least one safety filter, suggesting that prior assessments may have overestimated the practical success of these attacks; second, while safety filters are effective in detection, there remains room to better balance recall and precision to further optimize protection and user experience. We highlight critical gaps and call for further refinement of detection accuracy and usability in LLM safety systems.

To what extent are users surveilled on the web, by what technologies, and by whom? We answer these questions by combining passively observed, anonymized browsing data of a large, representative sample of Americans with domain-level data on tracking from Blacklight. We find that nearly all users ($ > 99\%$) encounter at least one ad tracker or third-party cookie over the observation window. More invasive techniques like session recording, keylogging, and canvas fingerprinting are less widespread, but over half of the users visited a site employing at least one of these within the first 48 hours of the start of tracking. Linking trackers to their parent organizations reveals that a single organization, usually Google, can track over $50\%$ of web activity of more than half the users. Demographic differences in exposure are modest and often attenuate when we account for browsing volume. However, disparities by age and race remain, suggesting that what users browse, not just how much, shapes their surveillance risk.

Mixture-of-Experts architectures have become the standard for scaling large language models due to their superior parameter efficiency. To accommodate the growing number of experts in practice, modern inference systems commonly adopt expert parallelism to distribute experts across devices. However, the absence of explicit load balancing constraints during inference allows adversarial inputs to trigger severe routing concentration. We demonstrate that out-of-distribution prompts can manipulate the routing strategy such that all tokens are consistently routed to the same set of top-$k$ experts, which creates computational bottlenecks on certain devices while forcing others to idle. This converts an efficiency mechanism into a denial-of-service attack vector, leading to violations of service-level agreements for time to first token. We propose RepetitionCurse, a low-cost black-box strategy to exploit this vulnerability. By identifying a universal flaw in MoE router behavior, RepetitionCurse constructs adversarial prompts using simple repetitive token patterns in a model-agnostic manner. On widely deployed MoE models like Mixtral-8x7B, our method increases end-to-end inference latency by 3.063x, degrading service availability significantly.

Confronting the substantial challenges of malware detection in cybersecurity necessitates solutions that are both robust and adaptable to the ever-evolving threat environment. The paper introduces Meta Learning Malware Detection (MeLeMaD), a novel framework leveraging the adaptability and generalization capabilities of Model-Agnostic Meta-Learning (MAML) for malware detection. MeLeMaD incorporates a novel feature selection technique, Chunk-wise Feature Selection based on Gradient Boosting (CFSGB), tailored for handling large-scale, high-dimensional malware datasets, significantly enhancing the detection efficiency. Two benchmark malware datasets (CIC-AndMal2020 and BODMAS) and a custom dataset (EMBOD) were used for rigorously validating the MeLeMaD, achieving a remarkable performance in terms of key evaluation measures, including accuracy, precision, recall, F1-score, MCC, and AUC. With accuracies of 98.04% on CIC-AndMal2020 and 99.97% on BODMAS, MeLeMaD outperforms the state-of-the-art approaches. The custom dataset, EMBOD, also achieves a commendable accuracy of 97.85%. The results underscore the MeLeMaD's potential to address the challenges of robustness, adaptability, and large-scale, high-dimensional datasets in malware detection, paving the way for more effective and efficient cybersecurity solutions.

Convolutional Neural Networks (CNNs) and their quantized counterparts are vulnerable to extraction attacks, posing a significant threat of IP theft. Yet, the robustness of quantized models against these attacks is little studied compared to large models. Previous defenses propose to inject calculated noise into the prediction probabilities. However, these defenses are limited since they are not incorporated during the model design and are only added as an afterthought after training. Additionally, most defense techniques are computationally expensive and often have unrealistic assumptions about the victim model that are not feasible in edge device implementations and do not apply to quantized models. In this paper, we propose DivQAT, a novel algorithm to train quantized CNNs based on Quantization Aware Training (QAT) aiming to enhance their robustness against extraction attacks. To the best of our knowledge, our technique is the first to modify the quantization process to integrate a model extraction defense into the training process. Through empirical validation on benchmark vision datasets, we demonstrate the efficacy of our technique in defending against model extraction attacks without compromising model accuracy. Furthermore, combining our quantization technique with other defense mechanisms improves their effectiveness compared to traditional QAT.

