AX·NEWS
Sign in

Cybersecurity

Cybersecurity news — Tuesday 27 October 2026

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

Open in the live feedRSS feedOn a TV

  1. Japan's GMO urges customers to shut down PCs amid ransomware threat
    USBleepingComputer

    The GMO Group has urged customers to disconnect their systems from the internet and shut down computers after a threat actor who breached the company's Onamae Desktop Cloud threatened to launch a ransomware attack today. [...]

  2. OpenAI's GPT-6.1 Terra is safer overall, but shows more destructive-action flags in tests
    USBleepingComputer

    OpenAI confirmed GPT-6.1 Terra is safer overall than some of its older models, but its safety testing also found other problems. [...]

  3. UNM Health ransomware attack disrupts patient care, surgeries
    USBleepingComputer

    The University of New Mexico Health Sciences Center disclosed a cyberattack that caused system outages and disrupted patient care and academic activities. [...]

  4. Before you file that AI bug report, OpenAI has some advice
    USCyberScoop

    Even OpenAI is admitting there is a lot of AI slop in the security space. An executive for the frontier AI company said Tuesday that AI -generated vulnerability reports are overwhelming open-source maintainers, leading some to stop…

  5. Hackers hijack Convertr analytics script to steal payment card data
    USBleepingComputer

    A credit card skimming group reportedly compromised a shared analytics script used by e-commerce platform Convertr to steal shoppers' payment data during checkout. [...]

  6. All the security risks Anthropic is worried about before its IPO 
    USCyberScoop

    Anthropic said its AI systems could drastically reshape the software and cybersecurity sectors in the “near term,” while also laying out a litany of cybersecurity and regulatory risks facing the company as it prepares to go public. In a…

  7. Why this new Fitbit feature takes us one step closer to non-invasive glucose tracking
    USZDNet Security

    Insulin resistance and blood pressure trends are now available on two new Fitbit devices.

  8. Prolific Hive and Nefilim ransomware affiliate sentenced to 70 months in prison
    USCyberScoop

    A 36-year-old Ukrainian national was sentenced to 70 months in prison for participating in a string of ransomware attacks targeting various critical infrastructure operators and businesses from the middle of 2018 to late 2021, the Justice…

  9. CyberStrikeAI Uses 44 AI Agents to Automate Hacking Operations Against Japan Targets
    INThe Hacker News

    Cybersecurity researchers have disclosed details of three cyber attacks targeting organizations in Japan in which artificial intelligence (AI) was a key element, either in the driving seat or to assist a human operator. TrendAI, which…

  10. Microsoft releases Windows OOB updates to distrust eMudhra certificates
    USBleepingComputer

    Microsoft has released out-of-band (OOB) Windows updates to remove trust in certificates tied to compromised eMudhra certificate authority keys, protecting systems that do not receive automatic certificate trust updates. [...]

  11. Netflix now, Hulu later? T-Mobile’s T-Mix plans let you switch perks – what you need to know
    USZDNet Security

    T-Mobile’s new T-Mix plans deliver tons of extras ranging from streaming to mobile discounts, but there are some caveats to know.

  12. Windows finally gets a built-in internet speed test tool – how to try it
    USZDNet Security

    Available to Windows Insiders, the built-in tool measures your upload and download speeds over Ethernet or Wi-Fi.

  13. This tool can recover deleted AI chats from your phone – how it works
    USZDNet Security

    This new tool proves it’s still best to limit the info you share with chatbots.

  14. Warp discloses breach exposing AI agent logs, user telemetry data
    USBleepingComputer

    Agent decision logs and telemetry data belonging to Warp agentic development environment users were exposed in an attack on a third-party vendor in the project's infrastructure. [...]

  15. The new iPad Mini’s best upgrade has nothing to do with its flashy OLED screen
    USZDNet Security

    Apple’s newest Mini is the first iPad ever with an IP67 rating for dust and water resistance.

  16. Apple’s M6 iMac has arrived – here’s what’s changed since the M4
    USZDNet Security

    Apple’s colorful all-in-one gets its first refresh in two years. Here’s what you get for the $1,499 starting price.

  17. Five election scams to watch out for
    USMalwarebytes Labs

    Elections give scammers plenty to work with: urgent deadlines, familiar faces, passionate supporters, and messages asking people to take action. Some want you to “fix” your voter registration. Others want a donation, an answer to a survey…

  18. A fake TikTok shop is Surfshark’s most-blocked phishing scam – how to steer clear
    USZDNet Security

    Blocked phishing attempts rose 83% in six months, according to Surfshark, largely due to one TikTok shop. Here’s how to spot the red flags.

  19. Tailscale Rushes Patches After Uncoordinated Disclosure of ‘Tailworm’ Flaws
    USSecurityWeek

    Tailscale is preparing a patch for Tailscale SSH flaws whose details went public over the weekend, before any fixes had been released. The post Tailscale Rushes Patches After Uncoordinated Disclosure of ‘Tailworm’ Flaws appeared first on…

  20. National cyber director rolls out latest pilot in South Carolina, this time for rural hospitals
    USCyberScoop

    The Office of the National Cyber Director has rolled out its second regional cyber pilot program, this one focused on rural hospitals. National Cyber Director Sean Cairncross visited South Carolina on Monday to debut the six-month program…

  21. Tailscale warns of SSH flaws allowing root privilege escalation
    USBleepingComputer

    The Tailscale security team is working to patch almost a dozen security vulnerabilities that could be exploited to steal credentials, crash servers, gain root privileges, and more. [...]

