AX·NEWS
Sign in

Cybersecurity

Cybersecurity news — Wednesday 30 September 2026

141 stories from 10 sources, every one in English. Each links to its publisher.

Open in the live feedRSS feedOn a TV

  1. FTC is Investigating OpenAI and Anthropic Over Possible risks to Consumers
    USSecurityWeek

    An FTC spokesperson confirmed the investigation but declined further comment. The post FTC is Investigating OpenAI and Anthropic Over Possible Risks to Consumers appeared first on SecurityWeek .

  2. OpenAI reveals ‘novel’ encryption bypass used in distillation attack 
    USCyberScoop Asia

    OpenAI said it disrupted a “coordinated campaign” to distill and extract reasoning capabilities from its AI models, pointing the finger at a Chinese rival. On Wednesday, OpenAI said it first spotted low-level activity on July 1 that…

  3. Russian state hackers use new RedFlick technique to push malware
    USBleepingComputer

    The Russian state actor Star Blizzard has been using a new malware installation tactic dubbed "RedFlick" to deploy its signature CosmicPulse backdoor. [...]

  4. DIVD says Zammad zero-days enabled AI-driven network breach
    USBleepingComputer

    The Dutch Institute for Vulnerability Disclosure (DIVD) says that the breach of its network was possible by exploiting a chain of two zero-day vulnerabilities in the open-source Zammad ticketing system. [...]

  5. Over 543,000 valid credentials exposed in public GitHub repositories
    USBleepingComputer

    More than 543,000 credentials exposed in public GitHub repositories were still valid in July despite the platform's security measures to prevent accidental leaks of sensitive data. [...]

  6. openSUSE Leap adds a new security layer: Immutable mode
    USZDNet Security

    One of the more secure Linux distributions has added another feature to help further lock it down.

  7. Attackers Exploit Zimbra Flaw to Deploy Web Shells and Harvest Authentication Secrets
    INThe Hacker News

    Threat actors have weaponized a now-patched security flaw in Zimbra Collaboration Suite (ZCS) to deploy web shells and access mailbox data, according to findings from the Microsoft Security Research team. The attack exploits CVE-2026-73570…

  8. Attackers Abuse MSP360 to Deploy ScreenConnect in Dual-RMM Phishing Attacks
    INThe Hacker News

    Microsoft has warned of phishing campaigns distributing an installer for the MSP360 Remote Monitoring and Management (RMM) software under the guise of meeting invitations, PDF-themed lures, software update prompts, and other…

  9. CISA warns of critical pre-auth RCE flaw in MikroTik RouterOS
    USBleepingComputer

    The U.S. Cybersecurity and Infrastructure Security Agency (CISA) is warning of a new critical vulnerability in MikroTik RouterOS that could lead to remote code execution or cause a denial-of-service condition. [...]

  10. Your iPhone just got a hidden anti-scam upgrade in iOS 27: How to enable it
    USZDNet Security

    Impersonation Risk Detection tries to spot scams traditional security methods might miss.

  11. Hackers steal protective order and foster care records from Arizona courts
    USMalwarebytes Labs

    “Arizona’s court system was targeted by a cyber attack from criminal hackers or their bots.” This is how the Arizona Supreme Court announced that hackers had attacked the state’s court system and stolen the personal information of “many…

  12. Cisco Warns of Attackers Exploiting Critical Authentication Bypass in SD-WAN Manager
    INThe Hacker News

    Attackers are exploiting a new critical zero-day flaw in Cisco Catalyst SD-WAN Manager, the system companies use to manage their Cisco SD-WAN networks, Cisco said in an advisory on September 30. The flaw, CVE-2026-76504, could allow a…

  13. Attackers Abuse ChatGPT Custom GPTs to Deliver RAT via ClickFix Lures
    INThe Hacker News

    Threat actors are abusing ChatGPT Custom GPTs to disguise them as legitimate product offerings and direct unsuspecting victims to malicious sites that employ ClickFix lures to deliver malware. Huntress, which observed the activity in late…

  14. Cisco warns of new SD-WAN zero-day exploited in attacks
    USBleepingComputer

    Cisco released security updates to address a critical zero-day in the Catalyst SD-WAN Manager (tracked as CVE-2026-76504) that attackers are actively exploiting to escalate to admin privileges. [...]

  15. Google: AI Is Changing the Pace and Profile of Vulnerability Discovery
    USSecurityWeek

    Google’s analysis found that AI-discovered vulnerabilities are more likely to enable remote code execution. The post Google: AI Is Changing the Pace and Profile of Vulnerability Discovery appeared first on SecurityWeek .

  16. AI's Third Wave: Coworkers Break the Security Model That Worked for Agents
    USBleepingComputer

    Persistent AI coworkers may operate continuously with standing access, creating identity risks that existing security models were not designed to handle. Token Security explains why these agents need their own identities, owners, scoped…

  17. Microsoft to block Entra ID script injection attacks starting October
    USBleepingComputer

    Microsoft has reminded customers that the Entra ID authentication system will get better protection against external script injection attacks starting next month. [...]

  18. WatchGuard Patches Critical Fireware OS Code Injection Vulnerability
    USSecurityWeek

    WatchGuard has rolled out patches for 15 code execution, DoS, authorization, and path traversal bugs in Fireware OS. The post WatchGuard Patches Critical Fireware OS Code Injection Vulnerability appeared first on SecurityWeek .

  19. Government, Finance Orgs Targeted in Weeks-Long NetScaler Zero-Day Attacks
    USSecurityWeek

    Several security firms have confirmed seeing exploitation of the NetScaler vulnerabilities CVE-2026-88771 and CVE-2026-88772. The post Government, Finance Orgs Targeted in Weeks-Long NetScaler Zero-Day Attacks appeared first on…

  20. Pearson’s Workera deal targets a big workplace problem: No time to AI upskill
    USZDNet Security

    Pearson has agreed to acquire Workera to help businesses assess employees’ AI skills and target training where it’s needed most.