We present an openly documented methodology for fine-tuning language models to detect temporal attack patterns in multi-agent AI workflows using OpenTelemetry trace analysis. We curate a dataset of 80,851 examples from 18 public cybersecurity sources and 35,026 synthetic OpenTelemetry traces. We apply iterative QLoRA fine-tuning on resource-constrained ARM64 hardware (NVIDIA DGX Spark) through three training iterations with strategic augmentation. Our custom benchmark accuracy improves from 42.86% to 74.29%, a statistically significant 31.4-point gain. Targeted examples addressing specific knowledge gaps outperform indiscriminate scaling. Key contributions include: (1) synthetic trace generation methodology for multi-agent coordination attacks and regulatory violations, (2) empirical evidence that training data composition fundamentally determines behavior, and (3) complete open release of datasets, training scripts, and evaluation benchmarks on HuggingFace. While practical deployment requires human oversight due to false positive rates, this work establishes the first reproducible framework enabling practitioners to build custom agentic security models adapted to their threat landscapes.

Audio-language models combine audio encoders with large language models to enable multimodal reasoning, but they also introduce new security vulnerabilities. We propose a universal targeted latent space attack, an encoder-level adversarial attack that manipulates audio latent representations to induce attacker-specified outputs in downstream language generation. Unlike prior waveform-level or input-specific attacks, our approach learns a universal perturbation that generalizes across inputs and speakers and does not require access to the language model. Experiments on Qwen2-Audio-7B-Instruct demonstrate consistently high attack success rates with minimal perceptual distortion, revealing a critical and previously underexplored attack surface at the encoder level of multimodal systems.

Sophisticated attackers can evade detection-based security by using encryption, stealth tactics, and low-rate attack patterns. This challenge is particularly acute in Internet of Things (IoT) and edge environments, where limited resources make ML-based intrusion detection systems impractical. Hereby, we present Economic Denial Security (EDS), a framework that renders attacks economically infeasible rather than trying to detect them. EDS exploits a fundamental asymmetry. Defenders control their own environment, whereas attackers do not. The four mechanisms in this framework amplify attack costs superlinearly. These mechanisms include adaptive computational puzzles, decoy-driven interaction entropy, temporal stretching, and bandwidth taxation. This paper uses game theory to mathematically prove the optimal configuration of EDS, and we found that combining multiple safety mechanisms usually costs 2.1 times as much as using them separately, a key trade-off to consider during design. The good news is that EDS is extremely efficient, using less than 12 KB of memory, making it practical to run on small embedded devices like microcontrollers rather than on expensive servers. EDS is tested on 20 different IoT devices under four attack scenarios. The results showed that attacks slow down significantly, costs become asymmetric, attack success rates drop, and the system adds only 20 ms of latency with no false positive results. When tested against real malware (Mirai, Torii, and Hajime), combining EDS with machine learning detection improved protection from 67 % to 88 %. Adding both techniques together reached 94 % protection, a 27 % improvement overall. Unlike traditional detection-based approaches, EDS operates independently, without requiring attack identification, making it practical for resource-limited IoT devices where other methods simply don't work.

Recent attacks on critical infrastructure, including the 2021 Oldsmar water treatment breach and 2023 Danish energy sector compromises, highlight urgent security gaps in Industrial IoT (IIoT) deployments. While Federated Learning (FL) enables privacy-preserving collaborative intrusion detection, existing frameworks remain vulnerable to Byzantine poisoning attacks and lack robust agent authentication. We propose Zero-Trust Agentic Federated Learning (ZTA-FL), a defense in depth framework combining: (1) TPM-based cryptographic attestation achieving less than 0.0000001 false acceptance rate, (2) a novel SHAP-weighted aggregation algorithm providing explainable Byzantine detection under non-IID conditions with theoretical guarantees, and (3) privacy-preserving on-device adversarial training. Comprehensive experiments across three IDS benchmarks (Edge-IIoTset, CIC-IDS2017, UNSW-NB15) demonstrate that ZTA-FL achieves 97.8 percent detection accuracy, 93.2 percent accuracy under 30 percent Byzantine attacks (outperforming FLAME by 3.1 percent, p less than 0.01), and 89.3 percent adversarial robustness while reducing communication overhead by 34 percent. We provide theoretical analysis, failure mode characterization, and release code for reproducibility.