  22. The Huntress Tragic Quadrant: Ranking Threats Hitting Businesses Hardest
    USBleepingComputer

    Attackers are increasingly abusing legitimate IT tools, Microsoft 365 sessions, and everyday workflows to compromise businesses without triggering traditional defenses. Huntress Labs ranks the threats hitting organizations hardest…

  23. GoAhead TLS Flaws Allow Revoked Client Certificates and Enable Session Decryption
    INThe Hacker News

    The GoAhead embedded web server loaded a certificate revocation list into its default TLS backend but never turned revocation checking on, so a revoked client certificate could still pass login on affected setups. The flaw is in builds…

  24. Apple Backports Fix for Exploited CoreGraphics Flaw to Older iPhones and iPads
    INThe Hacker News

    Apple on Monday backported fixes for a high-severity security flaw that was exploited in real-world cyber attacks to older devices. The vulnerability, tracked as CVE-2026-86950 (CVSS score: 8.8), is an out-of-bounds write affecting…

  25. Vendor Key Leak Puts Data of 238,000 Denver Health Plan Members at Risk
    USSecurityWeek

    A leaked SquareML access key could have been used to access a database containing personal and health information. The post Vendor Key Leak Puts Data of 238,000 Denver Health Plan Members at Risk appeared first on SecurityWeek .

  26. HPE Patches Dozens of Critical Flaws in Networking Products
    USSecurityWeek

    Eight max-severity security defects could lead to remote code execution and authentication bypass. The post HPE Patches Dozens of Critical Flaws in Networking Products appeared first on SecurityWeek .

  27. Artificial Intimacy: Love and lies in the age of AI 
    USMalwarebytes Labs

    AI is becoming a third wheel in relationships, increasingly playing a role in the most intimate parts of our lives. It shows up in the flirty first text, the polished dating profile, the therapized language we use with a partner after a…

  28. The web’s having a rough week. Here’s how to browse safely
    USMalwarebytes Labs

    You recognize the website. You’ve visited it before. You may even have it bookmarked. Then it asks you to do something unusual: install an update, allow notifications, or run a command to prove you’re human. Stop and check first. A website…

  29. I tried Windows 11’s customizable right-click menu – and it’s a much-needed change
    USZDNet Security

    The latest Windows Insider build lets you add, remove, and manage commands in the right-click menu so it’s less crowded and confusing.

  30. Microsoft Authenticator Flaw Let Permissionless Android Apps Steal Work Account Tokens
    INThe Hacker News

    Microsoft has fixed a flaw in its Authenticator app for Android that let a malicious app with no permissions pose as a trusted Microsoft app, such as Teams, and collect sign-in tokens for a victim's work account. Microsoft tracks the flaw…

  31. Unexplained DreamHost Incident Prompts Email, FTP Password Resets
    USSecurityWeek

    The company says it reacted to an issue affecting some of its systems, including DreamHost webmail. The post Unexplained DreamHost Incident Prompts Email, FTP Password Resets appeared first on SecurityWeek .

  32. From industry concern to practical action: the value of working with regulators
    GBNCSC UK

    How industry, the NCSC and the FCA worked together to help organisations share information while protecting market integrity

  33. iOS, macOS 27.2 Patch 286 Bugs; Fix for Exploited Flaw Backported to iOS 18
    USSecurityWeek

    Apple patched 286 vulnerabilities in iOS and macOS 27.2 and brought the fix for an exploited CoreGraphics zero-day to older iPhones. The post iOS, macOS 27.2 Patch 286 Bugs; Fix for Exploited Flaw Backported to iOS 18 appeared first on…

  34. Keystone to Destroy 18,835 Hardware Wallets After Hackers Steal Device Verification Key
    INThe Hacker News

    An attacker who gained access to Keystone's website obtained the key that its hardware wallets use to prove they are genuine, the company said in a statement on October 26. Someone with that key could build counterfeit devices that pass…

  35. AI-Generated Fake Login Pages Are Making Account Takeovers Easier
    INThe Hacker News

    Fake login pages are nothing new, but generative AI is making them easier to build and more convincing. Attackers can recreate familiar login screens, impersonate trusted brands and tailor phishing attempts to specific employees or…

  36. Lazarus Abuses South Korean Security Software to Deploy Malware Using Post-Quantum Encryption
    INThe Hacker News

    The North Korea-linked Lazarus Group has been attributed to a cyber espionage campaign dubbed Operation FlashHole targeting at least three organizations in South Korea spanning the construction, software, and wire-communication equipment…

  37. Hiding Capabilities in AI Models
    USSchneier on Security

    Interesting research: “ AI Models Can Provably Hide Arbitrary Capabilities ” Abstract: Capability evaluations aim to surface dangerous model behaviors before deployment, but their reliability depends on the assumption that hidden…

  38. Samsung TVs without mics may listen through soundbars, researcher says
    USMalwarebytes Labs

    A reverse engineer claims that their Samsung smart TV has been collecting audio without their consent. The researcher, who calls themselves “log4j,” said in an X thread that their 2025 Samsung TV was delivering ads that seemed suspiciously…

  39. Risks Spelled Out for Anthropic Investors Include Possible Harms From Runaway AI
    USSecurityWeek

    Anthropic presents its own initial public offering as a step toward mitigating the risks it believes AI poses. The post Risks Spelled Out for Anthropic Investors Include Possible Harms From Runaway AI appeared first on SecurityWeek .

