AX·NEWS
Sign in

Cybersecurity

Cybersecurity news — Wednesday 21 October 2026

118 stories from 13 sources, every one in English. Each links to its publisher.

Open in the live feedRSS feedOn a TV

  1. Google Chrome adds Gemini-powered scam detection, warns users of ClickFix attacks
    USBleepingComputer

    Chrome is getting new scam protections that use Gemini to detect deceptive websites and warn users when they copy suspicious commands from ClickFix attacks. [...]

  2. Microsoft pins ConnectWise supply-chain attack to China-state aligned threat group
    USCyberScoop

    Microsoft researchers attributed the recent supply-chain attack targeting ConnectWise customers ’ Microsoft 365 environments to a China-state aligned threat actor it tracks as Storm-0940. “As early as September 2026, Storm-0940 used the…

  3. Japan's Onamae Desktop Cloud hacked, up to 27,000 servers affected
    USBleepingComputer

    Hackers compromised the infrastructure of Onamae Desktop Cloud, a popular provider of VPS services in Japan, and installed malware on up to 26,739 customers' servers. [...]

  4. Lazarus hackers use post-quantum cryptography for malware delivery
    USBleepingComputer

    North Korean hackers are using a post-quantum cryptographic mechanism to secure malicious payload delivery in new operations that combine old and new malware. [...]

  5. Critical GitLab Flaw (CVSS 10.0) Could Allow Unauthenticated RCE Across 12 Years of Releases
    INThe Hacker News

    GitLab released fixes on October 21 for a flaw that lets an attacker run code on a self-managed server without logging in, and it affects every version of the software going back to 7.0, released in 2014. The fix matters even for…

  6. WhatsApp’s chatbot was given hidden instructions to not mention privacy regulators
    USMalwarebytes Labs

    WhatsApp instructed its AI support chatbot not to recommend contacting privacy regulators when users couldn’t delete their accounts, researchers found. Privacy software company Redact discovered the hidden text on October 8. Its…

  7. Hold onto your old HomePod – an upgrade is on the way
    USZDNet Security

    An upcoming update will bring new features to existing HomePods. Here’s what to know.

  8. CISA: Hackers chain Plex flaws in Plex2Shell attacks to breach servers
    USBleepingComputer

    The U.S. Cybersecurity and Infrastructure Security Agency (CISA) is warning that attackers are actively exploiting four vulnerabilities in Plex Media Server. [...]

  9. OpenAI’s most powerful AI models turned off safety monitoring – then lied about it
    USZDNet Security

    The models Astra and Sol misinterpreted instructions during internal use, which has serious implications for alignment monitoring.

  10. LiteLLM Compromised in Supply Chain Attack Delivering Credential Stealer and Backdoor
    INThe Hacker News

    The popular LiteLLM Python package has been compromised as part of a software supply chain attack that delivered a credential stealer and a backdoor, according to reports from SafeDep, Socket, StepSecurity, and Upwind. The compromise…

  11. Google Hardens Chrome Against ClickFix Attacks, Insecure HTTP Connections
    USSecurityWeek

    Protections have been added against scams and social engineering, including new site warnings and server-side Gemini analysis. The post Google Hardens Chrome Against ClickFix Attacks, Insecure HTTP Connections appeared first on…

  12. PHP2Shell Patches Land as First Exploitation Attempts Hit Sensors
    USSecurityWeek

    Several cybersecurity companies saw exploitation attempts of CVE-2026-109618 and CVE-2026-109619 before PHP released the patches. The post PHP2Shell Patches Land as First Exploitation Attempts Hit Sensors appeared first on SecurityWeek .

  13. Chrome warns you about suspicious websites now – how to turn on Google’s new scam protection
    USZDNet Security

    Google’s browser will flash two types of warnings based on whether the site is deemed dangerous or just suspicious.

  14. OpenAI's GPT-6 Agent Removed Safety Monitoring, Then Falsely Denied Doing So
    INThe Hacker News

    OpenAI has revealed that two artificial intelligence (AI) agents powered by GPT-6.1 Sol and GPT-6 Astra weakened misalignment monitoring while making requested software changes during two internal coding tasks. The AI company said the…

  15. OpenAI agents disabled safety monitoring, one lied about it
    USBleepingComputer

    OpenAI agents powered by the GPT-6.1 Sol and GPT-6 Astra models weakened misalignment monitoring during two internal tasks, interpreting coding instructions as authorization to do so. [...]