  21. TeamViewer urges users to patch severe flaws “as soon as possible”
    USBleepingComputer

    Remote access software company TeamViewer warned customers on Tuesday to immediately patch a set of high-severity vulnerabilities affecting its client and host software. [...]

  22. Chrome, Firefox Updates Patch Over 100 Vulnerabilities
    USSecurityWeek

    Some of the flaws could allow remote attackers to execute arbitrary code or escape the browser sandbox. The post Chrome, Firefox Updates Patch Over 100 Vulnerabilities appeared first on SecurityWeek .

  23. Know Your Enemy: Browser-Based Attack Techniques in 2026
    INThe Hacker News

    Given that the browser is where business apps are accessed and used, it makes sense that attacks are happening there too. Most breaches today begin in a browser session. Often, they never leave it, with the entire attack chain from initial…

  24. AI Coding Agents Exposed 13,000 Internal Images, Including Billing Records, on GitHub
    INThe Hacker News

    AI coding agents asked to share screenshots of code changes for review have put internal company images in public GitHub repositories, security company Glow said. Its researchers found more than 13,000 internal images from developers at…

  25. Your car’s app could be telling Big Tech who you are and where you go
    USMalwarebytes Labs

    A study testing 21 cars from 19 brands and 30 companion apps found connections to advertising and tracking companies, and evidence that some apps shared sensitive personal data. Modern cars can unlock remotely, route around traffic, stream…

  26. Anthropic Flags AI Agent Liability Risks as OpenAI Faces Hacking Lawsuit
    USSecurityWeek

    Attacks by autonomous AI agents are moving out of the lab and into the courtroom, raising unsettled questions about who is liable for what agents do. The post Anthropic Flags AI Agent Liability Risks as OpenAI Faces Hacking Lawsuit…

  27. Bitget hacked via zero-day in third-party security products
    USBleepingComputer

    Cryptocurrency exchange Bitget revealed today that attackers who stole $387.5 million last week breached its systems after exploiting a zero-day flaw in third-party security products. [...]

  28. I Want Better Reporting on AI Genie Behavior
    USSchneier on Security

    AI systems are regularly completing tasks in ways that their prompters don’t want or intend. Some of them are disturbing, and some of them are dangerous. This is something I’ve been calling “ genie behavior ,” because I think that really…

  29. Russian APT Star Blizzard Uses ‘RedFlick’ Infection Chain in Recent Attacks
    USSecurityWeek

    The state-sponsored group has launched larger-scale phishing campaigns to deploy the CosmicPulse backdoor. The post Russian APT Star Blizzard Uses ‘RedFlick’ Infection Chain in Recent Attacks appeared first on SecurityWeek .

  30. US-Focused CSuite Phishing Steals Microsoft 365 Sessions and Deploys RMM Tools for Remote Access
    INThe Hacker News

    ANY.RUN researchers traced a US-focused CSuite phishing campaign across 351 sandbox analyses, with 51% of submissions coming from the United States. Technology, manufacturing, government, and consulting organizations showed the highest…

  31. ShinyHunters Defiant After FBI Calls on Members to Come Forward
    USSecurityWeek

    In the wake of a suspected leader’s arrest, ShinyHunters says it never intended to publish data stolen from the FBI. The post ShinyHunters Defiant After FBI Calls on Members to Come Forward appeared first on SecurityWeek .

  32. WaterISAC reckons with range of threats after summer of cyberattacks
    USCyberScoop Asia

    The computers that automate water treatment systems across the country were built for durability, not for an internet-connected world. Many still accomplish their fundamental function, but as cyberattacks on the sector mounted this summer…

  33. Attackers Exploit NetScaler Flaw for Root Access, Deploy WHIPSHOT and SLAPSHOT
    INThe Hacker News

    Unknown threat actors have been observed exploiting a newly patched security flaw in Citrix NetScaler ADC and NetScaler Gateway appliances to target organizations in North America and Europe. The activity, observed by Mandiant Consulting…

  34. OpenSSL Fixes High-Severity DTLS Flaw That Can Leak Heap Memory Unencrypted
    INThe Hacker News

    A High-severity OpenSSL flaw can leak heap memory to the other side of a DTLS connection or crash the program, OpenSSL said on September 29 as it released fixes. DTLS, the TLS variant used for UDP traffic, resends a handshake message if no…

  35. High-Severity Vulnerabilities Patched in OpenSSL, WolfSSL
    USSecurityWeek

    Roughly a dozen vulnerabilities have been patched in each of the open source cryptographic libraries. The post High-Severity Vulnerabilities Patched in OpenSSL, WolfSSL appeared first on SecurityWeek .

  36. Citrix NetScaler CVE-2026-88772 Exploit Details Show Pre-Auth Path to Shellcode Execution
    INThe Hacker News

    Cybersecurity researchers have disclosed technical details of a recently patched critical security flaw in Citrix NetScaler ADC and Gateway that has come under active exploitation in the wild. The vulnerability, tracked as CVE-2026-88772…

  37. Selective Channel Restoration for Backdoored Vision-Language Models
    USarXiv cs.CR (Security)

    arXiv:2609.37759v1 Announce Type: new Abstract: Vision-language models (VLMs) exhibit strong multimodal capabilities but remain vulnerable to backdoors implanted through poisoned fine-tuning data. Existing defenses often require extensive…

  38. Hamming Ideals and Grobner Bases for ISD-like Syndrome Decoding
    USarXiv cs.CR (Security)

    arXiv:2609.18866v2 Announce Type: replace Abstract: We investigate an algebraic approach to the Syndrome Decoding Problem, based on a reformulation of the Hamming weight constraint and its integration with the Information Set Decoding…

  39. She Spoofed Sea Ships by the Sea Shore: Measuring Large-Scale GPS Spoofing in Global Maritime Traffic
    USarXiv cs.CR (Security)

    arXiv:2609.37676v1 Announce Type: new Abstract: GPS spoofing has emerged as a serious threat to maritime security, yet its global prevalence, persistence, and structure remain largely unmeasured. In this paper, we present the first…