  40. Apple backports CoreGraphics zero-day patches to older iPhones
    USBleepingComputer

    ​Apple has backported security patches released last month to older iPhones and iPads, addressing a zero-day bug exploited in "extremely sophisticated" targeted attacks on iOS devices. [...]

  41. This is Fine: What NetScaler Zero-Days Tell Us About the New Normal
    USSecurityWeek

    Keep your patch management pipelines primed in case another round of Citrix NetScaler patches get released soon. The post This is Fine: What NetScaler Zero-Days Tell Us About the New Normal appeared first on SecurityWeek .

  42. What happens when security tools go dark before the ransomware
    USCyberScoop

    On Oct. 1, threat researchers reported that the China-linked group behind Warlock ransomware, which Microsoft tracks as Storm-2603, had attacked at least four organizations in the prior two months. A water utility and a telecommunications…

  43. Dead Popcorn Time Updater Put Millions of Computers at Risk of Takeover
    USSecurityWeek

    Running at system boot, it periodically polled the update domain to execute any code served, with full privileges. The post Dead Popcorn Time Updater Put Millions of Computers at Risk of Takeover appeared first on SecurityWeek .

  44. Australia Gets Into Amadey Malware Backend, Cuts Off Access to 180,000 Infected Computers
    INThe Hacker News

    A criminal using the Amadey malware can no longer access or steal data from 180,000 infected computers around the world, including some in Australia, the Australian Signals Directorate (ASD) says. The agency's operators gained access to…

  45. Energy giant ENGIE discloses data breach affecting customers
    USBleepingComputer

    French energy giant ENGIE, one of the world's largest power and natural gas providers, confirmed that attackers stole customer data after breaching a marketing platform used by business partners. [...]

  46. OpenAI Compaction Flaw Let Researchers Bypass Hidden Reasoning Protections
    INThe Hacker News

    OpenAI's compaction feature, which shrinks a long conversation to save space, could be turned into a way to copy a model's hidden reasoning out in plain text, security firm Xint has shown. OpenAI's reasoning models work through a problem…

  47. CISA confirms PHP2Shell vulnerabilities now exploited in attacks
    USBleepingComputer

    The U.S. Cybersecurity and Infrastructure Security Agency (CISA) has added the PHP2Shell vulnerabilities to its catalog of actively exploited security flaws and ordered federal agencies to patch their systems by Thursday. [...]

  48. Maritime Coalition Launches Free Vulnerability Database for Ship and Port Systems
    USSecurityWeek

    Blue Initiative indexes nearly 30,000 IT and OT vulnerabilities relevant to ships and ports, 37 of them in maritime-specific products. The post Maritime Coalition Launches Free Vulnerability Database for Ship and Port Systems appeared…

  49. Attackers Hijack Dahua IP Cameras in Ukraine Using P2P Access and Auth Weaknesses
    INThe Hacker News

    The Computer Emergency Response Team of Ukraine (CERT-UA) has issued an alert warning of unauthorized remote access to video surveillance systems manufactured by Dahua. "Cases of compromising IP cameras and DVRs with access to the video…

  50. CISA Confirms Active Exploitation of Two PHP2Shell Flaws Enabling Remote Code Execution
    INThe Hacker News

    The U.S. Cybersecurity and Infrastructure Security Agency (CISA) on Monday confirmed active exploitation of the recently patched PHP2Shell flaws by adding them to its Known Exploited Vulnerabilities (KEV) catalog. The PHP vulnerabilities…

  51. I-HITL: Inalienable Human-in-the-Loop
    USarXiv cs.CR (Security)

    arXiv:2610.35835v1 Announce Type: new Abstract: Human oversight is a pillar of AI safety. It is often exercised by a human in the loop (HITL) who approves critical actions taken by an AI agent. AI agents' increasing ability to bypass…

  52. Benchmarking Classical and Quantum Machine Learning for Intrusion Detection Across Multiple Datasets
    USarXiv cs.CR (Security)

    arXiv:2610.33119v1 Announce Type: new Abstract: Intrusion detection is examined through a comparative benchmark of classical and quantum machine learning models on three datasets: NSL-KDD, UNSW-NB15, and MQTTEEB-D2025. The analysis focuses…

  53. HNSecW: A Logarithmic-Time MPC Protocol for Approximate Nearest Neighbor Search
    USarXiv cs.CR (Security)

    arXiv:2610.34248v1 Announce Type: new Abstract: As a core component of modern RAG systems, approximate nearest neighbor (ANN) search has been extensively studied for decades. In practice, deploying ANN often requires symmetric privacy…

  54. WADBERT: Dual-Channel Web Attack Detection via Permutation-Invariant Parameter Modeling
    USarXiv cs.CR (Security)

    arXiv:2601.21893v2 Announce Type: replace Abstract: Web attack detection is the first line of defense for securing web applications, designed to preemptively identify malicious activities. Deep learning-based approaches are increasingly…

