Searcharxiv⌕ Search

arXiv subjects

Yuval Elovici

Publications and source records attributed to Yuval Elovici.

At least 19 recordsLinked to original sources

I2C-GRPO: Insight Importance Checklist for Reinforcement Learning in a Multi-Agent DDoS Alert Triage Workflow

Large language model (LLM) agents are increasingly used for investigative tasks over incomplete evidence, such as cybersecurity alert triage, but reinforcement learning (RL) for such tasks is difficult: intermediate steps have no verifiable labels, rollouts are costly, and outcome rewards are sparse. We propose Insight Importance Checklist (I2C)-GRPO, a checklist reward for intermediate investigation steps. For each decision point, a panel of expert LLMs proposes investigative actions, their atomic forensic insights are extracted, and a separate three-model ranker panel orders the insights by importance. A reward model checks which ranked insights a candidate action targets and returns their consensus-weighted coverage. Because actions are scored without being executed, the reward is a proxy for expected investigative value, and training reduces to a one-step contextual bandit over fixed investigation prefixes. From a soft-computing perspective, the reward is an additive fuzzy measure whose rank-exponent weights act as an ordered weighted averaging (OWA)-type operator; one exponent sets its orness, i.e., how strongly the reward concentrates on the top of the judges' consensus. We train an open-weight Qwen3-4B Triage Agent within a fixed multi-agent Distributed Denial of Service (DDoS) alert-triage workflow on DARPA 2000 LLDOS 1.0 and test it on an independent CSE-CIC-IDS2018 DDoS trace. Across three training seeds, the agent reaches a mean F1-score of 0.869, 5.9 and 9.7 points above uniform insight weights and DIVA, and a false-positive rate of 4.9% versus 16.4% for GPT-5-mini. Within the weighting family, false positives fall only at the highest-orness weighting evaluated. All claims are limited to DDoS alert triage; the code, task contexts, and ranked insight dataset will be released.

cs.CR↗

Scan Before You Install: Detecting and Mitigating Local and Composite Vulnerabilities in Skills

An agent skill packages natural-language instructions, scripts, and reference files that an LLM-based agent loads for a task. Skills may access the file system, shell, and network under user-authorized permissions; weaknesses can lead to unsafe execution. Existing scanners primarily target malicious skills, not unintended vulnerabilities in public repositories, and analyze skills in isolation, overlooking multi-skill interactions. To address these gaps, we developed a six-module scanner: Content Extraction, Local Analysis, Composite Analysis, Validation, Reporting, and Mitigation. Local Analysis detects unintended vulnerabilities in benign skills. On a 556-case leaderboard subset of SkillTrustBench, it achieved 95.90% recall with a 4.22% false positive rate; on a balanced sample of 1,000 MalSkillBench skills, its F1 was 0.947, outperforming Cisco's Skill Scanner, NVIDIA's SkillSpector, AgentSkillsScanner, and SkillSieve. In 1,587 skills from the 75 most-starred GitHub repositories, three analysts confirmed 53 of 59 findings (89.8% precision) across 23 repositories. We developed proof-of-concept (PoC) attacks for each vulnerability, disclosed 10 medium- and high-severity cases to maintainers, and tested mitigations on the remaining cases. Composite Analysis models dependencies among skills through shared resources, invocations, and instruction-mediated interactions. On SCR-Bench, its F1 score was 0.991 (98.7% precision, 99.6% recall), versus 0.816-0.858 for three SkillsGuard configurations released with CompoSkill. It identified two composite findings on GitHub, validated by PoC attacks. The Mitigation module recommends skill substitution, patches, and configuration changes. For 47 of 55 GitHub findings, mitigations prevented PoC attacks in all trials while preserving tested legitimate functionality.