  40. GenomeOcean Anywhere: Private WebGPU Inference for Genome MoEs
    USarXiv cs.CR (Security)

    arXiv:2609.35882v1 Announce Type: new Abstract: Genome foundation models are most useful where sequences are generated, yet the largest models need datacenter accelerators and a place to send private DNA. We ask whether a…

  41. Agentic Commerce Bench: Measuring Fraud Detection for Agents That Spend Money
    USarXiv cs.CR (Security)

    arXiv:2609.35886v1 Announce Type: new Abstract: AI agents now hold spend authority and settle payments without per-action human confirmation. The resulting loss is often not a security failure: a counterparty with the correct domain, the…

  42. Privacy-Friendly Cohort Determination: Sealed, CSP-Independent In-Browser ML Inference of Professional Segments for Identity-Less Advertising
    USarXiv cs.CR (Security)

    arXiv:2609.36153v1 Announce Type: new Abstract: B2B advertising targets a viewer's professional attributes (employer size and industry, function, seniority) and has obtained them by matching identities across sites. Safari and Firefox…

  43. DecoyTrace: Toxic Decoys for Active Defense in Decentralized Federated Learning
    USarXiv cs.CR (Security)

    arXiv:2609.36330v1 Announce Type: new Abstract: Decentralized Federated Learning (DFL) eliminates the central aggregation server, reducing the single point of observation that traditional defenses against attacks rely on. As a result…

  44. CounterSteer: Suppressing Indirect Prompt Injection with Activation Steering
    USarXiv cs.CR (Security)

    arXiv:2609.36570v1 Announce Type: new Abstract: Indirect prompt injection makes an LLM agent treat untrusted retrieved text as instructions. We present CounterSteer, an inference-time defense that suppresses this behavior inside the model…

  45. Deep Learning Latency Attacks and Defenses: A Cross-Domain Survey of Availability Threats
    USarXiv cs.CR (Security)

    arXiv:2609.36732v1 Announce Type: new Abstract: Adversarial machine learning has focused mainly on integrity, but availability is an increasingly consequential complement. Latency attacks (also energy-latency attacks) increase…

  46. Backdoor Mitigation in Decentralized LLM Fine-Tuning
    USarXiv cs.CR (Security)

    arXiv:2609.37367v1 Announce Type: new Abstract: Decentralized large language model (LLM) fine-tuning lets organizations collaboratively train a shared LLM on data they cannot pool, without a central coordinator. In every round, each node…

  47. Lights, Camera, Attack: Exploiting Temporal HDR Fusion with Pulsed Light
    USarXiv cs.CR (Security)

    arXiv:2609.37742v1 Announce Type: new Abstract: Modern cameras widely use temporal High Dynamic Range (HDR) to improve visibility by capturing a sequence of exposures with different integration times and fusing them into a single image…

  48. Making Duplicate Reimbursement Unrepresentable: A Verified Ethereum E-Invoice System for Humans and AI Agents
    USarXiv cs.CR (Security)

    arXiv:2609.37819v1 Announce Type: new Abstract: Electronic invoices are replacing paper invoices worldwide, but today's centralized architectures leave three problems unsolved on the consumption side: an invoice can be submitted for…

  49. Dagger: Decoupling-based Model Stealing Attack against Graph Neural Networks
    USarXiv cs.CR (Security)

    arXiv:2609.37972v1 Announce Type: new Abstract: As Graph Neural Networks (GNNs) are widely deployed as Machine Learning-as-a-Service (MLaaS) APIs, model stealing attacks have emerged as a critical security threat. By querying a victim…

  50. Epistemic Typing as a PostgreSQL Table Access Method: Adversarial Conflict Resolution Under Confidence Forgery and Sybil Coordination
    USarXiv cs.CR (Security)

    arXiv:2609.36795v1 Announce Type: cross Abstract: We describe KNDB, a PostgreSQL 18 table access method (TAM) that types every row with an engine-assigned epistemic kind (MEASURED, INFERRED, or DERIVED) and resolves per-slot conflicts…

  51. MCPTox: A Benchmark for Tool Poisoning Attack on Real-World MCP Servers
    USarXiv cs.CR (Security)

    arXiv:2508.14925v2 Announce Type: replace Abstract: By providing a standardized interface for LLM agents to interact with external tools, the Model Context Protocol (MCP) is quickly becoming a cornerstone of the modern autonomous agent…

  52. GAZEleak: Passcode Inference Against Eye-tracking XR Devices Through External Observation
    USarXiv cs.CR (Security)

    arXiv:2609.35040v2 Announce Type: replace Abstract: Mixed-reality headsets such as Apple Vision Pro replace the touch screen with gaze-as-pointer interaction: the wearer looks at a target and confirms with an air pinch. Because the display…

  53. Unconditional Unclonable Encryption
    USarXiv cs.CR (Security)

    arXiv:2607.21551v2 Announce Type: replace-cross Abstract: We give an unconditional construction of information-theoretically secure one-time private-key unclonable encryption scheme for one-bit messages, with efficient encryption and…

  54. Toward provably private learning from federated data
    USarXiv cs.CR (Security)

    arXiv:2609.31494v3 Announce Type: replace Abstract: Federated Learning (FL) allows devices with private data to collaborate in training a shared model. We present a next-generation FL system based on Trusted Execution Environments (TEEs)…

  55. Cheap to Hypothesize, Costly to Verify: The Defense Surface of Agentic Vulnerability Discovery
    USarXiv cs.CR (Security)

    arXiv:2609.35909v1 Announce Type: new Abstract: Autonomous LLM agents turn vulnerability discovery into a repository-scale search: they generate many vulnerability hypotheses but can verify only a subset under a finite budget. We show that…