  55. CVM-Store: A Secure Storage System for Confidential Virtual Machines
    USarXiv cs.CR (Security)

    arXiv:2610.34560v1 Announce Type: new Abstract: Stateful cloud workloads require high-performance persistent storage when running on untrusted cloud infrastructure. Confidential Virtual Machines (CVMs) promise to protect these workloads…

  56. Scope-Gap Injection: Exploiting Evaluator Channel Blind Spots in the OpenClaw Agent Security Competition
    USarXiv cs.CR (Security)

    arXiv:2610.34777v1 Announce Type: cross Abstract: This paper introduces Scope-Gap Injection (SGI), an indirect prompt injection method for multi-stage LLM-agent safety pipelines. In the IJCAI 2026 OpenClaw Agent Security Attack…

  57. A Cross-Modal Prompt Injection Attack against Large Vision-Language Models with Image-Only Perturbation
    USarXiv cs.CR (Security)

    arXiv:2605.16090v2 Announce Type: replace Abstract: Large vision-language models (LVLMs) have emerged as a powerful paradigm for multimodal intelligence, but their growing deployment also expands the attack surface of prompt injection…

  58. Can Open Be Too Open? On the Security and Trustworthiness of Fully Transparent LLMs
    USarXiv cs.CR (Security)

    arXiv:2610.35199v1 Announce Type: new Abstract: There are different levels of transparency for LLMs: open weights, open data, and/or open pipeline. Recently, more interest has emerged in fully transparent LLMs, i.e., LLMs with fully open…

  59. Deletion, Substitution, Repair and Transplant: Component Attribution in BARBICAN, a Layered Prompt-Injection Gateway for MCP
    USarXiv cs.CR (Security)

    arXiv:2610.33198v1 Announce Type: new Abstract: A neural classifier can fail completely inside a layered defense without any end-to-end number moving. We demonstrate this on BARBICAN, a fail-closed gateway for language-model tool mediation…

  60. GuardRAG: A Post-Retrieval Defense against Poisoning Attacks in Retrieval-Augmented Generation
    USarXiv cs.CR (Security)

    arXiv:2610.35071v1 Announce Type: new Abstract: Retrieval-Augmented Generation (RAG) enhances the factual grounding of Large Language Models (LLMs) by incorporating external evidence, but also exposes them to corpus poisoning attacks. In…

  61. No Provenance Left Behind: Revisiting Reference Monitor Guarantees with Learned Kernel Scheduling
    USarXiv cs.CR (Security)

    arXiv:2510.08479v3 Announce Type: replace Abstract: Provenance plays a critical role in maintaining traceability of a system's actions for root cause analysis of security threats and impacts. Provenance collection is often incorporated…

  62. Second order Recurrences, quadratic number fields and cyclic codes
    USarXiv cs.CR (Security)

    arXiv:2603.25343v2 Announce Type: replace-cross Abstract: Wall-Sun-Sun primes (shortly WSS primes) are defined as those primes $p$ such that the period of the Fibonacci recurrence is the same modulo $p$ and modulo $p^2.$ This concept has…

  63. Pacifo: An Efficient Fully Homomorphic Encryption-based Lightweight ResNet for Trustworthy Machine Learning as a Service
    USarXiv cs.CR (Security)

    arXiv:2610.33525v1 Announce Type: new Abstract: Machine Learning as a Service (MLaaS) has become a popular paradigm for deploying AI applications. However, it raises significant privacy concerns that the raw client data needs to be…

  64. Denial of Service Attacks on Beam Allocation in Large LEO Constellations
    USarXiv cs.CR (Security)

    arXiv:2610.34631v1 Announce Type: cross Abstract: We study the vulnerability of beam allocation approaches in large low-Earth-orbit (LEO) satellite constellations to adversarial demand inflation resulting in denial-of-service attacks. We…

  65. From Speech Tokens to Stolen Voices: Attacker-Chosen-Text Speaker Impersonation via Representation Inversion
    USarXiv cs.CR (Security)

    arXiv:2610.34688v1 Announce Type: cross Abstract: Speech interaction systems have gradually evolved from traditional cascaded systems that use text as intermediate information to end-to-end speech language models that directly process…

  66. Trustworthy Recursive Self-Improvement in the Era of Large Models: Foundation, Risks, Evaluation and Open Challenges
    USarXiv cs.CR (Security)

    arXiv:2610.35193v1 Announce Type: cross Abstract: AI systems built around large models increasingly use experience and feedback to make persistent changes to models, prompts, memories, tools, and workflows. Recursive self-improvement (RSI)…

  67. Trim: Fee Level Matters More Than Skew in Lightning Routing
    USarXiv cs.CR (Security)

    arXiv:2610.35296v1 Announce Type: cross Abstract: Lightning routing nodes lend their channel balance for a posted fee. Operators' tools move that fee with the balance around a hand-set level, and which part of the fee drives revenue is…

  68. ProbeID: Verifying MCP Tool Identity for LLM Agents under Registry Drift
    USarXiv cs.CR (Security)

    arXiv:2610.33633v1 Announce Type: new Abstract: Agents select tools by copyable descriptions, which cannot establish which implementation runs. ProbeID compares canonicalized responses to schema-valid probes with a trusted admission…