Recent spam email techniques exploit visual effects in text messages, such as poisoning text, obfuscating words, and hidden text salting techniques. These effects were able to evade spam detection techniques based on the text. In this paper, we overcome this limitation by introducing a novel visual-based spam detection architecture, denoted as visual-based spam filter (VBSF). The multi-step process mimics the human eye's natural way of processing visual information, automatically rendering incoming emails and capturing their content as it appears on a user screen. Then, two different processing pipelines are applied in parallel. The first pipeline pertains to the perceived textual content, as it includes optical character recognition (OCR) to extract rendered textual content, followed by naive Bayes (NB) and decision tree (DT) content classifiers. The second pipeline focuses on the appearance of the email, as it analyzes and classifies the images of rendered emails through a specific convolutional neural network. Lastly, a meta classifier integrates text- and image-based classifier outputs, exploiting the stacking ensemble learning method. The performance of the proposed VBSF is assessed, showing that it achieves an accuracy of more than 98%, which is higher than the compared existing techniques on the designed dataset.

Side-channel attacks try to extract secret information from a system by analyzing different side-channel signatures, such as power consumption, electromagnetic emanation, thermal dissipation, acoustics, time, etc. Power-based side-channel attack is one of the most prominent side-channel attacks in cybersecurity, which rely on data-dependent power variations in a system to extract sensitive information. While there are related surveys, they primarily focus on power side-channel attacks on cryptographic implementations. In recent years, power-side channel attacks have been explored in diverse application domains, including key extraction from cryptographic implementations, reverse engineering of machine learning models, user behavior data exploitation, and instruction-level disassembly. In this paper, we provide a comprehensive survey of power side-channel attacks and their countermeasures in different application domains. Specifically, this survey aims to classify recent power side-channel attacks and provide a comprehensive comparison based on application-specific considerations.

Large Language Models (LLMs) can be driven into over-generation, emitting thousands of tokens before producing an end-of-sequence (EOS) token. This degrades answer quality, inflates latency and cost, and can be weaponized as a denial-of-service (DoS) attack. Recent work has begun to study DoS-style prompt attacks, but typically focuses on a single attack algorithm or assumes white-box access, without an attack-side benchmark that compares prompt-based attackers in a black-box, query-only regime with a known tokenizer. We introduce such a benchmark and study two prompt-only attackers. The first is an Evolutionary Over-Generation Prompt Search (EOGen) that searches the token space for prefixes that suppress EOS and induce long continuations. The second is a goal-conditioned reinforcement learning attacker (RL-GOAL) that trains a network to generate prefixes conditioned on a target length. To characterize behavior, we introduce Over-Generation Factor (OGF): the ratio of produced tokens to a model's context window, along with stall and latency summaries. EOGen discovers short-prefix attacks that raise Phi-3 to OGF = 1.39 +/- 1.14 (Success@>=2: 25.2%); RL-GOAL nearly doubles severity to OGF = 2.70 +/- 1.43 (Success@>=2: 64.3%) and drives budget-hit non-termination in 46% of trials.

In recent years, drone delivery, which utilizes unmanned aerial vehicles (UAVs) for package delivery and pickup, has gradually emerged as a crucial method in logistics. Since delivery drones are expensive and may carry valuable packages, they must maintain a safe distance from individuals until user-drone mutual authentication is confirmed. Despite numerous authentication schemes being developed, existing solutions are limited in authentication distance and lack resilience against sophisticated attacks. To this end, we introduce SyncGait, an implicit gait-based mutual authentication system for drone delivery. SyncGait leverages the user's unique arm swing as he walks toward the drone to achieve mutual authentication without requiring additional hardware or specific authentication actions. We conducted extensive experiments on 14 datasets collected from 31 subjects. The results demonstrate that SyncGait achieves an average accuracy of 99.84% at a long distance ($>18m$) and exhibits strong resilience against various spoofing attacks, making it a robust, secure, and user-friendly solution in real-world scenarios.