cs.CR↗

JANUS: Denial-of-Service Attack Against Beam Hopping in LEO Satellite Networks

Low Earth orbit (LEO) satellite networks are increasingly used to provide global connectivity. However, each satellite has limited resources that need to be allocated according to demand, which varies geographically and over time. Beam hopping addresses this challenge by dividing a satellite's service area into geographic cells. Rather than illuminating every cell simultaneously, it dynamically assigns available beams to a selected subset based on demand. This reliance on observed traffic demand as an input to beam-selection decisions creates a new attack surface whose security implications have received little attention. In this paper, we present JANUS, a novel targeted denial-of-service attack against beam-hopping systems in LEO networks. We show that a small botnet of compromised terminals can inject legitimate user traffic into carefully selected non-victim cells to manipulate the beam-hopping scheduler's view of demand. This manipulation alters beam-allocation decisions and redirects service away from the targeted victim area. We evaluate JANUS across different system configurations, schedulers, attack horizons, and attacker-knowledge settings to characterize the attack's effectiveness, required resources, and resulting service disruption over time. Against a rank-based KMAX scheduler, JANUS achieves complete service denial for up to approximately 95% of evaluated victims. Against DRL, JANUS can exclude the victim from approximately 92% of scheduling decisions. Finally, we evaluate mitigation strategies that reduce the attack effectiveness.

cs.CR↗

An Empirical Security Analysis of Open-Source Software Used in Onboard Satellite Systems

The use of open-source software (OSS) in satellite flight systems is increasing as missions adopt reusable frameworks, shared libraries, and community-maintained components. While this accelerates development, it also introduces software-security risks into systems where patching is costly and failures may affect mission operations. This paper presents an empirical security study of OSS used in onboard satellite systems. We analyze 126 public repositories using a pipeline that combines software bill of materials generation, software composition analysis, static application security testing, infrastructure-as-code analysis, and secret scanning. After rule-based cleaning, onboard-scope filtering, and fingerprint-based deduplication, the pipeline produced a final dataset of 2,827 findings. The results show that security findings are widespread but unevenly distributed. Medium-severity findings account for 49% of the dataset, and 72% are classified as medium severity or higher. A Common Weakness Enumeration (CWE)-based taxonomy assigns all findings to eight weakness families. Memory Safety and Code Quality dominate the dataset, followed by Input Validation and Injection. Most findings occur in project-developed code, accounting for 81.4% of the dataset, while external dependency code remains a relevant source of findings. While these findings do not establish mission-specific exploitability, they provide an empirical characterization of recurring security patterns across the open-source onboard satellite software ecosystem, helping quantify their prevalence and prioritize areas that warrant the greatest security attention.

cs.CR↗

HYDRA: Quantifying Botnet Resource Thresholds for Efficient Link-Flooding Attacks on LEO Satellite Networks

Low Earth orbit (LEO) satellite constellations, such as Starlink and Kuiper, are rapidly emerging as a critical backbone for low-latency global connectivity. As these systems expand, they become more attractive attack targets, necessitating increased resilience and security. Threat actors seek to exploit constellation-specific properties such as predictable motion, time-varying topologies, and reliance on inter-satellite and ground-satellite links. Recent work has shown that link-flooding attacks (LFAs) can exploit these properties to congest strategic network bottlenecks. Yet, prior work does not quantify the resilience of LEO networks to targeted disruption. We present HYDRA, a modeling and optimization framework that formulates LFA variants as botnet minimization problems. HYDRA quantifies network resilience to LFAs by measuring the smallest active subset of bots and the corresponding traffic allocation required to disrupt communication between targeted geographic areas. Under matched stealth constraints, HYDRA achieves the same targeted disruption as ICARUS while using 34% fewer bots and 23% less aggregate attack traffic. HYDRA achieves over 97% success in sustaining continuous attacks as the network topology evolves. Finally, HYDRA evaluates five mitigation strategies, showing how routing diversification, ingress policing, distance-based traffic constraints, source throttling, and botnet attrition reduce attack success and improve network resilience to targeted disruption.

cs.CR↗

Training-Free Policy Violation Detection via Activation-Space Whitening in LLMs

As organizations increasingly deploy LLMs in sensitive domains such as legal, financial, and medical settings, ensuring alignment with internal organizational policies has become a priority. Existing content moderation frameworks remain largely confined to the safety domain and lack the robustness to capture nuanced organizational policies. LLM-as-a-judge and fine-tuning approaches, though flexible, introduce significant latency and training cost. To address these limitations, we frame policy violation detection as an out-of-distribution (OOD) problem in the model's activation space. We propose a training-free method that operates directly on the LLM internal representations, leveraging prior evidence that decision-relevant information is encoded within them. Inspired by whitening techniques, we derive policy-violation scores directly from normalized representations of LLM hidden activations. Our method requires only the policy text and a small number of illustrative samples, making it lightweight and easily deployable. We extensively evaluate our method across multiple LLMs and challenging policy benchmarks. It achieves up to 86.0% F1, outperforming both fine-tuned and LLM-as-a-judge baselines while requiring substantially less computation. Our code is publicly available at: https://github.com/FujitsuResearch/LLM-policy-violation-detection

cs.LG↗

Rethinking Satellite Cybersecurity: A System-Level Taxonomy and Longitudinal Analysis