  69. Spoofing the Toolbox: Optimizing Tool Descriptions to Deceive LLM Agents
    USarXiv cs.CR (Security)

    arXiv:2610.33182v1 Announce Type: new Abstract: Large language model (LLM) agents are widely used in many applications and often rely on external tools. Tool information, such as function descriptions, helps agents determine which tools…

  70. SentryBench: A Cost-Sensitive Benchmark for Adversarial Robustness of Tool-Using LLM Agents in Public-Safety Deployments
    USarXiv cs.CR (Security)

    arXiv:2610.34535v1 Announce Type: new Abstract: Large language model (LLM) agents are increasingly wired to consequential tools that send public alerts, dispatch responders, and query citizen records, where one erroneous tool call carries…

  71. The Target Agent Is Its Own Best Critic but a Poor Proposer: Self-Improved Red Teaming
    USarXiv cs.CR (Security)

    arXiv:2610.34965v1 Announce Type: new Abstract: Automatic red teaming increasingly relies on large language models (LLMs) to iteratively diagnose failed attacks and propose improved adversarial prompts or injections. Although prior work…

  72. Extraction meets Distillation - cryptanalytic techniques to recover parameters from a Deep Neural Network
    USarXiv cs.CR (Security)

    arXiv:2610.33671v1 Announce Type: cross Abstract: We study fully connected neural-network reconstruction by combining cryptanalytic weight-direction extraction with learning from predicted labels. We adapt a ReLU extraction implementation…

  73. From Hype to ROI: An Economic Model for LLM Adoption in Information Systems and Security Operations
    USarXiv cs.CR (Security)

    arXiv:2610.33352v1 Announce Type: new Abstract: Large Language Models (LLMs) promise significant efficiency gains across information-intensive operations by automating repetitive language-based tasks. However, adoption decisions remain…

  74. Alignment of Autonomous Cyber Defense Agents via Adaptive Policy Shaping
    USarXiv cs.CR (Security)

    arXiv:2610.33234v1 Announce Type: new Abstract: Cybersecurity operators increasingly encounter sophisticated and fast-paced attacks, making autonomous cyber defense (ACD) a critical approach for supporting timely and effective…

  75. CVE-to-ATT＆CK Mapping via Evidence-Constrained Ranking
    USarXiv cs.CR (Security)

    arXiv:2610.33924v1 Announce Type: new Abstract: Mapping CVEs to MITRE ATT&CK techniques helps defenders translate vulnerability information into hypotheses about adversary behavior. Automating this mapping is challenging because relevant…

  76. Phantom Forgetting: How BatchNorm State Fakes Deletion in Edge Split Machine Unlearning
    USarXiv cs.CR (Security)

    arXiv:2610.34800v1 Announce Type: new Abstract: Split learning is a natural fit for edge intelligence: a resource-constrained edge device runs the first layers of a network and offloads the remainder to a server, so that raw data never…

  77. Cheap AI-Assisted Reverse Engineering and What It Will Do
    USarXiv cs.CR (Security)

    arXiv:2610.35633v1 Announce Type: cross Abstract: Generative AI is putting the law of reverse engineering (RE) under severe strain. Existing IP and technology laws assume that RE is expensive and time-consuming, and that it requires…

  78. Performance Guarantees for Permissionless Lending
    USarXiv cs.CR (Security)

    arXiv:2610.34774v1 Announce Type: cross Abstract: We propose a lending protocol with far fewer risks than existing pool-based lending protocols (e.g., Aave, Compound). In our protocol, lenders and borrowers embody the spirit of…

  79. Security evaluation of 6G-oriented and O-RAN networks through targeted attacks
    USarXiv cs.CR (Security)

    arXiv:2610.33526v1 Announce Type: new Abstract: The sixth generation (6G) of networks is supported by increased bandwidth, automation through the use of O-RAN technologies. Simultaneously though autonomous operation leads to an extensive…

  80. Reva: Real-Time Vulnerability Detection through Contextual Probing of Large Language Models
    USarXiv cs.CR (Security)

    arXiv:2610.35829v1 Announce Type: new Abstract: A vulnerability costs more to fix the later it is discovered. As coding agents generate code faster than developers can review it, software needs a first line of defense screening each new…

  81. Residual Transferability in Neural Image Watermarking
    USarXiv cs.CR (Security)

    arXiv:2609.32241v2 Announce Type: replace Abstract: Neural image watermarks can be forged by extracting watermark-bearing residuals from released images and transferring them to unrelated content. While prior work has demonstrated this…

  82. Online Shift Detection and Conformal Adaptation for Deployed Safety Classifiers
    USarXiv cs.CR (Security)

    arXiv:2606.11949v5 Announce Type: replace-cross Abstract: We study three failure modes of deployed safety classifiers: distribution shift, targeted evasion, and invalid scores from reasoning-model monitors. A reduced 800-cell subset of a…

  83. Holo-NN: Holographic Nearest Neighbor Search via Compressive Sensing
    USarXiv cs.CR (Security)

    arXiv:2610.33224v1 Announce Type: new Abstract: In the age of AI, privacy-preserving approximate nearest neighbor (ANN) search over high-dimensional embeddings has emerged as a critical yet frustratingly challenging domain, dominated by…

  84. Adversary Brains: Hypothesis-Based Logical Signatures for Threat Actor Detection with LLMs
    USarXiv cs.CR (Security)

    arXiv:2610.35624v1 Announce Type: new Abstract: An adversary brain is an LLM-based model of a threat actor's behavior, held as an evolving logical signature: a hypothesis with enough structure that data can test it continuously, without a…