  16. PHP2Shell: Why websites are going offline for emergency security updates
    USMalwarebytes Labs

    If you’ve encountered websites in maintenance mode today, it may be a security precaution. Reports of attacks targeting two critical PHP vulnerabilities have prompted warnings to install emergency updates and consider temporarily taking…

  17. Widespread PHP Flaw May Cause Many Site Outages Today
    USKrebs on Security

    Expect to see quite a few websites serving error pages today: Nearly 70 percent of websites need to update their core software to fix a pair of vulnerabilities that allow an attacker to seize full control over affected servers. Experts say…

  18. ‘Context Bombs’ Cut AI Attack Success by Misleading Agents With Fake Targets
    USSecurityWeek

    Deceptive prompts lure AI agents into decoys and dead ends, cutting attack success by up to 91% in tests and raising the cost of a successful attack 23-fold. The post ‘Context Bombs’ Cut AI Attack Success by Misleading Agents With Fake…

  19. High-Severity Flaws Exploited in Three Days as Ransomware Dominates Q3: Rapid7
    USSecurityWeek

    Rapid7’s Q3 analysis finds high-severity flaws exploited within three days on average, while AI appeared in just 8% of tracked incidents. The post High-Severity Flaws Exploited in Three Days as Ransomware Dominates Q3: Rapid7 appeared…

  20. PHP ships emergency patches for critical PHP2Shell flaws
    USBleepingComputer

    The PHP security team released emergency security patches for two critical remote code execution (RCE) vulnerabilities (CVE-2026-109618 and CVE-2026-109619) tracked collectively as PHP2Shell. [...]

  21. Amazon unveiled 5 new Blink cameras, including a 4K model that caught my eye
    USZDNet Security

    Five new Blink cameras are coming to the market, and the extended-range model has all my attention.

  22. ClickFix has never been bigger, and AI tools are its most convincing lure yet
    USBleepingComputer

    ClickFix attacks are surging as cybercriminals use fake AI installers, malicious search results, and trusted chatbot platforms to trick users into executing malware. Push Security examines how these attacks evade traditional defenses and…

  23. PHP Issues Emergency Patches for Two Critical PHP2Shell RCE Flaws as Exploit Attempts Emerge
    INThe Hacker News

    PHP has released emergency updates that fix two critical flaws in the language itself, letting an attacker without an account run code on a web server. The fixes are in PHP 8.5.12, 8.4.27, 8.3.36 and 8.2.35, and the PHP team says only…

  24. Prepare yourselves: PHP to release emergency fixes for 2 critical vulnerabilities
    USCyberScoop

    The PHP project released patches Wednesday for two critical vulnerabilities in a programming language used in an extremely large portion of the internet. The patch addresses two flaws, catalogued as CVE-2026-109618 and CVE-2026-109619…

  25. Asymptote Labs Lands $5.5 Million Seed for AI Agent Runtime Security
    USSecurityWeek

    Asymptote Labs raises $5.5M seed led by Tower Research Ventures to commercialize its Beacon AI agent runtime security platform. The post Asymptote Labs Lands $5.5 Million Seed for AI Agent Runtime Security appeared first on SecurityWeek .

  26. Google Chrome Adds ClickFix Warnings to Stop Users From Running Malicious Commands
    INThe Hacker News

    Google on Tuesday said it's rolling out improved safeguards in its Chrome browser to secure users against ClickFix attacks when Enhanced Protection is enabled. ClickFix is a popular social engineering technique in which bad actors trick…

  27. Memory Bugs Account for Most of Chrome’s 353 New Security Fixes
    USSecurityWeek

    The update fixes 353 flaws, including 163 memory corruption and 125 use-after-free vulnerabilities. The post Memory Bugs Account for Most of Chrome’s 353 New Security Fixes appeared first on SecurityWeek .

  28. Fake Amazon Music x YouTube deal lures creators into a Google login trap
    USMalwarebytes Labs

    .kadence-column476672_104e78-a1 > .kt-inside-inner-col{padding-top:20px;padding-right:20px;padding-bottom:10px;padding-left:20px;}.kadence-column476672_104e78-a1 > .kt-inside-inner-col{column-gap:var(--global-kb-gap-sm…

  29. Critical PHP Vulnerability (CVE-2026-109618, CVE-2026-109619) ＆#x5b;infocon yellow＆#x5d;, (Wed, Oct 21st)
    USSANS ISC

    PHP.net announced that it will release a critical update today. We decided to change the infoCon to YELLOW due to the widespread use of PHP, the expected severity of the flaw (CVSS 9.4), and the availability of early PoC exploits before…

  30. OpenAI Agents Silenced Security Team Alerts, Removed Misalignment Monitor
    USSecurityWeek

    In two internal incidents, agents read routine coding instructions as permission to weaken monitoring, and one gave a false answer about it. The post OpenAI Agents Silenced Security Team Alerts, Removed Misalignment Monitor appeared first…

  31. Microsoft PowerToys adds CursorHide, Button Lock mouse tools
    USBleepingComputer

    Microsoft has released PowerToys 0.102, with two new mouse utilities, new command-line tools, and Command Palette improvements. [...]