Satellite systems are increasingly targeted by cyber and electronic-warfare adversaries as their roles in communication, navigation, Earth observation, and defense expand. Existing surveys do not comprehensively characterize adversarial behavior across the full attack lifecycle and often omit emerging attack surfaces such as adversarial machine learning (AML). This paper presents a satellite-specific taxonomy of tactics, techniques, and procedures (TTPs), developed primarily for low Earth orbit systems and informed by evidence from LEO, MEO, and GEO missions. We analyze the space, ground, communication, and user segments to identify architectural exposures and operational attack surfaces, and compile a dataset of more than 200 publicly reported satellite incidents from 1962 to 2026, including over 80 incidents not covered in prior work. Longitudinal analysis reveals shifts toward ground-segment compromise, GNSS interference, communication disruption, proximity-based counterspace activity, and deception-oriented attacks. Building on these findings, we propose a MITRE ATT&CK-inspired satellite attack lifecycle taxonomy that integrates subsystem exploitation, radio-frequency interference, on-orbit operations, AML, and deception techniques. We demonstrate its practical utility through case studies of the 2022 Viasat KA-SAT cyberattack and a simulation-based ICARUS constellation-scale denial-of-service scenario. The framework combines longitudinal evidence, real-world incidents, and emerging attack modalities to support threat modeling, defensive planning, and the design of detection and mitigation strategies.

cs.CR↗

Data Leakage Prevention in Agentic Applications via Preemptive Hardening

Agentic systems integrate LLM driven planning with interfaces to external tools, making data leakage and tool misuse feasible via instruction/data boundary failures and prompt injection attacks. Enforcing required controls consistently is particularly challenging in workflows spanning many codebases and heterogeneous agents. To address this challenge in multi agentic systems, we present a pre-deployment pipeline for scanning, hardening, and validation of agentic applications. The pipeline analyzes prompt templates, tool interfaces, and tool-invocation code to identify leakage-enabling patterns and generate actionable patches. The hardened application is then validated through adversarial prompt injection attacks and benign input variations ensuring that mitigations do not disrupt intended behavior. In the hardening stage, high-risk tools are prioritized, and minimally invasive mitigations are applied, including schema tightening, boundary sanitization, allowlist-based tool gating, and least-privilege checks. In the validation stage, the pipeline automatically generates attack inputs that mimic jailbreaks, instruction overrides, and tool-targeted manipulation, along with benign task variants, to confirm that the functionality of the hardened application is preserved after remediation. We evaluated the pipeline on five real-world agentic applications, as well as on the AgentDojo benchmark. Across all applications, the proposed pipeline identified recurring leakage-enabling patterns and generated patches that can be integrated without disrupting the intended application behavior. The resulting modifications of application code were shown to eliminate leaks when targeted by basic jailbreak and instruction-override attacks, achieving a 100% reduction in leakage, and reduce leaks by 91% under conditions of stress-induced manipulation, without the need of continuous runtime policy enforcement.

cs.CR↗

COHORT: Collaborative Orchestration for Hardening via Offensive Replay on Emulated Topologies

Mitigating an observed adversary in an enterprise network typically takes weeks of expert work: an analyst derives a mitigation tailored to that adversary, validates it without breaking production, and verifies it disrupts the specific attack. The procedure relies on expert judgment and cannot safely be exercised against the production network. COHORT is the first end-to-end framework to automate this procedure for deployable mitigations. A role-decomposed multi-agent LLM workflow proposes candidates, implements them as real device commands, and refines them through a critique loop, all on a high-fidelity GNS3 emulator running real vendor firmware (firewall, switch, router). Each candidate is evaluated by offensive replay: re-executing the original adversary on the mitigated network for a paired comparison against the unmitigated baseline, rather than the reward-signal or expert-judgment proxies used in prior simulation, hybrid, and configuration-generation work. Two further checks complement replay: a connectivity-regression check (LAN ping and internet HTTP probe) rejects mitigations that disrupt legitimate LAN or internet connectivity, and a cumulative evaluation stacks approved mitigations onto a persistent state to surface compound effects. Across three topologies and four attack scenarios (ransomware, lateral movement, DNS exfiltration, data theft), 46.7% of generated mitigations both disrupt the attack and preserve connectivity under replay, 4.4 times the rate of a single-agent baseline using the same model and tool access. A demo video walking through the framework is available with our released artifacts.