API gateways serve as critical enforcement points for security, governance, and traffic management in cloud-native systems. As organizations increasingly adopt multi-cluster and hybrid cloud deployments, maintaining consistent policy enforcement, predictable performance, and operational stability across heterogeneous gateway environments becomes challenging. Existing approaches typically manage security, governance, and performance as loosely coupled concerns, leading to configuration drift, delayed policy propagation, and unstable runtime behavior under dynamic workloads. This paper presents a governance-aware, intent-driven architecture for coordinated API gateway management in multi-cluster cloud environments. The proposed approach expresses security, governance, and performance objectives as high-level declarative intents, which are systematically translated into enforceable gateway configurations and continuously validated through policy verification and telemetry-driven feedback. By decoupling intent specification from enforcement while enabling bounded, policy-compliant adaptation, the architecture supports heterogeneous gateway implementations without compromising governance guarantees or service-level objectives. A prototype implementation across multiple Kubernetes clusters demonstrates the effectiveness of the proposed design. Experimental results show up to a 42% reduction in policy drift, a 31% improvement in configuration propagation time, and sustained p95 latency overhead below 6% under variable workloads, compared to manual and declarative baseline approaches. These results indicate that governance-aware, intent-driven gateway orchestration provides a scalable and reliable foundation for secure, consistent, and performance-predictable cloud-native platforms.

Web applications increasingly face evasive and polymorphic attack payloads, yet traditional web application firewalls (WAFs) based on static rule sets such as the OWASP Core Rule Set (CRS) often miss obfuscated or zero-day patterns without extensive manual tuning. This work introduces WAMM, an AI-driven multiclass web attack detection framework designed to reveal the limitations of rule-based systems by reclassifying HTTP requests into OWASP-aligned categories for a specific technology stack. WAMM applies a multi-phase enhancement pipeline to the SR-BH 2020 dataset that includes large-scale deduplication, LLM-guided relabeling, realistic attack data augmentation, and LLM-based filtering, producing three refined datasets. Four machine and deep learning models are evaluated using a unified feature space built from statistical and text-based representations. Results show that using an augmented and LLM-filtered dataset on the same technology stack, XGBoost reaches 99.59% accuracy with microsecond-level inference while deep learning models degrade under noisy augmentation. When tested against OWASP CRS using an unseen augmented dataset, WAMM achieves true positive block rates between 96 and 100% with improvements of up to 86%. These findings expose gaps in widely deployed rule-based defenses and demonstrate that curated training pipelines combined with efficient machine learning models enable a more resilient, real-time approach to web attack detection suitable for production WAF environments.

This document presents a concise overview of the contemporary research directions in quantum computer cybersecurity. The aim of this document is not to be a survey, but rather a succinct summary of the major research directions in quantum computer cybersecurity at the end of the first half of the current decade. The document has been inspired by the presentations and discussions held at the 3$^{rd}$ Quantum Computer Cybersecurity Symposium, but goes beyond the contents of the symposium and aims to summarize at the high level the last five years of quantum computer cybersecurity work in academia. It is hoped that the document can provide researchers as well as government and industry leaders an overview of the current landscape of security threats and defenses against emergent quantum computing technologies. The document also includes a discussion of the current trends in cybersecurity research on quantum computers, and the perceived research gaps that should be filled with future funding and through academic and industry~research.

Powerful autonomous systems, which reason, plan, and converse using and between numerous tools and agents, are made possible by Large Language Models (LLMs), Vision-Language Models (VLMs), and new agentic AI systems, like LangChain and GraphChain. Nevertheless, this agentic environment increases the probability of the occurrence of multimodal prompt injection (PI) attacks, in which concealed or malicious instructions carried in text, pictures, metadata, or agent-to-agent messages may spread throughout the graph and lead to unintended behavior, a breach of policy, or corruption of state. In order to mitigate these risks, this paper suggests a Cross-Agent Multimodal Provenanc- Aware Defense Framework whereby all the prompts, either user-generated or produced by upstream agents, are sanitized and all the outputs generated by an LLM are verified independently before being sent to downstream nodes. This framework contains a Text sanitizer agent, visual sanitizer agent, and output validator agent all coordinated by a provenance ledger, which keeps metadata of modality, source, and trust level throughout the entire agent network. This architecture makes sure that agent-to-agent communication abides by clear trust frames such such that injected instructions are not propagated down LangChain or GraphChain-style-workflows. The experimental assessments show that multimodal injection detection accuracy is significantly enhanced, and the cross-agent trust leakage is minimized, as well as, agentic execution pathways become stable. The framework, which expands the concept of provenance tracking and validation to the multi-agent orchestration, enhances the establishment of secure, understandable and reliable agentic AI systems.