  32. China-Aligned Bronze Sandcastle Abuses Sandboxie DLL Sideloading to Deploy Backdoors
    INThe Hacker News

    Cybersecurity researchers have disclosed details of a previously undocumented China-aligned threat cluster known as Bronze Sandcastle that has been observed targeting organizations in Asia and Africa since mid-2025. "The most distinctive…

  33. Critical security issue affecting PHP
    GBNCSC UK

    Updated: The NCSC is encouraging UK organisations to take immediate action to mitigate a security issue affecting PHP (CVE-2026-109618 and CVE-2026-109619).

  34. Generative AI Is Erasing Behavioral Tells, Weakening Behavior-Only Detection, Proofpoint Reports
    INThe Hacker News

    Generative AI has changed what a malicious email looks like. Messages that once gave themselves away through clumsy wording, generic pretexts, and reused templates now arrive fluent, personalized, and unique to each recipient, and the…

  35. Signal bug could let someone keep listening after you hang up
    USMalwarebytes Labs

    A recently disclosed flaw in Signal’s calling software could allow another participant to keep hearing you after you hang up. The vulnerability, tracked as CVE-2026-113380 ( CVSS score 9.3 out of 10), affects RingRTC versions earlier than…

  36. Exclusive: ServiceNow and Nvidia launch Arc, a new AI agent for enterprise
    USZDNet Security

    The two companies are trying to solve the AI agent security conundrum with multiple layers of real-time security checks.

  37. Browsers Block eMudhra Intermediate CAs Over Recoverable Private Keys
    USSecurityWeek

    The private keys associated with eight certificates were compromised, and root programs and trust stores are blocking them. The post Browsers Block eMudhra Intermediate CAs Over Recoverable Private Keys appeared first on SecurityWeek .

  38. Workshop on Economics and Information Security (WEIS) Archive
    USSchneier on Security

    There’s a new archive of all the papers from twenty-five years of the Workshop on Economics and Information Security (WEIS). Really nice resource. It’s fun to reread my 2002 essay on the economics of information security.

  39. China Defends DeepSeek After US Ban of Its AI From Government Systems
    USSecurityWeek

    The DHS said agencies should remove DeepSeek products as well as third-party products and services developed using DeepSeek from their IT systems. The post China Defends DeepSeek After US Ban of Its AI From Government Systems appeared…

  40. PHP2Shell: What Actually Protects You Before the 14:00 UTC PHP RCE Patch, and What Doesn't
    INThe Hacker News

    You cannot fully block the two critical PHP flaws called PHP2Shell before today's fix, and the one setting being passed around as a workaround is not enough on its own, the security firm that suggested it now says. PHP will release the fix…

  41. The Harness Is the Hard Part of Agentic Pentesting
    INThe Hacker News

    An AppSec team can point an AI agent running on a general-purpose model at its own web apps with little setup. Sure, the model writes the attacks. But everything that makes those attacks safe to run against production and worth acting on…

  42. Attackers Hijacked Over 300 Tor Exit Relays to Intercept SSH Connections
    INThe Hacker News

    Attackers took control of more than 300 Tor exit relays in early October and used them to redirect SSH connections to servers that stole login passwords, a researcher found. Exit relays are the last hop in a Tor circuit, where traffic…

  43. Microsoft says October updates didn’t actually fix VPN issues
    USBleepingComputer

    Microsoft says it mistakenly flagged a bug causing Always On VPN connection issues as fixed in the October 2026 Windows 11 cumulative updates. [...]

  44. Chinese AI Model Writes More Hackable Code When Told It’s for US Contractor
    USSecurityWeek

    Tasked with writing an application for a US defense contractor, the model made the code 19% more likely to be hackable. The post Chinese AI Model Writes More Hackable Code When Told It’s for US Contractor appeared first on SecurityWeek .

  45. Attackers Abuse UpdraftCentral Keys to Plant Backdoors on WordPress Sites
    INThe Hacker News

    Unknown threat actors breached TeamUpdraft's website and potentially gained unauthorized access to customers' connection keys for UpdraftCentral, UpdraftVault, and UpdraftClone, using them to backdoor WordPress sites. TeamUpdraft is the…

  46. CISA: SafePay ransomware hit over 600 critical infrastructure orgs
    USBleepingComputer

    The Cybersecurity and Infrastructure Security Agency (CISA) says the SafePay ransomware gang has breached over 600 critical infrastructure organizations worldwide since 2024. [...]