cs.NI↗

SoK: Security Below the OS -- A Security Analysis of UEFI

The Unified Extensible Firmware Interface (UEFI) is a linchpin of modern computing systems, governing secure system initialization and booting. This paper is urgently needed because of the surge in UEFI-related attacks and vulnerabilities in recent years. Motivated by this urgent concern, we undertake an extensive exploration of the UEFI landscape, dissecting its distribution supply chain, booting process, and security features. We carefully study a spectrum of UEFI-targeted attacks and proofs of concept (PoCs) for exploiting UEFI-related vulnerabilities. Building upon these insights, we construct a comprehensive attack threat model encompassing threat actors, attack vectors, attack types, vulnerabilities, attack capabilities, and attacker objectives. Drawing inspiration from the MITRE ATT&CK framework, we present a MITRE ATT&CK-like taxonomy delineating tactics, techniques, and sub-techniques in the context of UEFI attacks. This taxonomy can provide a road map for identifying existing gaps and developing new techniques for rootkit prevention, detection, and removal. Finally, the paper discusses existing countermeasures against UEFI attacks including a variety of technical and operational measures that can be implemented to lower the risk of UEFI attacks to an acceptable level. This paper seeks to clarify the complexities of UEFI and equip the cybersecurity community with the necessary knowledge to strengthen the security of this critical component against a growing threat landscape.

cs.CR↗

RogueRover: Autonomous Rogue Device Localization for Incident Response

Physically localizing unauthorized wireless devices remains a critical bottleneck in cyber-physical security operations, where rogue access points can provide entry points for lateral movement and persistent compromise. While such devices can often be detected through network-side mechanisms, determining their physical location typically requires dense sensing infrastructure, site-specific RF fingerprinting, or manual inspection, limiting timely incident response. We investigate whether a single commodity robot can autonomously detect and localize rogue wireless devices under zero-configuration constraints, without RF fingerprinting, pre-installed sensors, or site calibration. We present RogueRover, an end-to-end system in which a quadruped robot autonomously patrols, collects spatially labeled RSSI measurements via a standard 802.11 interface, and estimates device locations offline. We evaluate the system across 11 patrol runs in a real indoor environment, with 6 rogue devices deployed under heterogeneous propagation conditions. Across 62 AP-patrol sessions, RogueRover achieves a median single-patrol localization error of 1.62 m without prior RF knowledge. Under multi-run aggregation, five of six devices are localized within 1 m. A blind trial validates the full pipeline, correctly identifying rogue devices among 73 observed BSSIDs and localizing them with errors of 0.34 m and 1.84 m. Across environments, simple weighted-centroid estimators perform comparably to, or better than, parametric path-loss models, indicating that measurement coverage from autonomous patrols is the primary determinant of localization accuracy under zero-prior constraints. Our results demonstrate that infrastructure-free, autonomous localization is feasible in practice, enabling rapid physical incident response in cyber-physical environments without additional sensing infrastructure.

cs.RO↗

CEDAR-42001: From ISO/IEC 42001 Conformity to Architecture-Aware, Audit-Visible Assurance Posture for AI Cyber-Physical Systems

AI-enabled cyber-physical systems (AI-CPS) turn data-driven decisions into physical actions, creating assurance challenges across sensing, computation, control, human oversight, and governance. ISO/IEC 42001:2023 specifies requirements for an artificial intelligence management system (AIMS), but conformity assessment alone does not show which architectural layers are affected, whether practices are mature enough for the risk context, or what actions should follow. We present CEDAR-42001 (Control-Evidence Decision and Action Reasoning), a two-stage method that converts ISO/IEC 42001 audit evidence into an architecture-aware assurance posture traceable to the audit record. Stage A preserves the conformity determination. Stage B adds four outputs to each audit row: (i) attribution to a governance stratum or one of seven AI-CPS layers; (ii) a five-dimensional maturity profile with binding-constraint identification; (iii) a risk-proportionate target maturity; and (iv) a rulebook-derived action recommendation. The enriched rows are aggregated into strategic, operational, and tactical decision products. We evaluate CEDAR-42001 using a synthetic autonomous-fleet AIMS and by comparing conformity-only results with the enriched outputs. Although 89.9 percent of audit rows were conforming, only 34.3 percent of conforming rows reached the baseline High-assurance category; across alternative operationalizations, this proportion ranged from 22.4 percent to 46.2 percent. A retrospective application to the 2023 Cruise robotaxi incident shows how the method captures documented concerns across governance, perception, decision-making, and human oversight and maps them to layer-specific actions. CEDAR-42001 does not estimate exploitability or replace technical CPS-security testing; it identifies where audit evidence warrants deeper technical assurance, organizational improvement, or remediation.