This paper presents a private transfer architecture for the Internet Computer (ICP) that decouples deposit and retrieval through two short-lived intermediaries, with sealed storage and attested teardown by an ephemeral witness. The protocol uses a non-interactive RDMPF-based encapsulation to derive per-transfer transport keys. A public notice hint is computed from the capsule to enable discovery without fingerprinting the recipient's key. Retrieval is authorized by a short proof of decapsulation that reveals no identities. All transaction intermediaries are ephemeral and issue certified destruction intents and proofs, allowing a noticeboard to publish auditable finalization records. The design provides sender identity privacy with respect to the recipient, content confidentiality against intermediaries, forward secrecy for transport keys after staged destruction, verifiable liveness and finality. We formalize the basic interfaces, provide the security arguments for encapsulation correctness, hint privacy, authorization soundness and timeout reclaim. In terms of implementation, it has been recently brought into production on the ICP under the name ICPP. It has been subject to exhaustive testing and incorporates a few enhancements, focusing on the operational possibilities offered by ICP's technology. This work hence serves as a broad reference for the protocol now publicly accessible.

The software supply chain attacks are becoming more and more focused on trusted development and delivery procedures, so the conventional post-build integrity mechanisms cannot be used anymore. The available frameworks like SLSA, SBOM and in toto are majorly used to offer provenance and traceability but do not have the capabilities of actively identifying and removing vulnerabilities in software production. The current paper includes an example of agentic artificial intelligence (AI) based on autonomous software supply chain security that combines large language model (LLM)-based reasoning, reinforcement learning (RL), and multi-agent coordination. The suggested system utilizes specialized security agents coordinated with the help of LangChain and LangGraph, communicates with actual CI/CD environments with the Model Context Protocol (MCP), and documents all the observations and actions in a blockchain security ledger to ensure integrity and auditing. Reinforcement learning can be used to achieve adaptive mitigation strategies that consider the balance between security effectiveness and the operational overhead, and LLMs can be used to achieve semantic vulnerability analysis, as well as explainable decisions. This framework is tested based on simulated pipelines, as well as, actual world CI/CD integrations on GitHub Actions and Jenkins, including injection attacks, insecure deserialization, access control violations, and configuration errors. Experimental outcomes indicate better detection accuracy, shorter mitigation latency and reasonable build-time overhead than rule-based, provenance only and RL only baselines. These results show that agentic AI can facilitate the transition to self defending, proactive software supply chains rather than reactive verification ones.

Despite holding over $1.7T in value, Bitcoin scales poorly: Layer-1 (L1) processes only a few transactions per second, and Layer-2 (L2) solutions suffer from operational and liquidity fragmentation that requires external bridges for cross-chain connectivity. We present Bitcoin-IPC, a protocol that horizontally scales Bitcoin via permissionless, interconnected, programmable Proof-of-Stake L2 subnets staked in L1 BTC. Subnets leverage Bitcoin L1 for interconnectivity, settlement, and security, including Sybil protection, equivocating validator slashing, unilateral exit for honest stake, long-range attack protection, and compromised subnet firewalling. Bitcoin-IPC cross-subnet transfers involve only source/destination subnet validators and L1, with no external bridges or per-path liquidity lock-up. Embedded and batched in Bitcoin's witness mechanism, cross-subnet settlements can approach 6 vB per transfer, 23x less than a native L1 payment. Filling Bitcoin blocks with such batches corresponds to L1 monetary throughput of about 273 transfers per second, without modifying Bitcoin.