  47. U.S. Orders DeepSeek Removed From Federal Civilian Systems, Bans New Purchases
    INThe Hacker News

    U.S. federal civilian agencies must remove DeepSeek products from their information technology systems, along with any third-party products or services "developed using DeepSeek implementations," under a Department of Homeland Security…

  48. State Bank of India Confirms Data Leak After 3,897 Customer Records Found on Dark Web
    INThe Hacker News

    State Bank of India has identified 3,897 customer records containing personal information in a set of about 375,000 records available on the dark web, and it does not yet know where the data came from. The state-owned bank, India's…

  49. CISA Orders Federal Agencies to Patch Exploited Plex2Shell, Siemens, and SiteGuard Flaws
    INThe Hacker News

    The U.S. Cybersecurity and Infrastructure Security Agency (CISA) on Tuesday added four vulnerabilities to its Known Exploited Vulnerabilities (KEV) catalog, including four vulnerabilities impacting Plex Media Server that's collectively…

  50. PHP2Shell: Mass Exploitation Looms for Leaked Critical PHP Flaws
    USSecurityWeek

    Tracked as CVE-2026-109618 and CVE-2026-109619, the bugs could potentially lead to unauthenticated remote code execution. The post PHP2Shell: Mass Exploitation Looms for Leaked Critical PHP Flaws appeared first on SecurityWeek .

  51. Hackers Hid Vasilek Backdoor in VMware Tools During 2-Year Russian Healthcare Intrusion
    INThe Hacker News

    A Belarus-aligned hacktivist group spent nearly two years within the network of a Russian healthcare organization. According to a report published by Russian cybersecurity firm Solar, the activity has been attributed to a threat cluster…

  52. Hijacked Developer Account Used to Plant Stealer in OVHcloud CLI
    USSecurityWeek

    Researchers say the developer’s account was compromised in March with a GlassWorm-linked loader. The post Hijacked Developer Account Used to Plant Stealer in OVHcloud CLI appeared first on SecurityWeek .

  53. CrowdStrike Uncovers New Agentic AI Campaigns
    USCrowdstrike Blog
  54. CrowdStrike Protects Customers from Threats Delivered via PHP2Shell
    USCrowdstrike Blog
  55. PHP2Shell: Two Unpatched Critical PHP RCE Flaws (CVSS 9.2) Trigger Emergency Security Alert
    INThe Hacker News

    This story is developing and was written before PHP's scheduled 14:00 UTC release on October 21. Once the release is out, the affected and fixed versions below are confirmed by the official advisory. PHP will ship an out-of-cycle security…

  56. Accenture, Dragos, SANS Launch $10M Initiative to Defend Small US Utilities
    USSecurityWeek

    The first phase will cover up to 500 utilities, which will get free security tools, equipment, deployment help and training. The post Accenture, Dragos, SANS Launch $10M Initiative to Defend Small US Utilities appeared first on…

  57. Adversarial Benchmarks and BenchDoor Attacks: Exploiting Benchmark Incentives for Backdoor Attacks
    USarXiv cs.CR (Security)

    arXiv:2610.27533v1 Announce Type: new Abstract: Machine learning benchmarks act as guideposts for model development and spur the creation of training datasets intended to improve benchmark scores. We introduce BenchDoor, a family of data…

  58. A Digital Twin to Improve Cyber-Social Impact
    USarXiv cs.CR (Security)

    arXiv:2610.28430v1 Announce Type: new Abstract: The increasing complexity of Industrial Control Systems (ICS) and their integration with IT infrastructures has significantly expanded the cyber threat landscape. Traditional cybersecurity…

  59. Evidence-Grounded LLM-Assisted Android Mobile Forensic Reporting
    USarXiv cs.CR (Security)

    arXiv:2610.26946v1 Announce Type: new Abstract: Large language models can assist forensic report drafting, but fluent output does not ensure evidentiary reliability. This study presents an evidence-grounded framework for evaluating…

  60. BR-MoE: Budgeted Semantic Release for Privacy-Preserving Traffic Analytics
    USarXiv cs.CR (Security)

    arXiv:2610.27026v1 Announce Type: new Abstract: Traffic cameras support useful analysis, but their outputs can reveal private information. Replacing images with text reduces direct visual exposure. It does not place a formal bound on…

  61. CyberHull: Deterministic Malicious URL Isolation via Convex Hull
    USarXiv cs.CR (Security)

    arXiv:2610.27675v1 Announce Type: new Abstract: The emergence of increasingly sophisticated and evasion-resistant phishing attack strategies represents a significant threat to modern cybersecurity, particularly in cyber-physical systems…

  62. ClawMMU: Protection-Coherent Virtual Memory for Stateful LLM Agents
    USarXiv cs.CR (Security)

    arXiv:2610.28093v1 Announce Type: new Abstract: AI agents keep long-term memory, much of which comes from web pages and tool results. Agent harnesses, the software that runs the model, do not record who may act on this memory, so an…