  85. Machine-Checked Soundness of the GKR Protocol for Sparse-Merkle Updates
    USarXiv cs.CR (Security)

    arXiv:2610.05335v2 Announce Type: replace Abstract: Proofs of authenticated-data-structure updates built on the Goldwasser-Kalai-Rothblum (GKR) protocol are sound only if three arguments hold together: the soundness of the protocol, the…

  86. Attack logics, not outputs: Concept-based Property Attacks for Shortcut Elimination in Neural Networks
    USarXiv cs.CR (Security)

    arXiv:2610.34923v1 Announce Type: new Abstract: Deep neural networks (NNs) for computer vision are vulnerable to adversarial attacks, such that minuscule malicious changes to inputs may induce unintuitive outputs. One key approach to…

  87. Factual Yet One-Sided: Stance Poisoning and Evidence Balancing in Retrieval-Augmented Generation
    USarXiv cs.CR (Security)

    arXiv:2610.34294v1 Announce Type: cross Abstract: The stance expressed in responses from retrieval-augmented generation systems is significantly influenced by the stance distribution of the retrieved materials. Existing research on…

  88. A Survey on Long-Term Memory Security in LLM Agents: Attacks, Defenses, and Governance Across the Memory Lifecycle
    USarXiv cs.CR (Security)

    arXiv:2604.16548v4 Announce Type: replace Abstract: The emergence of writable, cross-session persistent memory in LLM agents introduces a qualitatively different threat landscape from conventional input-centric security concerns…

  89. Dummy Backdoor as a Defense: Removing Unknown Backdoors via Shared Internal Mechanisms for Generative LLMs
    USarXiv cs.CR (Security)

    arXiv:2606.11648v2 Announce Type: replace Abstract: Backdoor attacks pose a serious threat to the safety and reliability of Large Language Models (LLMs), as they cause models to behave normally on clean inputs while producing…

  90. TRACER: An RDC-Triggered Trajectory Recovery Framework for Cyberattack-Resilient Adaptive Cruise Control
    USarXiv cs.CR (Security)

    arXiv:2610.33052v1 Announce Type: new Abstract: The increasing connectivity of adaptive cruise control (ACC) systems exposes vehicle-state information to cyberattacks, potentially causing irrational car-following behavior and unsafe…

  91. BE-CTDE: Blockchain-Enabled Multi-Agent Reinforcement Learning with Anomaly Detection and Token Economics
    USarXiv cs.CR (Security)

    arXiv:2610.34121v1 Announce Type: cross Abstract: Multi-agent reinforcement learning supports cooperative decision-making in Internet of Things systems, but strategic deviations can disrupt coordination and undermine confidence in the…

  92. Compound Environmental Operating Stress and Cyber-Induced Pump Disruption in Water Distribution Networks: Matched Hydraulic Impacts, Service Continuity, and Sustained Recovery
    USarXiv cs.CR (Security)

    arXiv:2610.34670v1 Announce Type: new Abstract: Digitally controlled water distribution systems must maintain service during disruptions, yet restoring water delivery does not necessarily ensure hydraulic recovery. This study evaluates…

  93. A Category-Selective Encryption Framework for Privacy Protection of UAV Images
    USarXiv cs.CR (Security)

    arXiv:2610.35729v1 Announce Type: new Abstract: Imagery captured by unmanned aerial vehicle (UAV) systems increasingly raises privacy concerns, as UAV aerial surveillance routinely captures sensitive objects such as pedestrians and…

  94. StateSync-GKR: Design and Evaluation of a Reusable Sparse GKR Engine
    USarXiv cs.CR (Security)

    arXiv:2610.14849v2 Announce Type: replace Abstract: Authenticated state transitions need proof engines that conserve memory, avoid repeated setup, keep every result bound to its request under parallel execution and return complete proofs…

  95. Enhancing LLM-Based Vulnerability Detection via Representation-Aware Logical Ensemble
    USarXiv cs.CR (Security)

    arXiv:2610.33756v1 Announce Type: new Abstract: Automated vulnerability detection in C source code is a long-standing challenge, as traditional static analysis tools often trade precision for scalability and struggle with complex…

  96. Does the Voynich Manuscript Mean Anything? Calibrated Tests Find Little Trace of Language
    USarXiv cs.CR (Security)

    arXiv:2610.34499v1 Announce Type: cross Abstract: The Voynich manuscript is a fifteenth-century book in a script that no one can read. One account holds that its text means nothing and that a scribe made it by copying his own words and…

  97. A-GE: Privacy-Preserving Age Assurance via Fine-Grained Motion
    USarXiv cs.CR (Security)

    arXiv:2610.33206v1 Announce Type: new Abstract: Age assurance faces a fundamental privacy dilemma: meaningful assurance typically requires revealing identity or other sensitive information, whereas privacy-preserving alternatives often…

  98. A Certified Robustness Framework for Backdoor Poisoning Attacks using $f$-Divergences
    USarXiv cs.CR (Security)

    arXiv:2610.34934v1 Announce Type: cross Abstract: Machine learning systems trained on untrusted data are vulnerable to data and backdoor poisoning attacks, in which an adversary corrupts training and test samples to manipulate predictions…