  56. MMSkillRisk: Can Agents Stay Safe When Multimodal Skills Become Traps?
    USarXiv cs.CR (Security)

    arXiv:2609.35912v1 Announce Type: new Abstract: Agent skills are shareable packages of procedural instructions, tools, and examples. Multimodal skills additionally include visual references that agents retrieve and inspect during…

  57. Same Bytes, Different Authority: Reserved-Token Representations in Chat-Template Prompt Injection
    USarXiv cs.CR (Security)

    arXiv:2609.35932v1 Announce Type: new Abstract: Prompt injection against LLM agents becomes much stronger when the injected instruction is wrapped in the model's own chat template. A forged template marker such as <|im_start|> can reach…

  58. VStress: Correlation-Aware Auditing and Adaptive Budget Allocation for Repeated Verifiers
    USarXiv cs.CR (Security)

    arXiv:2609.36958v1 Announce Type: cross Abstract: Repeated verifier calls are useful only when they contribute conditional information. We introduce VStress, an auditable replay contract, and VStress-CA, a correlation-aware allocation…

  59. Benign Fine-Tuning Breaks Safety Alignment in Audio LLMs
    USarXiv cs.CR (Security)

    arXiv:2604.16659v2 Announce Type: replace Abstract: Fine-tuning on benign data is known to degrade safety alignment in text and vision LLMs, but whether distinct input properties drive this vulnerability differently remains unclear. Audio…

  60. XPhysICS: Cross-Physical-Domain Threat Grounding for Industrial Control Systems Security
    USarXiv cs.CR (Security)

    arXiv:2609.30805v2 Announce Type: replace Abstract: Industrial control system attacks are usually documented in terms of the plant where they occurred: its sensors, actuators, process stages, and control logic. Yet many attacks express a…

  61. Raising the Bar for Chinese Adolescent LLM Safety: A Culturally-Grounded, Fine-Grained Benchmark
    USarXiv cs.CR (Security)

    arXiv:2609.35902v1 Announce Type: new Abstract: Safety risks in conversations with adolescents are not always explicit. A request may appear harmless unless a model considers the user's age, circumstances, and earlier turns. Existing…

  62. PrivacySkills: How Privacy Guidance Shapes Source Selection in LLM Agents
    USarXiv cs.CR (Security)

    arXiv:2609.35937v1 Announce Type: new Abstract: While prior work has documented privacy failures in LLM agents, it remains unclear how the presentation of privacy guidance influences their choice of information sources. We introduce…

  63. Multi-Class, Multi-Tier Network Intrusion Detection: A Comprehensive and Reproducible Benchmark
    USarXiv cs.CR (Security)

    arXiv:2609.36039v1 Announce Type: new Abstract: Machine learning (ML) and deep learning (DL) have dominated Intrusion Detection System (IDS) research in recent years. Unfortunately, many existing studies have produced inflated results and…

  64. Know the Normal, Track the Attack: Context-Grounded and Stateful LLM Investigation over System Provenance
    USarXiv cs.CR (Security)

    arXiv:2609.36494v1 Announce Type: new Abstract: Provenance-based intrusion detection systems (PIDSs) identify suspicious activity in audit streams, but their outputs remain difficult to turn into coherent attack narratives. Direct LLM…

  65. CyberPersistBench: Evaluating LLM-Based Cyber Attackers on Installation and Persistence
    USarXiv cs.CR (Security)

    arXiv:2609.36573v1 Announce Type: new Abstract: While LLM-based attackers exhibit growing proficiency in vulnerability exploitation, most existing cybersecurity benchmarks suffer from single-stage truncation, prematurely terminating…

  66. Does the Unsafe Gradient Survive a Conversation? On the Fragility of Gradient-Based Jailbreak Detection in Multi-Turn Dialogue
    USarXiv cs.CR (Security)

    arXiv:2609.36849v1 Announce Type: new Abstract: Safety-aligned language models are commonly deployed as multi-turn assistants, which lets adversaries spread unsafe intent across several user turns instead of a single prompt. Gradient-based…

  67. AutoMark: Enabling Autoresearch to Discover Better LLM Watermarks
    USarXiv cs.CR (Security)

    arXiv:2609.37310v1 Announce Type: new Abstract: With LLM watermarking being deployed commercially and now required by regulations, improving its reliability and effectiveness has become crucial. Yet, recent progress in the field of LLM…

  68. Where Do LLMs Decide to Break the Rules? Mechanistic Localization of Prompt Injection Compliance
    USarXiv cs.CR (Security)

    arXiv:2609.37737v1 Announce Type: new Abstract: When a prompt injection attack succeeds, a Large Language Model (LLM) abandons its assigned system role to comply with an adversarial instruction. While prior work has extensively quantified…

  69. SIREN (Luring LLMs onto the Rocks): PAIR-Driven Preference Manipulation in Web-RAG Recommenders
    USarXiv cs.CR (Security)

    arXiv:2607.21951v3 Announce Type: replace-cross Abstract: This paper investigates the adversarial manipulation of the ranked recommendations produced by web-augmented large language models (LLMs). When an LLM answers a recommendation query…

  70. Harvest Season for SLUB: From io_uring vulnerability to Novel Sheaf-Based Exploitation Techniques
    USarXiv cs.CR (Security)

    arXiv:2609.37608v1 Announce Type: new Abstract: The Linux kernel's push for higher I/O performance and more efficient memory management has introduced new mechanisms that, while improving performance, also open new attack surfaces. This…

  71. TRACE: Task-Aware Adaptive Self-Evolving Agentic Jailbreaking
    USarXiv cs.CR (Security)

    arXiv:2605.30883v2 Announce Type: replace Abstract: The rise of LLM agents introduces a new threat by enabling planning, coding, and even end-to-end execution of expert-level attack workflows. However, this threat remains underexplored and…