Modern CPUs are black boxes, proprietary, and increasingly characterized by sophisticated microarchitectural flaws that evade traditional analysis. While some of these critical vulnerabilities have been uncovered through cumbersome manual effort, building an automated and systematic vulnerability detection framework for real-world post-silicon processors remains a challenge. In this paper, we present Fuzzilicon, the first post-silicon fuzzing framework for real-world x86 CPUs that brings deep introspection into the microcode and microarchitectural layers. Fuzzilicon automates the discovery of vulnerabilities that were previously only detectable through extensive manual reverse engineering, and bridges the visibility gap by introducing microcode-level instrumentation. At the core of Fuzzilicon is a novel technique for extracting feedback directly from the processor's microarchitecture, enabled by reverse-engineering Intel's proprietary microcode update interface. We develop a minimally intrusive instrumentation method and integrate it with a hypervisor-based fuzzing harness to enable precise, feedback-guided input generation, without access to Register Transfer Level (RTL). Applied to Intel's Goldmont microarchitecture, Fuzzilicon introduces 5 significant findings, including two previously unknown microcode-level speculative-execution vulnerabilities. Besides, the Fuzzilicon framework automatically rediscover the $\mu$Spectre class of vulnerabilities, which were detected manually in the previous work. Fuzzilicon reduces coverage collection overhead by up to 31$\times$ compared to baseline techniques and achieves 16.27% unique microcode coverage of hookable locations, the first empirical baseline of its kind. As a practical, coverage-guided, and scalable approach to post-silicon fuzzing, Fuzzilicon establishes a new foundation to automate the discovery of complex CPU vulnerabilities.

The rapid growth of Artificial Intelligence (AI) models and applications has led to an increasingly complex security landscape. Developers of AI projects must contend not only with traditional software supply chain issues but also with novel, AI-specific security threats. However, little is known about what security issues are commonly encountered and how they are resolved in practice. This gap hinders the development of effective security measures for each component of the AI supply chain. We bridge this gap by conducting an empirical investigation of developer-reported issues and solutions, based on discussions from Hugging Face and GitHub. To identify security-related discussions, we develop a pipeline that combines keyword matching with an optimal fine-tuned distilBERT classifier, which achieved the best performance in our extensive comparison of various deep learning and large language models. This pipeline produces a dataset of 312,868 security discussions, providing insights into the security reporting practices of AI applications and projects. We conduct a thematic analysis of 753 posts sampled from our dataset and uncover a fine-grained taxonomy of 32 security issues and 24 solutions across four themes: (1) System and Software, (2) External Tools and Ecosystem, (3) Model, and (4) Data. We reveal that many security issues arise from the complex dependencies and black-box nature of AI components. Notably, challenges related to Models and Data often lack concrete solutions. Our insights can offer evidence-based guidance for developers and researchers to address real-world security threats across the AI supply chain.

Neural ranking models have achieved remarkable progress and are now widely deployed in real-world applications such as Retrieval-Augmented Generation (RAG). However, like other neural architectures, they remain vulnerable to adversarial manipulations: subtle character-, word-, or phrase-level perturbations can poison retrieval results and artificially promote targeted candidates, undermining the integrity of search engines and downstream systems. Existing defenses either rely on heuristics with poor generalization or on certified methods that assume overly strong adversarial knowledge, limiting their practical use. To address these challenges, we propose RobustMask, a novel defense that combines the context-prediction capability of pretrained language models with a randomized masking-based smoothing mechanism. Our approach strengthens neural ranking models against adversarial perturbations at the character, word, and phrase levels. Leveraging both the pairwise comparison ability of ranking models and probabilistic statistical analysis, we provide a theoretical proof of RobustMask's certified top-K robustness. Extensive experiments further demonstrate that RobustMask successfully certifies over 20% of candidate documents within the top-10 ranking positions against adversarial perturbations affecting up to 30% of their content. These results highlight the effectiveness of RobustMask in enhancing the adversarial robustness of neural ranking models, marking a significant step toward providing stronger security guarantees for real-world retrieval systems.