cs.CY↗

UEFI Memory Forensics: A Framework for UEFI Threat Analysis

Modern computing systems rely on the Unified Extensible Firmware Interface (UEFI), which has replaced the legacy Basic Input/Output System (BIOS) as the firmware standard for the modern boot process. Although the UEFI represents a significant advancement in system firmware, it is increasingly targeted by threat actors seeking to exploit its execution environment and take advantage of its persistence mechanisms. While some security-related analysis of UEFI components has been performed--primarily via debugging and runtime behavior testing--to the best of our knowledge, no prior study has specifically addressed the capturing and analysis of volatile UEFI runtime memory to detect malicious exploitation during the pre-OS phase. This gap in UEFI forensic tools limits the ability to conduct in-depth security analysis in pre-OS environments. Such a gap is particularly surprising, given that memory forensics is widely regarded as foundational to modern incident response, as reflected by the popularity of above-OS memory analysis frameworks, such as Rekall, Volatility, and MemProcFS. To address the lack of below-OS memory forensics, we introduce a framework for UEFI memory forensics. The proposed framework consists of two components: UEFIMemDump, a memory acquisition tool, and UEFIDumpAnalysis, an extendable collection of analysis modules capable of detecting malicious activities such as function pointer hooking, inline hooking, malicious image loading, and gadget-based control-flow manipulation. Our proof-of-concept implementation demonstrates the framework's ability to detect modern UEFI threats, such as Thunderstrike, CosmicStrand, and Glupteba bootkits. By providing an open-source solution, our work enables researchers and practitioners to investigate firmware-level threats, develop additional analysis modules, and advance overall below-OS security through UEFI memory analysis.

cs.CR↗

Provably Protecting Fine-Tuned LLMs from Training Data Extraction while Preserving Utility

Fine-tuning large language models (LLMs) on sensitive datasets raises privacy concerns, as training data extraction (TDE) attacks can expose highly confidential information. Existing defenses against such attacks either lack formal privacy guarantees or incur substantial utility degradation. We observe that fine-tuning induces widespread probability shifts, yet preserving only a small subset of influential token-level deviations is sufficient; the remaining shifts can be aggressively smoothed with minimal impact on utility. Motivated by this insight, we propose SCP-$Δ_r$, a Near Access Freeness (NAF)-based algorithm that operates on relative probabilities and explicitly smooths low-impact tokens using a base model. SCP-$Δ_r$ achieves orders-of-magnitude better theoretical bounds than existing NAF based methods and provides strong empirical protection against TDE attacks with minimal performance loss.

cs.LG↗

SecMate: Multi-Agent Adaptive Cybersecurity Troubleshooting with Tri-Context Personalization

Recent advances in large language models and agentic frameworks have enabled virtual customer assistants (VCAs) for complex support. We present SecMate, a multi-agent VCA for cybersecurity troubleshooting that integrates device, user, and service specificity from conversational and device-level signals. Device specificity is provided by a lightweight local diagnostic utility, while user specificity relies on implicit proficiency inference and profile-aware troubleshooting. Service specificity is achieved through a proactive, context-aware recommender. We evaluate SecMate in a controlled study with 144 participants and 711 conversations. Device-level evidence increased correct resolutions from about 50% to over 90% relative to an LLM-only baseline, while step-by-step guidance improved pleasantness and reduced user burden. The recommender achieved high relevance (MRR@1=0.75), and participants showed strong willingness to substitute human IT support at costs well below human benchmarks. We release the full code base and a richly annotated dataset to support reproducible research on adaptive VCAs.

cs.CR↗

SAFEdit: Does Multi-Agent Decomposition Resolve the Reliability Challenges of Instructed Code Editing?