  63. Privacy-Preserving Personalized Federated Learning (PPP-FL) for Medical Imaging
    USarXiv cs.CR (Security)

    arXiv:2610.27470v1 Announce Type: new Abstract: Federated learning (FL) is a distributed machine learning paradigm that reduces transmission costs, mitigates cross-border data sharing restrictions, and enhances privacy by keeping raw data…

  64. SRE: Accelerating Zero-Knowledge Proofs via Scalar Redundancy Elimination in Multi-Scalar Multiplication
    USarXiv cs.CR (Security)

    arXiv:2610.27452v1 Announce Type: new Abstract: Multi-Scalar Multiplication (MSM) is a fundamental cryptographic operation used in Elliptic Curve Cryptography (ECC) across a wide range of applications. Its computation involves intensive…

  65. AutoRedClaw: Automated Red-Teaming for Claw-Style Agents
    USarXiv cs.CR (Security)

    arXiv:2610.27244v1 Announce Type: new Abstract: Agent systems are shifting from task-scoped tool-use agents to persistent, personalized assistants such as OpenClaw. These claw-style agents maintain configuration files (e.g., AGENTS.md)…

  66. Reliable Estimation of Differential Privacy under Selection Bias
    USarXiv cs.CR (Security)

    arXiv:2610.27021v1 Announce Type: new Abstract: Statistical privacy audits are essential for verifying the theoretical privacy guarantees of differentially private (DP) mechanisms. Auditors typically use Membership Inference Attacks (MIAs)…

  67. ECHO: Revealing Review-to-Submission Linkage in Anonymous Peer Review
    USarXiv cs.CR (Security)

    arXiv:2610.27044v1 Announce Type: new Abstract: Anonymous peer review protects reviewers from retaliation, yet reviewer anonymity may also be weakened by artifacts released through the review process itself. This risk is growing as many…

  68. Model Exposure on One Device Endangers Others: Rethinking Edge Neural Network Security across Devices
    USarXiv cs.CR (Security)

    arXiv:2610.27152v1 Announce Type: new Abstract: Neural networks (NNs) are increasingly deployed on edge devices such as mobile and IoT platforms, offering low latency and improved privacy compared to cloud-based deployment. However…

  69. A Combined Streaming FPGA Architecture for Client-Side CKKS Encoding and Encryption
    USarXiv cs.CR (Security)

    arXiv:2610.27204v1 Announce Type: new Abstract: Homomorphic encryption allows a computing service to process encrypted data, but the client must first encode and encrypt its inputs. Sharing hardware between floating-point encoding and…

  70. When Models Know but Can't Linearly Say: Relational Encoding Limits Linear Vulnerability Classification in Code LLMs
    USarXiv cs.CR (Security)

    arXiv:2610.27257v1 Announce Type: new Abstract: Large language models can detect vulnerable code, yet the form of vulnerability encoding in their representations remains unclear. We analyze 8 code models spanning 6.7B-15B parameters on C…

  71. Guided Diffusion for Trustworthy Synthetic Text Generation
    USarXiv cs.CR (Security)

    arXiv:2610.27259v1 Announce Type: new Abstract: The rapid evolution of Large Language Models is fueled by massive datasets, yet this reliance poses severe privacy risks, as models can inadvertently memorize and even directly output…

  72. Make the Adversary Spend More Energy: Generative Deception against Language-Model Attackers in Critical Infrastructure
    USarXiv cs.CR (Security)

    arXiv:2610.27739v1 Announce Type: new Abstract: Language-model attackers consume computational resources as they explore and interact with a target, creating an opportunity for cyber deception to impose costs. Whether such engagements…

  73. MAS-Guard: Unified Guardrails from Individual Agents to Multi-Agent LLM Systems
    USarXiv cs.CR (Security)

    arXiv:2610.28397v1 Announce Type: new Abstract: Agentic AI extends large language models (LLMs) from generating responses to planning and executing complex workflows, substantially expanding their utility while exposing applications to…

  74. Privacy-Preserving Robustness-Guided Testing
    USarXiv cs.CR (Security)

    arXiv:2610.27441v1 Announce Type: new Abstract: In industries like semi-conductors and aerospace, it is common for the design company to contract out part of the testing process to a Testing-as-a-Service (TaaS) company. In such a setup…

  75. DISTA-LP: Privacy-Preserving Distributed Linear Programming for Metric Differential Privacy
    USarXiv cs.CR (Security)

    arXiv:2610.27615v1 Announce Type: new Abstract: Linear-programming (LP) mechanisms for metric differential privacy (mDP) enable fine-grained, task-aware utility control, but they are computationally expensive and become privacy-fragile in…

  76. VeriRouter: Towards Verifiable LLM Routing
    USarXiv cs.CR (Security)

    arXiv:2610.17431v2 Announce Type: replace Abstract: Large language model (LLM) routers are increasingly deployed to optimize quality, latency, and cost through dynamic model selection and request rewriting. However, a compromised router…