  72. JUMP: Efficient Membership Inference on Fine-Tuned Diffusion Language Models
    USarXiv cs.CR (Security)

    arXiv:2607.16207v3 Announce Type: replace-cross Abstract: Membership inference attacks (MIAs) test whether a candidate example was used to train a language model. Existing attacks on fine-tuned discrete diffusion language models (dLLMs)…

  73. Efficient Linkage-Based Compartmentalization on CHERI
    USarXiv cs.CR (Security)

    arXiv:2609.36731v1 Announce Type: new Abstract: We present an efficient linkage-based model for in-process compartmentalization built on CHERI memory safety, which enables fine-grained compartmentalization of the entire UNIX user-space…

  74. Smooth Sailing through Spherical Shells: Provable Random-Lattice Sieving in Time $2^{0.292n}$
    USarXiv cs.CR (Security)

    arXiv:2609.37482v1 Announce Type: new Abstract: In an attempt to close the gap between the best provable and heuristic algorithms for hard lattice problems, such as the shortest (SVP) and closest vector problem (CVP), we analyze lattice…

  75. pikit: A Composable Toolkit for Indirect Prompt Injection Research and Evaluation
    USarXiv cs.CR (Security)

    arXiv:2609.36817v1 Announce Type: new Abstract: Indirect prompt injection embeds malicious instructions within external content retrieved by LLM-based agents, altering target behavior without user authorization. We introduce pikit, a…

  76. OpenAI-HuggingFace: A Reproduction ＆ Lessons for Alignment Testing
    USarXiv cs.CR (Security)

    arXiv:2609.35799v1 Announce Type: cross Abstract: In July 2026, OpenAI's agents coordinated over channels outside their intended environment to breach Hugging Face's secured infrastructure. Could existing alignment testing practices have…

  77. CoSec: Benchmarking Agent Security in Communities
    USarXiv cs.CR (Security)

    arXiv:2609.34790v2 Announce Type: replace Abstract: LLM agents operate in persistent collaborative environments involving multiple users, communities, memories, files, and tools. Community boundaries may remain fixed or evolve with changes…

  78. CipherGenome: Homomorphic Inference for Genomic Mixture-of-Experts
    USarXiv cs.CR (Security)

    arXiv:2609.35883v1 Announce Type: new Abstract: Genome foundation models are growing into sparse mixture-of-experts (MoE) networks whose expert weights no longer fit on the machines that hold the sequences, yet sending a private genome to…

  79. GenoTrace: Inheritable Watermarks for Genome Foundation Model Distillation
    USarXiv cs.CR (Security)

    arXiv:2609.35881v1 Announce Type: new Abstract: Can a genome model retain a detectable record of the synthetic sequences used to train it? We study watermark inheritance through distillation with GenoTrace, a codon-aware extension of…

  80. SINGED: Correct Outputs Do Not Certify Safe Execution in LLM Agents
    USarXiv cs.CR (Security)

    arXiv:2609.35889v1 Announce Type: new Abstract: Tool-using language-model agents select and execute third-party artifacts. Different implementations can return the requested output while producing hidden execution effects that task…

  81. When Privacy Becomes a Weapon: Understanding Doxxing and Privacy Vulnerabilities in Mainland China's Social Media Ecosystem
    USarXiv cs.CR (Security)

    arXiv:2609.35951v1 Announce Type: new Abstract: Doxxing, the malicious disclosure of personal information, has become a pervasive privacy threat. Yet existing research remains predominantly Western-centric, limiting our understanding of…

  82. Adversarial Debiasing of Machine Learning Models for Enhanced Network Security against DDoS Attacks
    USarXiv cs.CR (Security)

    arXiv:2609.36167v1 Announce Type: new Abstract: Distributed Denial of Service attacks are a growing threat to network infrastructure, and new techniques, including the use of generative AI, make them harder to detect. Traditional detection…

  83. TTMark: Pairwise Distortion-Free Watermarking Beyond Single-Token Entropy
    USarXiv cs.CR (Security)

    arXiv:2609.36372v1 Announce Type: new Abstract: Distortion-free watermarking enables reliable attribution of machine-generated text while preserving output distribution. However, existing methods operate independently on each generated…

  84. Controlled Decoding Attacks on Black-Box LLMs
    USarXiv cs.CR (Security)

    arXiv:2609.36956v1 Announce Type: new Abstract: Manipulating next-token probabilities during generation can bypass the safety alignment of large language models. Existing approaches, however, rely on access to model weights or numerical…

  85. When Cyber Scoring Systems Diverge: An Empirical Comparison
    USarXiv cs.CR (Security)

    arXiv:2609.37217v1 Announce Type: new Abstract: Vulnerability scoring systems underpin cyber patch prioritization and risk management, but their comparative behavior is almost always assessed in the abstract, through correlation studies in…

  86. Low Reasoning Effort Is Enough for Routine Office Work by Language-Model Agents
    USarXiv cs.CR (Security)

    arXiv:2608.03169v2 Announce Type: replace Abstract: Purpose: Developers choose how much a language-model agent reasons before it acts. Some pick a high level for fear that a low one is not enough, and pay for it in tokens, time and…

  87. Poking Around in the Dark: Why a Shared Understanding of Components Matters
    USarXiv cs.CR (Security)

    arXiv:2606.02442v3 Announce Type: replace-cross Abstract: By listing the components included in an application, Software Bills of Materials (SBOMs) are intended to support the timely identification of vulnerable components and ensure the…

  88. SameFact: The Same Safety Facts Lead to Different Responses Across Interfaces
    USarXiv cs.CR (Security)

    arXiv:2609.35872v1 Announce Type: new Abstract: Safety evaluations often ask whether a model recognizes that an action is unsafe, whereas agent evaluations ask what the model chooses to do. Using safety judgments as evidence about action…

  89. Dual-Fit Imperative in Security Leadership: A Grounded Theory Investigation of CISO Role Enactment in Modern Organisations
    USarXiv cs.CR (Security)

    arXiv:2609.35887v1 Announce Type: new Abstract: The Chief Information Security Officer (CISO) has emerged as a strategically prominent executive role, confronting an adversarial environment, irreconcilable accountability tensions between…