In todays scenario, various organizations store their sensitive data in the cloud environment. Multiple problems are present while retrieving and storing vast amounts of data, such as the frequency of data requests (increasing the computational overhead of the server) and data leakage while storing. To cope with said problem, Attribute-Based Encryption (ABE) is one of the potential security and access control techniques for secure data storage and authorization. The proposed work divides into two objectives: (i) provide access to authorized users and (ii) secure data storage in a cloud environment. The improved ABE using Functional Based Stream Cipher (FBSE) is proposed for data storage. The proposed technique uses simple scalar points over a parabolic curve to provide multiparty authorization. The authorization points are generated and share only with the authorized recipients. The Shamir secret sharing technique generate the authorization points and 2D-Lagrange Interpolation is used to reconstruct the secret points from regular parabola. The proposed scheme has specified the threshold (Ts>3) legally authorized users to reconstruct the attribute-associated keys for decryption. The encryption of data is evaluated using Statistical analysis (NIST Statistical Test Suite, Correlation Coefficient, and Histogram) test to investigate image pixel deviation. The parameters like encryption and decryption are used for performance analysis, where an increase in the number of attributes for the authorization policy will increase the encryption time. The proposed scheme imposes minimal storage overhead, irrespective of the users identity. The security analysis evidence that it resists collision attacks. The security and performance analysis results demonstrate that the proposed scheme is more robust and secure.

Large language models (LLMs), such as ChatGPT, have achieved remarkable success across a wide range of fields. However, their trustworthiness remains a significant concern, as they are still susceptible to jailbreak attacks aimed at eliciting inappropriate or harmful responses. However, existing jailbreak attacks mainly operate at the natural language level and rely on a single attack strategy, limiting their effectiveness in comprehensively assessing LLM robustness. In this paper, we propose Equacode, a novel multi-strategy jailbreak approach for large language models via equation-solving and code completion. This approach transforms malicious intent into a mathematical problem and then requires the LLM to solve it using code, leveraging the complexity of cross-domain tasks to divert the model's focus toward task completion rather than safety constraints. Experimental results show that Equacode achieves an average success rate of 91.19% on the GPT series and 98.65% across 3 state-of-the-art LLMs, all with only a single query. Further, ablation experiments demonstrate that EquaCode outperforms either the mathematical equation module or the code module alone. This suggests a strong synergistic effect, thereby demonstrating that multi-strategy approach yields results greater than the sum of its parts.

Federated unlearning has become an attractive approach to address privacy concerns in collaborative machine learning, for situations when sensitive data is remembered by AI models during the machine learning process. It enables the removal of specific data influences from trained models, aligning with the growing emphasis on the "right to be forgotten." While extensively studied in horizontal federated learning, unlearning in vertical federated learning (VFL) remains challenging due to the distributed feature architecture. VFL unlearning includes sample unlearning that removes specific data points' influence and label unlearning that removes entire classes. Since different parties hold complementary features of the same samples, unlearning tasks require cross-party coordination, creating computational overhead and complexities from feature interdependencies. To address such challenges, we propose FedORA (Federated Optimization for data Removal via primal-dual Algorithm), designed for sample and label unlearning in VFL. FedORA formulates the removal of certain samples or labels as a constrained optimization problem solved using a primal-dual framework. Our approach introduces a new unlearning loss function that promotes classification uncertainty rather than misclassification. An adaptive step size enhances stability, while an asymmetric batch design, considering the prior influence of the remaining data on the model, handles unlearning and retained data differently to efficiently reduce computational costs. We provide theoretical analysis proving that the model difference between FedORA and Train-from-scratch is bounded, establishing guarantees for unlearning effectiveness. Experiments on tabular and image datasets demonstrate that FedORA achieves unlearning effectiveness and utility preservation comparable to Retrain with reduced computation and communication overhead.

Machine learning (ML) underpins foundation models in finance, healthcare, and critical infrastructure, making them targets for data poisoning, model extraction, prompt injection, automated jailbreaking, and preference-guided black-box attacks that exploit model comparisons. Larger models can be more vulnerable to introspection-driven jailbreaks and cross-modal manipulation. Traditional cybersecurity lacks ML-specific threat modeling for foundation, multimodal, and RAG systems. Objective: Characterize ML security risks by identifying dominant TTPs, vulnerabilities, and targeted lifecycle stages. Methods: We extract 93 threats from MITRE ATLAS (26), AI Incident Database (12), and literature (55), and analyze 854 GitHub/Python repositories. A multi-agent RAG system (ChatGPT-4o, temp 0.4) mines 300+ articles to build an ontology-driven threat graph linking TTPs, vulnerabilities, and stages. Results: We identify unreported threats including commercial LLM API model stealing, parameter memorization leakage, and preference-guided text-only jailbreaks. Dominant TTPs include MASTERKEY-style jailbreaking, federated poisoning, diffusion backdoors, and preference optimization leakage, mainly impacting pre-training and inference. Graph analysis reveals dense vulnerability clusters in libraries with poor patch propagation. Conclusion: Adaptive, ML-specific security frameworks, combining dependency hygiene, threat intelligence, and monitoring, are essential to mitigate supply-chain and inference risks across the ML lifecycle.