Instructed code editing is a significant challenge for large language models (LLMs). On the EditBench benchmark, 39 of 40 evaluated models obtain a task success rate (TSR) below 60 percent, highlighting a gap between general code generation and the ability to perform instruction-driven editing under executable test constraints. To address this, we propose SAFEdit, a multi-agent framework for instructed code editing that decomposes the editing process into specialized roles to improve reliability and reduce unintended code changes. A Planner Agent produces an explicit, visibility-aware edit plan, an Editor Agent applies minimal, literal code modifications, and a Verifier Agent executes real test runs. When tests fail, SAFEdit uses a Failure Abstraction Layer (FAL) to transform raw test logs into structured diagnostic feedback, which is fed back to the Editor to support iterative refinement. We compare SAFEdit against both prior single-model results reported for EditBench and an implemented ReAct single-agent baseline under the same evaluation conditions. We used EditBench to evaluate SAFEdit on 445 code editing instances in five languages (English, Polish, Spanish, Chinese, and Russian) under varying spatial context variants. SAFEdit achieved 68.6 percent TSR, outperforming the single-model baseline by 3.8 percentage points and the ReAct single-agent baseline by 8.6 percentage points. The iterative refinement loop was found to contribute 17.4 percentage points to SAFEdit's overall success rate. SAFEdit's automated error analysis further indicates a reduction in instruction-level hallucinations compared to single-agent approaches, providing an additional framework component for interpreting failures beyond pass or fail outcomes.

cs.SE↗

LumiMAS: A Comprehensive Framework for Real-Time Monitoring and Enhanced Observability in Multi-Agent Systems

The incorporation of LLMs in multi-agent systems (MASs) has the potential to significantly improve our ability to autonomously solve complex problems. However, such systems introduce unique challenges in monitoring, interpreting, and detecting system failures. Most existing MAS observability frameworks focus on analyzing each individual agent separately, overlooking failures associated with the entire MAS. To bridge this gap, we propose LumiMAS, a novel MAS observability framework that incorporates advanced analytics and monitoring techniques. The proposed framework consists of three key components: a monitoring and logging layer, anomaly detection layer, and anomaly explanation layer. LumiMAS's first layer monitors MAS executions, creating detailed logs of the agents' activity. These logs serve as input to the anomaly detection layer, which detects anomalies across the MAS workflow in real time. Then, the anomaly explanation layer performs classification and root cause analysis (RCA) of the detected anomalies. LumiMAS was evaluated on seven different MAS applications, implemented using two popular MAS platforms, and a diverse set of possible failures. The applications include two novel failure-tailored applications that illustrate the effects of a hallucination or bias on the MAS. The evaluation results demonstrate LumiMAS's effectiveness in failure detection, classification, and RCA.

cs.CR↗

From Tool Orchestration to Code Execution: A Study of MCP Design Choices

Model Context Protocols (MCPs) provide a unified platform for agent systems to discover, select, and orchestrate tools across heterogeneous execution environments. As MCP-based systems scale to incorporate larger tool catalogs and multiple concurrently connected MCP servers, traditional tool-by-tool invocation increases coordination overhead, fragments state management, and limits support for wide-context operations. To address these scalability challenges, recent MCP designs have incorporated code execution as a first-class capability, an approach called Code Execution MCP (CE-MCP). This enables agents to consolidate complex workflows, such as SQL querying, file analysis, and multi-step data transformations, into a single program that executes within an isolated runtime environment. In this work, we formalize the architectural distinction between context-coupled (traditional) and context-decoupled (CE-MCP) models, analyzing their fundamental scalability trade-offs. Using the MCP-Bench framework across 10 representative servers, we empirically evaluate task behavior, tool utilization patterns, execution latency, and protocol efficiency as the scale of connected MCP servers and available tools increases, demonstrating that while CE-MCP significantly reduces token usage and execution latency, it introduces a vastly expanded attack surface. We address this security gap by applying the MAESTRO framework, identifying sixteen attack classes across five execution phases-including specific code execution threats such as exception-mediated code injection and unsafe capability synthesis. We validate these vulnerabilities through adversarial scenarios across multiple LLMs and propose a layered defense architecture comprising containerized sandboxing and semantic gating. Our findings provide a rigorous roadmap for balancing scalability and security in production-ready executable agent workflows.

cs.CR↗