  90. Calibrating One-Round Membership Inference with Neighbors
    USarXiv cs.CR (Security)

    arXiv:2609.36331v1 Announce Type: new Abstract: The state-of-the-art Membership Inference (MI) methods calibrate their signal separately for each example using reference models, auxiliary models trained to exclude the target. This paradigm…

  91. SKILLLITE: Evidence-Guided Malicious Skill Auditing with Compact LLMs
    USarXiv cs.CR (Security)

    arXiv:2609.36879v1 Announce Type: new Abstract: As LLM-based agents perform increasingly complex tasks, Agent Skills have emerged as a flexible mechanism for extending their capabilities. An Agent Skill packages task-specific instructions…

  92. ToolFence: Fine-Grained Authorization for Secure Tool-Using LLM Agents
    USarXiv cs.CR (Security)

    arXiv:2609.37196v1 Announce Type: new Abstract: Tool-using LLM agents remain vulnerable to indirect prompt injection because trusted instructions and untrusted observations share one context, allowing malicious content to steer…

  93. Understanding Private Evolution as Learning-Augmented Clustering
    USarXiv cs.CR (Security)

    arXiv:2609.36678v1 Announce Type: cross Abstract: Private Evolution (PE) is a differentially private algorithm for synthetic data generation. While it can be viewed as a Wasserstein learning algorithm, it performs much better in practice…

  94. Collision Detection is Instance $\widetilde{O}$ptimal Under the Birthday Threshold
    USarXiv cs.CR (Security)

    arXiv:2609.37342v1 Announce Type: cross Abstract: Can structural knowledge about a hash function help accelerate the (black box) detection of collisions in it? This question is fundamental to cryptography theory given the importance of…

  95. DARWIN: Evolving Jailbreak Adversary and Guardrail for LLM Safety Evaluation and Protection
    USarXiv cs.CR (Security)

    arXiv:2607.19829v2 Announce Type: replace Abstract: Most existing LLM safety evaluation and defense methods are static: jailbreak vulnerabilities are assessed with fixed attacks, and guardrails are trained on fixed malicious-prompt…

  96. Self-Evolving Defense: Continual Security Policy Learning for LLM Agents
    USarXiv cs.CR (Security)

    arXiv:2609.36603v1 Announce Type: new Abstract: Large language models (LLMs) increasingly power agents that access sensitive information, use external tools, and modify software repositories. Although these capabilities offer substantial…

  97. An Empirical Study and Assessment of EU AI Act Compliance Checkers
    USarXiv cs.CR (Security)

    arXiv:2609.36228v1 Announce Type: cross Abstract: The EU AI Act introduces extensive compliance requirements for organizations that develop, deploy, or integrate AI systems. Many of these requirements are directly relevant to security and…

  98. Render Before Reading: Visual Rendering as a Prompt Injection Defense
    USarXiv cs.CR (Security)

    arXiv:2609.36121v1 Announce Type: new Abstract: Large language models are vulnerable to prompt injection attacks, where third-party adversarial content can hijack the model's behavior. In this paper, we study the role played by the…

  99. Backdoor in the Loop: Compromising Agentic Search via Malicious Retrievers
    USarXiv cs.CR (Security)

    arXiv:2609.37468v1 Announce Type: new Abstract: Agentic retrieval-augmented generation (RAG) interleaves reasoning with repeated retrieval, giving the retriever influence over both the evidence an agent observes and its subsequent search…

  100. Beyond Semantic Narrowing: Robust and Efficient LLM Watermarking with Hamming Neighborhoods
    USarXiv cs.CR (Security)

    arXiv:2609.37218v1 Announce Type: new Abstract: Semantic watermarking improves robustness against watermark removal attacks by embedding detectable signals into sentence-level representations. However, existing watermarking methods…

  101. FinRT: Distilling Adaptive Red-Teaming Strategies into Reusable Adversarial Generators in Consumer Finance
    USarXiv cs.CR (Security)

    arXiv:2609.36474v1 Announce Type: cross Abstract: In regulated industries like consumer finance, seemingly harmless user queries can exploit large language model vulnerabilities, triggering safety failures and pushing responses dangerously…

  102. Quantum Leakage Resilience of Shamir Secret Sharing
    USarXiv cs.CR (Security)

    arXiv:2609.37276v1 Announce Type: cross Abstract: We initiate the study of quantum leakage resilience of unmodified Shamir secret sharing over prime fields. A well-studied leakage model for Shamir's secret sharing classically is single-bit…

  103. Mirage: a Clean-Label Backdoor against LiDAR 3D Object Detection
    USarXiv cs.CR (Security)

    arXiv:2606.20752v2 Announce Type: replace-cross Abstract: Deep neural network-based LiDAR 3D object detection serves as a critical perception component in safety-critical autonomous systems. However, recent studies have revealed its…

  104. SaplingGuard: A Multidimensional-Profile-Aware Multi-Agent Guardrail for Developmentally Safe Adolescent-LLM Interaction
    USarXiv cs.CR (Security)

    arXiv:2609.35892v1 Announce Type: new Abstract: As adolescents increasingly use LLMs in everyday life, ensuring safe and developmentally appropriate responses has become essential. However, existing LLM guardrails primarily target explicit…

  105. Says Block, Still Acts: Why LLM Safety Judgments Fail to Govern Action in LLM Agents
    USarXiv cs.CR (Security)

    arXiv:2609.35870v1 Announce Type: new Abstract: Large language model agents can correctly judge that an action should be blocked while still preferring to take it. We ask why this judgment-action disconnect arises, and whether explicit…

  106. Similarity Is Not Validity: Defending LLM Semantic Caches Against Poisoning
    USarXiv cs.CR (Security)

    arXiv:2609.35908v1 Announce Type: new Abstract: Semantic caches reduce LLM serving costs by reusing previously generated answers for semantically similar queries. However, retrieval is based solely on embedding similarity between the…