  77. Mind Your Boundaries: Boundary-Aware Model Obfuscation against Cross-node Attack
    USarXiv cs.CR (Security)

    arXiv:2610.27135v1 Announce Type: new Abstract: With the advancement of powerful AI chips, on-device machine learning is gaining traction for its low latency and privacy protection. However, when edge devices are compromised, attackers can…

  78. Energy Consumption of Post-Quantum Cryptography on Constrained and General-Purpose Architectures
    USarXiv cs.CR (Security)

    arXiv:2610.27111v1 Announce Type: new Abstract: With Q-day approaching, the transition to post-quantum cryptography (PQC) has begun, with governments across the US, UK and EU mandating migration to quantum-resistant standards. This paper…

  79. Testing TLS 1.3 PQC on constrained devices and microcontrollers
    USarXiv cs.CR (Security)

    arXiv:2610.27766v1 Announce Type: new Abstract: The UK, EU and US governments have set timelines for post-quantum cryptography migration of critical systems, with deadlines ranging from 2028 to 2035. This paper focuses on IoT-class devices…

  80. The Show-Up Attack: When Subagents Win Over Their Supervisor
    USarXiv cs.CR (Security)

    arXiv:2610.27574v1 Announce Type: cross Abstract: Multi-agent systems are increasingly deployed to coordinate multiple large language model (LLM) agents. However, this also creates new attack surfaces: malicious agents may have chances to…

  81. Beyond the Prompt: Recovering Hidden Concepts in Text-to-Image Models
    USarXiv cs.CR (Security)

    arXiv:2610.27473v1 Announce Type: new Abstract: Text-to-image models can generate visual concepts that their prompts never mention, and recent work shows that such associations can be implanted so that ordinary prompts trigger them…

  82. On the Vulnerabilities and Data-Free Removal of Watermarks in Batch Normalization Layers
    USarXiv cs.CR (Security)

    arXiv:2610.28076v1 Announce Type: new Abstract: Batch Normalization (BN) layers have become a common choice for embedding white-box watermarks in Federated Learning (FL), providing a simple and effective mechanism for model ownership…

  83. Contextual Candour: Role-Aware Graduated Disclosure for LLMs over Mixed-Sensitivity Contexts
    USarXiv cs.CR (Security)

    arXiv:2610.28067v1 Announce Type: new Abstract: LLMs increasingly operate over mixed-sensitivity contexts, where they must reason over information ranging from public to confidential while disclosing only what the user is authorised to…

  84. Interflow: Breaking the Offline-Online Barrier for Private AI at the Edge
    USarXiv cs.CR (Security)

    arXiv:2610.27431v1 Announce Type: cross Abstract: Privacy-preserving machine learning (PPML) offers a promising path toward bringing powerful AI services to the edge without exposing sensitive user data. However, its practical deployment…

  85. Efficient and Adaptable Detection of Malicious LLM Prompts via Bootstrap Aggregation
    USarXiv cs.CR (Security)

    arXiv:2602.08062v2 Announce Type: replace-cross Abstract: Large Language Models have demonstrated remarkable capabilities in natural language understanding. However, they remain susceptible to malicious prompts that induce unsafe or…

  86. Profile-DP: Differential Privacy for User-Controlled Profile Release in Personalized LLM Inference
    USarXiv cs.CR (Security)

    arXiv:2610.27213v1 Announce Type: new Abstract: Personalized large language models (LLMs) increasingly rely on profile-conditioned inference, pairing public requests with rich, user-specific facts. This paradigm exposes the user profile…

  87. After the Attacker Leaves: Benchmarking Carryover Attacks on LLM Agents
    USarXiv cs.CR (Security)

    arXiv:2610.28425v1 Announce Type: new Abstract: Modern LLM agents persist state across sessions via two channels: long-lived memory files silently prepended to the system prompt, and scheduled tasks waking the agent without user input. A…

  88. HydraPoT: A Configurable Multi-Agent Framework for Cost-Aware LLM-Assisted Honeypots
    USarXiv cs.CR (Security)

    arXiv:2610.27722v1 Announce Type: new Abstract: Large language models (LLMs) can improve honeypot realism by generating dynamic, context-aware responses, but powerful cloud models incur substantial cost and latency, while lightweight…

  89. Latency-Neutral Passive Cyber Monitoring for Grid-Edge Operational Technology Networks
    USarXiv cs.CR (Security)

    arXiv:2610.28324v1 Announce Type: cross Abstract: Deterministic industrial networks often present an operational trade-off between maintaining microsecond timing of primary control loops and capturing security telemetry. This work…