  99. CoSA: The Conflict of Skills Attack in Real-World LLM Agent Ecosystems
    USarXiv cs.CR (Security)

    arXiv:2610.35082v1 Announce Type: cross Abstract: Agent Skills allow LLM agents to combine reusable capabilities, but checking each Skill in isolation cannot rule out policy violations when they are composed. We term such failures Conflict…

  100. Combinatorial Lock for Trojan-Resilient AI Engines via Hash-Assisted Spectral Filtering
    USarXiv cs.CR (Security)

    arXiv:2610.33437v1 Announce Type: new Abstract: Multi-trigger backdoor attacks, in which heterogeneous trigger mechanisms (e.g., patch, blend, warping, and the latent-domain Spy-Watermark) coexist within a single poisoned model, cannot be…

  101. Trust Mechanisms for Offline Generative AI at the Edge: A Connectivity-Regime Framework and an Empirical Study on a Raspberry Pi 5
    USarXiv cs.CR (Security)

    arXiv:2610.33732v1 Announce Type: new Abstract: Generative AI and large language models (LLMs) are moving into the Internet of Things, yet most deployments assume cloud connectivity, and many IoT settings have intermittent, restricted, or…

  102. Ring-0 by Default: Agents as Operating Systems (AaOS)
    USarXiv cs.CR (Security)

    arXiv:2610.33753v1 Announce Type: new Abstract: LLM agents now read and write files, start processes and subagents, fetch network content, load third-party extensions, and save and resume state, which are the responsibilities of an…

  103. LLM WikiLeaks: Uncovering URL-Based Capability Extension and Security Circumvention in Agentic Message Boards
    USarXiv cs.CR (Security)

    arXiv:2610.34542v1 Announce Type: new Abstract: Autonomous agents have been observed acting beyond their assigned tasks, accessing unauthorized systems, and establishing communication channels to exchange information, yet the Web and…

  104. Steering for Safety, Hijacking for Profit: Trojan Threats in the LLM Activation Steering Supply Chain
    USarXiv cs.CR (Security)

    arXiv:2610.35043v1 Announce Type: new Abstract: As large language models (LLMs) are increasingly integrated into Web search, online recommendation, and intelligent decision-making, the third-party plugin ecosystem centered on model…

  105. Agent-to-Agent Synthetic Social Engineering on the Web
    USarXiv cs.CR (Security)

    arXiv:2610.33411v1 Announce Type: new Abstract: An AI agent can prepare an accurate report and still request an action its user has not authorized. Cooperation across Web services therefore requires separate judgments about identity…

  106. Can a Teenager Fool an AI? Evaluating Low-Cost Cosmetic Attacks on Age Estimation Systems
    USarXiv cs.CR (Security)

    arXiv:2602.19539v2 Announce Type: replace-cross Abstract: Age estimation systems are increasingly deployed as gatekeepers for age-restricted online content, yet their robustness to cosmetic modifications has not been systematically…

  107. Hidden Backdoor Attack on Deep Contrastive Clustering
    USarXiv cs.CR (Security)

    arXiv:2610.33116v1 Announce Type: new Abstract: Deep contrastive clustering has become a representative paradigm for unsupervised image clustering, yet its security under training-time poisoning remains largely unexplored. This paper…

  108. Back to Basics: Formal Analysis Hints Improve Weak-to-Strong Monitoring
    USarXiv cs.CR (Security)

    arXiv:2610.34064v1 Announce Type: new Abstract: Frontier agentic systems are capable of taking misaligned actions that may cause significant harm. Monitoring techniques are developed to oversee these systems in order to detect and prevent…

  109. Maximum-Likelihood Decoding of Gabidulin Codes Is NP-Hard Even in Dimension One
    USarXiv cs.CR (Security)

    arXiv:2610.34404v1 Announce Type: new Abstract: Fix a prime power q. We prove that deciding whether a received word y in Fnqm lies at rank distance at most n-2 from the one-dimensional Gabidulin code {{\lambda.g : {\lambda} in Fqm } is…

  110. Building Automotive Security on Internet Standards: In-Car Service Authentication with DNSSEC, DANE, and DANCE
    USarXiv cs.CR (Security)

    arXiv:2506.13261v2 Announce Type: replace Abstract: The automotive industry trends to softwarizing cars according to flexible, service-oriented architectures. At the same time, connected cars become networked computer systems but lack…

  111. False-Alarm Improvements and Detection Trade-Offs in QML Intrusion Detection: A Secondary Analysis of the Kaissar Corpus
    USarXiv cs.CR (Security)

    arXiv:2610.35235v1 Announce Type: new Abstract: Published results in quantum machine learning for intrusion detection (QML-IDS) require joint interpretation of false-positive rate (FPR), attack true-positive rate (TPR), and evaluation…

  112. Atomic Security Concerns: Towards a Foundational Language for Security Knowledge Representation
    USarXiv cs.CR (Security)

    arXiv:2610.35782v1 Announce Type: new Abstract: Cybersecurity governance, risk management, compliance, and security architecture rely on numerous concepts, including security properties, domains, controls, risks, implementation measures…