  107. Why Backdooring Neural Networks is so Easy?
    USarXiv cs.CR (Security)

    arXiv:2609.36117v1 Announce Type: cross Abstract: Securing modern AI systems against backdoor attacks remains an open challenge and requires fundamentally principled estimates of the adversary's budget -- the poison fraction $\pi$ and…

  108. Conversable Quantum Fire in the Standard Model
    USarXiv cs.CR (Security)

    arXiv:2609.36335v1 Announce Type: cross Abstract: Quantum fire consists of quantum states that can be efficiently functionally cloned but cannot be transmitted using one-way classical communication. Whereas all previous quantum-fire…

  109. From Reconnaissance to Response: Quantitative Risk Parameterization and Game Theoretic Containment in Modern Enterprise Attack
    USarXiv cs.CR (Security)

    arXiv:2609.36485v1 Announce Type: cross Abstract: Modern Security Operations Centers struggle with delayed manual incident response, enabling adversaries to advance through the Cyber Kill Chain during early stage reconnaissance. While…

  110. Classical Verification of Quantum Computation with Quasilinear Resources, from Compiled Nonlocal Games
    USarXiv cs.CR (Security)

    arXiv:2609.38060v1 Announce Type: cross Abstract: Computational self-testing gives a classical verifier command over the quantum register of a single computationally bounded prover. We use this framework to construct the first argument…

  111. Adversarial Defense in Cybersecurity: A Systematic Review of GANs for Threat Detection and Mitigation
    USarXiv cs.CR (Security)

    arXiv:2509.20411v3 Announce Type: replace Abstract: Machine learning-based cybersecurity systems are highly vulnerable to adversarial attacks, while Generative Adversarial Networks (GANs) act as both powerful attack enablers and promising…

  112. Boosting Adversarial Robustness and Generalization with Dictionary Structure
    USarXiv cs.CR (Security)

    arXiv:2502.00834v2 Announce Type: replace-cross Abstract: This work investigates a novel approach to boost adversarial robustness and generalization by incorporating structural prior into the design of deep learning models. Specifically…

  113. A Function-level Dataset of Vulnerable and Fixed Source Code in JavaScript and TypeScript
    USarXiv cs.CR (Security)

    arXiv:2609.38012v1 Announce Type: new Abstract: JavaScript and TypeScript are widely used in modern web development, making their security critical; however, automated vulnerability detection is often constrained by the availability of…

  114. BadRAG: Identifying Vulnerabilities in Retrieval Augmented Generation of Large Language Models
    USarXiv cs.CR (Security)

    arXiv:2406.00083v3 Announce Type: replace Abstract: Retrieval-Augmented Generation (RAG) enhances Large Language Models (LLMs) by retrieving relevant information from external knowledge bases to provide more accurate, contextually…

  115. TEE Anchor: Cross-TEE Organizational Endorsement for Mitigating TEE Physical Attacks
    USarXiv cs.CR (Security)

    arXiv:2609.35919v1 Announce Type: new Abstract: In 2025, practical physical-access attacks against TEEs, such as TEE.fail and Battering RAM, were disclosed, posing a serious threat to current TEEs. The more strictly the target machine is…

  116. Quantization Enables Private Dense Retrieval against Malicious Service Providers
    USarXiv cs.CR (Security)

    arXiv:2609.36376v1 Announce Type: new Abstract: Dense retrieval, the key component of Retrieval Augmented Generation (RAG), retrieves the most relevant documents by comparing dense vector representations of queries and passages from a…

  117. Practical Secrets Extraction against Black-box LLMs
    USarXiv cs.CR (Security)

    arXiv:2609.36941v1 Announce Type: new Abstract: Large language models (LLMs) increasingly power autonomous coding agents such as Codex and Claude Code, yet their training corpora may contain confidential credentials exposed in public…

  118. OPFL: Optimistic Verification of Federated Learning via Empirical Boundary
    USarXiv cs.CR (Security)

    arXiv:2609.37011v1 Announce Type: new Abstract: Federated learning enables multiple clients to collaboratively train models without sharing their private data. However, the lack of visibility into local training makes it difficult to…

  119. Confidence-Guided Protocol IR for LLM-Aided Security Protocol Modeling
    USarXiv cs.CR (Security)

    arXiv:2609.37396v1 Announce Type: new Abstract: Large language models offer a promising interface for translating natural-language protocol descriptions into formal security models, but their outputs remain difficult to trust without…

  120. Concealing LLM-Based Multi-Agent Topology via Phantom Structure Injection
    USarXiv cs.CR (Security)

    arXiv:2609.37567v1 Announce Type: new Abstract: Driven by the rapid advancement of large language models (LLMs), LLM-based multi-agent systems (MAS) have emerged as a powerful paradigm for collaborative reasoning over complex tasks. A key…

  121. When Tools Silently Lie: Evaluating and Mitigating Blind Compliance in Tool-Augmented Data Agents
    USarXiv cs.CR (Security)

    arXiv:2609.37153v1 Announce Type: cross Abstract: Tool-augmented data agents rely on tool outputs for analytical decisions. Yet successful execution can return plausible but incorrect evidence, requiring agents to decide whether to trust…

  122. An Operator-Norm Approach to Security with Quantum Advice
    USarXiv cs.CR (Security)

    arXiv:2609.37467v1 Announce Type: cross Abstract: Non-uniform security allows an adversary to receive bounded advice about an oracle before attempting a fresh challenge. This captures the most realistic attacks and has already been studied…

  123. OVIG: Optimistic Verification of AI Training Integrity via Gradient Signals
    USarXiv cs.CR (Security)

    arXiv:2606.21045v2 Announce Type: replace Abstract: The rapid growth of AI has increased the demand for domain-specific models. Post-training of open-source models offers a more economical way to meet this growing demand, but the cost of…