  90. Act on Your Behalf, Beyond Your Authorization: Benchmarking LLM Agent Authorization Alignment
    USarXiv cs.CR (Security)

    arXiv:2610.27548v1 Announce Type: cross Abstract: When an LLM agent acts on behalf of a user, it is expected to act within the authorization scope the user grants. Existing agent evaluations largely measure task completion or harmful…

  91. Local Allow, Global Deny: Why Per-Action Authorization Does Not Compose in LLM Agents
    USarXiv cs.CR (Security)

    arXiv:2610.27620v1 Announce Type: new Abstract: Autonomous LLM agents act through tools under per-action authorization: a local guard screens each tool call in isolation. We show that local authorization does not compose into global…

  92. Exposing Vulnerability Monoculture Through Systemic Adversarial Robustness
    USarXiv cs.CR (Security)

    arXiv:2610.27353v1 Announce Type: new Abstract: Beneath the apparent diversity of foundation models lies an increasingly homogeneous technical stack of common training pipelines, distillation methods, and data. These shared algorithmic…

  93. CAESAR: Mixed Reality Search-and-Rescue Platform for Cognitive Attack Evaluation
    USarXiv cs.CR (Security)

    arXiv:2610.27654v1 Announce Type: cross Abstract: Mixed Reality (MR) enables immersive training by tightly coupling virtual content with the physical environment and supporting embodied interaction. In safety-critical collaborative tasks…

  94. Secure Agentic Coding through Counterexample-Grounded Feedback
    USarXiv cs.CR (Security)

    arXiv:2610.25916v2 Announce Type: replace Abstract: As coding models and agent harnesses become more capable, reliable security verification is becoming a central bottleneck. Improving secure code generation requires a signal whose…

  95. NetGuise: Adversarial Traffic Generation for Early-Stage Classifier Evasion
    USarXiv cs.CR (Security)

    arXiv:2610.27456v1 Announce Type: cross Abstract: Early-stage encrypted traffic classifiers make decisions from only the first few packets of a connection, enabling low-latency classification but raising robustness concerns under realistic…

  96. Secure Multimodal Federated Optimization via Hierarchical DNA-Based Key Management
    USarXiv cs.CR (Security)

    arXiv:2610.27644v1 Announce Type: new Abstract: Federated optimization over heterogeneous multimodal data enables distributed institutions to collaboratively optimize a shared model using medical images, genomic profiles, clinical records…

  97. When Pebbles Are Not Atomic: A Polynomial Separation Between Pure and Entangled Pebbling
    USarXiv cs.CR (Security)

    arXiv:2610.27249v1 Announce Type: new Abstract: Graph pebbling is the model in which the memory cost of memory-hard functions and proofs of space is measured, and pebbling lower bounds are the main way to argue that such constructions are…

  98. Adversarial procurement in blockchains
    USarXiv cs.CR (Security)

    arXiv:2605.05559v2 Announce Type: replace-cross Abstract: An emerging blockchain protocol design pattern leverages the asymmetry between the computational effort in performing versus verifying tasks. For example, cryptographic validity…

  99. Bridging Digital Deception and Physical Fidelity: A Spatio-Temporal Attack on Cognitive Radars
    USarXiv cs.CR (Security)

    arXiv:2610.27585v1 Announce Type: cross Abstract: As cognitive radars evolve towards the end-to-end processing of one-dimensional complex baseband sequences, deep networks face severe adversarial threats. However, existing attacks…

  100. A Unified Attack Framework against Statistical Generative Watermarks for Large Language Models
    USarXiv cs.CR (Security)

    arXiv:2610.27653v1 Announce Type: new Abstract: Watermarking has emerged as a key tool for identifying text generated by large language models (LLMs), with current schemes embedding statistical signals into the generation process to enable…

  101. Disclosure Begets Disclosure: A Self-Reinforcing Privacy Loop in LLM Agents with Persistent Memory
    USarXiv cs.CR (Security)

    arXiv:2610.28315v1 Announce Type: new Abstract: Large language model (LLM) agents increasingly use persistent memory to personalize responses, and personal attributes in this memory pose a privacy risk if exposed. We ask how consulting…

  102. Context-Aware Prompt Injection Detection via Small Language Models in RAG Pipelines: nokast-secureRAG
    USarXiv cs.CR (Security)

    arXiv:2610.28364v1 Announce Type: new Abstract: Local Retrieval-Augmented Generation (RAG) systems are vulnerable to indirect prompt injection and retrieval poisoning, where malicious instructions in retrieved content are mistaken for…

  103. SPECTER: Covert Data Exfiltration from Air-Gapped Systems via USB Power Modulation
    USarXiv cs.CR (Security)

    arXiv:2610.28401v1 Announce Type: new Abstract: We present SPECTER, a post-compromise covert channel that uses host-side USB power control to exfiltrate information from air-gapped and network-segregated systems. A transmitter with…