Zero Trust architectures strengthen continuous verification and least-privilege enforcement, but general-purpose implementations do not explicitly treat financial exposure and transaction-level operational cost as first-class decision inputs. This paper presents SecureBank, a financially-aware Zero Trust architecture that combines identity and device trust, contextual consistency, financial risk, business-aligned segmentation, and policy actions such as allow, step-up verification, and block. The empirical study validates the architecture's financial-risk decision component rather than claiming end-to-end production validation of the complete architecture. A public credit-card fraud dataset containing 284,807 transactions was divided into independent training, validation, and test partitions using a 60/20/20 stratified design. A class-weighted logistic regression model produced a risk score. The operational threshold was selected only on the validation partition using an asymmetric cost function in which a false negative was assigned a cost 50 times greater than a false positive, and the frozen threshold was then evaluated once on the independent test partition. The final test set contained 56,962 transactions and 99 fraud cases. At the validation-selected risk-score threshold of 0.9991, the model achieved ROC-AUC 0.9664, average precision 0.7119, precision 71.03%, recall 76.77%, F1-score 0.7379, and a false-positive rate of 0.0545%. These results provide reproducible component-level evidence that financially asymmetric operating costs can be integrated into a Zero Trust decision process while keeping the scope of the empirical claims explicit.

This paper focuses on solving unconstrained privacy-preserving black-box optimization (PBBO), its corresponding least Frobenius norm updating of quadratic models, and the differentially privacy mechanisms for PBBO. Optimization problems with transformed/encrypted objective functions aim to minimize F(x), which is encrypted/transformed/encrypted to F_k(x) as the output at the k-th iteration. A new derivative-free solver named DFOp, with its implementation, is proposed in this paper, which has a new updating formula for the quadratic model functions. The convergence of DFOp for solving problems with transformed/encrypted objective functions is given. Other analyses, including the new model updating formula and the analysis of the transformation's impact to model functions are presented. We propose two differentially private noise-adding mechanisms for privacy-preserving black-box optimization. Numerical results show that DFOp performs better than compared algorithms. To the best of our knowledge, DFOp is the first derivative-free solver that can solve black-box optimization problems with step-encryption and privacy-preserving black-box problems exactly, which also tries to answer the open question about the combination of derivative-free optimization and privacy.

Reinforcement learning is increasingly used to transform large language models into agentic systems that act over long horizons, invoke tools, and manage memory under partial observability. While recent work has demonstrated performance gains through tool learning, verifiable rewards, and continual training, deployed self-improving agents raise unresolved security and governance challenges: optimization pressure can incentivize reward hacking, behavioral drift is difficult to audit or reproduce, and improvements are often entangled in opaque parameter updates rather than reusable, verifiable artifacts. This paper proposes Audited Skill-Graph Self-Improvement (ASG-SI), a framework that treats self-improvement as iterative compilation of an agent into a growing, auditable skill graph. Each candidate improvement is extracted from successful trajectories, normalized into a skill with an explicit interface, and promoted only after passing verifier-backed replay and contract checks. Rewards are decomposed into reconstructible components derived from replayable evidence, enabling independent audit of promotion decisions and learning signals. ASG-SI further integrates experience synthesis for scalable stress testing and continual memory control to preserve long-horizon performance under bounded context. We present a complete system architecture, threat model, and security analysis, and provide a fully runnable reference implementation that demonstrates verifier-backed reward construction, skill compilation, audit logging, and measurable improvement under continual task streams. ASG-SI reframes agentic self-improvement as accumulation of verifiable, reusable capabilities, offering a practical path toward reproducible evaluation and operational governance of self-improving AI agents.