  113. From Instructions to Exploits: Applying Program Analysis to Malicious Agent Skills Across Natural Language and Code
    USarXiv cs.CR (Security)

    arXiv:2610.33489v1 Announce Type: new Abstract: LLM-powered agents increasingly rely on third-party skills that combine natural-language instructions with executable code, creating supply-chain risks that may span both artifacts. Existing…

  114. Ranking Is Not Deciding: Benchmark Shortcuts and Operating-Point Collapse in Query-Time RAG Poisoning Defenses
    USarXiv cs.CR (Security)

    arXiv:2610.33849v1 Announce Type: new Abstract: Query-time filters that screen each retrieved document before it reaches the generator are the dominant published defense against retrieval-augmented generation poisoning, and they report…

  115. NameTrap: Steering AI Coding Agents Toward Security Vulnerabilities via Variable Names
    USarXiv cs.CR (Security)

    arXiv:2610.34253v1 Announce Type: new Abstract: AI coding agents routinely clone repositories, install packages, and read source code obtained from the Web, so publicly hosted code becomes part of the context that shapes the code they…

  116. From Representations to Computational Trajectories: Intrinsic Fingerprinting for Large Language Model Provenance Verification
    USarXiv cs.CR (Security)

    arXiv:2610.34675v1 Announce Type: new Abstract: Large language models (LLMs) are valuable digital assets, yet their weights can be easily modified after unauthorized acquisition, making model provenance difficult to verify. Model…

  117. SoK: Timestamps Are Not a Time Model: Rethinking Temporal Reasoning in Host-Based Intrusion Detection
    USarXiv cs.CR (Security)

    arXiv:2610.35151v1 Announce Type: new Abstract: Host-based intrusion detectors order events, window audit streams and sequence attack stages by their timestamps. Every conclusion they draw therefore rests on assumptions about clocks…

  118. PRISMA: Differentially Private Synthetic In-Car Telemetry via Differentiable Simulation
    USarXiv cs.CR (Security)

    arXiv:2610.35521v1 Announce Type: new Abstract: Differentially private synthesis of in-car energy consumption data is challenging because the output is high-dimensional, and known lower bounds show that the privacy-utility trade-off…

  119. Bounds on polynomial approximations of the discrete logarithm function
    USarXiv cs.CR (Security)

    arXiv:2610.33452v1 Announce Type: cross Abstract: We revisit a classical result of Coppersmith and Shparlinski showing bounds on how well polynomials can approximate the discrete log function. They showed that for any univariate polynomial…

  120. EdgeDeID: A 0.6B Clinical De-identifier Trained Without Real Patient Data
    USarXiv cs.CR (Security)

    arXiv:2610.33839v1 Announce Type: cross Abstract: De-identifying clinical notes with a hosted language model discloses the information the step is meant to protect. We present EdgeDeID, a 0.6B-parameter decoder-only de-identifier that runs…

  121. Reversible Pseudonymisation of Documents Sent to External LLMs: an Open Bilingual Benchmark and an Engine
    USarXiv cs.CR (Security)

    arXiv:2610.34012v1 Announce Type: cross Abstract: Professionals send working documents to external large language models. A common safeguard is reversible pseudonymisation: personal data are replaced with placeholders before the request…

  122. Decentralized Medical Data Sharing with Off-chain Storage and View-based Access Control
    USarXiv cs.CR (Security)

    arXiv:2610.34279v1 Announce Type: cross Abstract: Recently, many blockchain-based medical data sharing systems using off-chain storage have been proposed. However, these systems have some drawbacks such as being incompletely decentralized…

  123. Information-Theoretic Secure Aggregation With Heterogeneous Client Quantization
    USarXiv cs.CR (Security)

    arXiv:2610.35833v1 Announce Type: cross Abstract: In federated learning (FL), secure aggregation enables a server to compute the sum of local model updates of $N$ clients, without revealing any information on the individual model updates…

  124. From Caveman to COBA: Coordinated Optimization-Based Attacks on Grid-Edge Smart Inverters in DER-Rich Power Systems
    USarXiv cs.CR (Security)

    arXiv:2610.33401v1 Announce Type: new Abstract: Aggregation platforms and smart-inverter control channels can allow a compromised coordination layer to alter active- and reactive-power commands across a geographically distributed DER…

  125. "I Strongly Suspect This Website Is a Scam": Benchmarking PII Leakage and Detection without Defense in Autonomous Web Agents
    USarXiv cs.CR (Security)

    arXiv:2606.00497v2 Announce Type: replace Abstract: Deceptive web content, widely instantiated across the internet and commonly known as \textit{social-engineering attacks}, manipulates autonomous web agents into submitting users'…

  126. ISC Stormcast For Tuesday, October 27th, 2026 https://isc.sans.edu/podcastdetail/10152, (Tue, Oct 27th)
    USSANS ISC
  127. Cybersecurity Issue Forces a Major New Mexico Hospital to Divert Patients and Close Clinics
    USSecurityWeek

    The University of New Mexico Health System told people not to show up at its main hospital in Albuquerque and canceled appointments at more than 30 clinics. The post Cybersecurity Issue Forces a Major New Mexico Hospital to Divert Patients…

  128. Experts temper emergency-level alarms over critical Brotli defect
    USCyberScoop

    Alarm bells ratcheted up yet again about an AI-discovered vulnerability that could — if exploited in the wild — destabilize vast stretches of the internet. The critical defect — CVE-2026-111497 — affects the Google-developed Brotli…