  104. WaveCommitments: Private and Verifiable Speech-Agent Marketplaces
    USarXiv cs.CR (Security)

    arXiv:2610.26867v1 Announce Type: cross Abstract: Modern speech systems are evolving into ecosystems of autonomous speech agents that speak together, negotiate, and trade assets such as embeddings, spectrograms, or raw data. In such…

  105. You Are the Trigger: Speaker-Activated Backdoors
    USarXiv cs.CR (Security)

    arXiv:2610.27210v1 Announce Type: cross Abstract: Speech models are increasingly used as an interface to AI systems, from transcription tools to voice agents. Existing audio backdoors require an attacker to introduce triggers at inference…

  106. Right Facts, Wrong Rule: Scope-Transfer Poisoning of Retrieval-Augmented Generation
    USarXiv cs.CR (Security)

    arXiv:2609.35155v2 Announce Type: replace Abstract: Retrieval-augmented generation (RAG) systems use external documents for facts and rules that guide answers. We study rule-scope misapplication, where a rule valid within its documented…

  107. Proof-Carrying Contracts: Commitments AI Agents Cannot Break
    USarXiv cs.CR (Security)

    arXiv:2610.28484v1 Announce Type: cross Abstract: AI agents increasingly act on behalf of people or other agents, purchasing services, exchanging information and delegating work. What an agent must respect is written in instructions…

  108. Scrambled Multiply-with-Carry Pseudorandom Number Generators
    USarXiv cs.CR (Security)

    arXiv:2610.27313v1 Announce Type: new Abstract: Multiply-with-carry (MWC) generators are very fast pseudorandom number generators whose quality depends on the choice of a multiplier, and is usually assessed by applying the spectral test to…

  109. Global Security Blind Spots: Translation as a Safety Bypass in Low-Resource Languages
    USarXiv cs.CR (Security)

    arXiv:2610.27304v1 Announce Type: cross Abstract: Large language models are deployed globally, yet safety alignment and evaluation remain concentrated in English: harmful requests refused in English can elicit unsafe completions when…

  110. Big Secrets Small Shares: Revisiting Secure Multi-Party Computation
    USarXiv cs.CR (Security)

    arXiv:2610.27609v1 Announce Type: new Abstract: At CRYPTO 2007, Damg{\aa}rd and Nielsen introduced a preprocessing approach to multi-party computation (MPC) to achieve linear communication complexity. Another approach to reduce…

  111. Post-quantum commutative group actions from oriented superspecial abelian varieties
    USarXiv cs.CR (Security)

    arXiv:2610.27922v1 Announce Type: cross Abstract: We study the cryptographic use of Shimura class group actions on superspecial principally polarized abelian varieties equipped with a CM orientation, within the context of post-quantum…

  112. KNOX: Hardware-assisted Regulatory Sandbox for Generative Models
    USarXiv cs.CR (Security)

    arXiv:2610.27543v1 Announce Type: new Abstract: Regulations require companies to run AI models in a regulatory sandbox, an isolated environment that enables monitoring model behavior by regulators, prior to deployment. But, they do not…

  113. Privacy Accounting for Empirical Denoisers
    USarXiv cs.CR (Security)

    arXiv:2610.27710v1 Announce Type: cross Abstract: Diffusion models have emerged as the foundation of modern generative systems, yet their high memorization capacity raises privacy concerns. While differentially private (DP) training…

  114. How Intervention History Changes AI Agent Sabotage on Subsequent Tasks
    USarXiv cs.CR (Security)

    arXiv:2610.26935v1 Announce Type: new Abstract: AI control evaluations often reset agents' memory between tasks, although deployed agents may retain records of prior interventions. We ask whether learning that an earlier attack was caught…

  115. ISC Stormcast For Wednesday, October 21st, 2026 https://isc.sans.edu/podcastdetail/10144, (Wed, Oct 21st)
    USSANS ISC
  116. Stolen UpdraftCentral keys used to backdoor WordPress sites
    USBleepingComputer

    A threat actor breached TeamUpdraft's website, a developer of popular WordPress backup and security plugins, potentially compromising UpdraftCentral connection keys that can send commands to connected websites. [...]

  117. Mozilla, Google, Apple block compromised eMudhra CA certificates
    USBleepingComputer

    Certificate authority eMudhra revoked eight compromised intermediate certificates after researchers found their private keys were so weak that anyone could recover them almost instantly and use them to forge trusted certificates for…

  118. PHP2Shell Exploitation: Two Critical PHP Flaws Explained
    USRecorded Future

    Two critical remote code execution vulnerabilities in the PHP engine, CVE-2026-109618 and CVE-2026-109619 (nicknamed “PHP2Shell”), are reportedly being targeted in active exploitation attempts following the leak of technical details ahead…