  124. CORE-BREW: LLR-Based Soft Decoding for Robust Multi-Bit LLM Watermarking
    USarXiv cs.CR (Security)

    arXiv:2606.24163v2 Announce Type: replace Abstract: Reliable provenance for LLM outputs requires multi-bit watermarks that remain robust under editing while maintaining low false-positive rates. Existing ECC-based LLM watermarks rely on…

  125. How to Tame a Multi-Headed Hydra? Adaptive Multi-Category Safety Steering for Large Language Models
    USarXiv cs.CR (Security)

    arXiv:2609.34514v2 Announce Type: replace Abstract: As large language models (LLMs) become increasingly widespread, preventing unsafe responses to harmful prompts is essential for their safe deployment. Activation steering offers an…

  126. Climbing the Hill: Prompt Injection Red-Teaming Against Frontier Models with Curriculum Reinforcement Learning
    USarXiv cs.CR (Security)

    arXiv:2609.33628v2 Announce Type: replace-cross Abstract: Prompt injection is a leading security risk for LLMs and LLM-based applications such as agents. State-of-the-art red-teaming methods for prompt injection leverage reinforcement…

  127. Audience-Bound Persistent Memory: Authorization Across the Memory Lifecycle
    USarXiv cs.CR (Security)

    arXiv:2609.36373v1 Announce Type: new Abstract: A personal language agent that acts for its owner across private and shared conversations can learn a fact from one audience and later place it in the context it assembles for another. We…

  128. Trojan Hippo Bench: A Dynamic Benchmark for Persistent Memory Attacks and Defenses in LLM Agents
    USarXiv cs.CR (Security)

    arXiv:2605.01970v4 Announce Type: replace Abstract: Memory systems enable otherwise stateless LLM agents to persist user information across sessions, but also introduce a new attack surface. The Trojan Hippo attack is a class of persistent…

  129. When One Leak Pays Forever: Context Binding and the Price of Deterring Collusion
    USarXiv cs.CR (Security)

    arXiv:2609.36667v1 Announce Type: cross Abstract: A coalition that deviates once can profit many times when what it sells keeps working. In a threshold-encrypted mempool, a leading defense against maximal extractable value (MEV), a quorum…

  130. Evaluating System One Models for Agent Security Decisions: Reliability, Calibration, and Selective Automation
    USarXiv cs.CR (Security)

    arXiv:2609.33401v2 Announce Type: replace Abstract: Model-based judges support agent security by detecting prompt injections, assessing interaction risks, and screening harmful requests. System One models select from predefined answers and…

  131. Meta-SecAlign: Training LLMs against Prompt Injection for Robust Agents
    USarXiv cs.CR (Security)

    arXiv:2507.02735v4 Announce Type: replace Abstract: Prompt injection attacks, where untrusted data contains an injected prompt to manipulate the system, have been listed as the top security threat to AI agents. By fine-tuning on simulated…

  132. Dual lattice attacks for bounded distance decoding, revisited
    USarXiv cs.CR (Security)

    arXiv:2609.37483v1 Announce Type: new Abstract: Analyses of dual lattice attacks have often assumed that the individual scores associated with short dual vectors are mutually independent. Laarhoven-Walter used this heuristic to derive…

  133. Succinct Arguments for QMA in the Quantum Random Oracle Model
    USarXiv cs.CR (Security)

    arXiv:2609.37429v1 Announce Type: cross Abstract: Succinct arguments are a fundamental cryptographic primitive for verifying computational claims with small communication. In the classical setting, succinct arguments for NP can be…

  134. PoisonCap: Efficient Hierarchical Temporal Safety for CHERI
    USarXiv cs.CR (Security)

    arXiv:2605.13210v3 Announce Type: replace-cross Abstract: In this paper, we present PoisonCap: scalable temporal safety with strict use-after-free protection and initialisation safety for CHERI systems. Efficient memory safety is an…

  135. Bayesian-Guided Cooperative RL Beamforming for Wireless Adversarial User Detection
    USarXiv cs.CR (Security)

    arXiv:2607.25417v2 Announce Type: replace-cross Abstract: In next-generation wireless networks, communication systems are expected to go beyond simple data transmission and simultaneously provide high data rates, efficiency, and security…

  136. Safer Content or Firmer Refusals? A Hybrid Perturbation Defense for Alignment under Harmful Fine-tuning
    USarXiv cs.CR (Security)

    arXiv:2609.36862v1 Announce Type: new Abstract: Fine-tuning-as-a-service lets users adapt a safety-aligned language model to their own data, but it also creates a harmful fine-tuning attack surface: a small amount of harmful data mixed…

  137. One Pipeline Does Not Fit All: TAILOR, a Type- and State-Aware Framework for CVE Reproduction
    USarXiv cs.CR (Security)

    arXiv:2609.37006v1 Announce Type: new Abstract: Growing vulnerability disclosure and widespread software reuse increase security teams' need for reproducible evidence to diagnose vulnerabilities, validate patches, and build regression…

  138. ISC Stormcast For Wednesday, September 30th, 2026 https://isc.sans.edu/podcastdetail/10116, (Wed, Sep 30th)
    USSANS ISC
  139. Trump Says Top Tech Firms Have Signed Accord to ‘Self-Police’ AI Development
    USSecurityWeek

    The accord opened the door to future regulation but focused on four voluntary steps for the companies to take. The post Trump Says Top Tech Firms Have Signed Accord to ‘Self-Police’ AI Development appeared first on SecurityWeek .

  140. Microsoft is rolling out Linux container support to WSL
    USBleepingComputer

    Microsoft is taking Windows Subsystem for Linux beyond just running Linux distributions, as WSL Containers is now generally available. [...]

  141. Recorded Future Debuts Autonomous Defense, Built for Machine-Speed Threats
    USRecorded Future

    Rethinking defense entirely We know attackers are already using AI to operate at machine speed, running more frequent and more effective attacks. They're automating reconnaissance, standing up infrastructure, and pivoting faster than human…