All news with #research tag

🛡️ Security researchers detailed a local privilege escalation that gains root on OpenBSD 7.9 by chaining a decades-old integer overflow in shmat(4) with a separate kernel reference-counting flaw. Calif published a full writeup and proof-of-concept after OpenBSD issued errata 018 and 026 in September, and patches are available via syspatch for all architectures. The chain bypasses KARL, SMAP, and SMEP by corrupting shared-memory counters and leveraging deterministic kernel allocator behavior. Calif also proposed multiple hardening measures to reduce similar risks.

🔍 A security report from HiddenLayer argues that TypeSafe’s Jev, a streamlined model designed for machine-to-machine decisions, is more vulnerable to certain adversarial attacks than traditional LLMs. The researchers found Jev susceptible to prompt injection and TokenBreak-style perturbations that can alter outputs and trigger harmful actions, despite its performance and lower token use. Analysts note the testing methodology is sound, though some call the conclusions premature given Jev’s recent release. The report urges caution, application-level testing, and independent checks around critical outcomes.

🛡️ A critical Linux kernel networking flaw, tracked as CVE-2026-99999 and nicknamed PwnEscape, lets code running inside a VM read host memory and potentially crash or execute code on the host. The issue was discovered in a Vercel security challenge by researcher Paulos Yibelo of pwn.ai and confirmed by Vercel; no evidence of real-world exploitation has been reported. Fixes have been merged and backported to multiple stable kernel branches; hosts must be updated and rebooted to be protected.

🔎 Cloudflare and OpenAI ran a research pilot to see if network signals can help ChatGPT find more useful web pages. The pilot surfaced hundreds of millions of previously undiscovered pages into OpenAI’s index, helping ChatGPT provide better answers and citations. The work focuses on efficient discovery, filtering noisy or harmful pages, and preserving site owner control over access.

🔍 Cloudflare examined traffic impacts after reported strikes on Yandex data centers, which caused dramatic drops in HTTP volume for two Yandex ASNs: AS13238 (Search) and AS200350 (Cloud). Using Cloudflare Radar and Bot Management, the team found both ASNs saw reduced traffic but with different automation profiles: Yandex Search was overwhelmingly verified crawler activity, while Yandex.Cloud showed unverified, app-like bot traffic. The outage reduced requests from those networks but did not produce measurable, lasting reductions in global malicious or social media manipulation traffic.

🔒 Dashlane's analysis of millions of aggregated, anonymized authentications from July 2025 to June 2026 shows passkey authentications nearly doubled year-over-year to 2.5 million monthly in 2026, and 57% of users now store at least one passkey. Despite growth, passkeys remain a minority of logins on major platforms like Google, PayPal, Amazon and eBay. Dashlane urges businesses and developers to make passkeys visible, default and easier to use to close the adoption gap and reduce account takeover risk.

🚀 Space-based services underpin modern life, yet cyber risks to orbital and ground infrastructure remain largely unaddressed by corporate leadership. Rapid commercialisation has introduced enterprise-style vulnerabilities into satellite systems, while nation-state and criminal actors increasingly target space-sector assets. A coordinated outage could cause severe economic and physical impacts, prompting urgent calls for unified, platform-based security across the entire space supply chain.

🔍 New research from Chainguard shows that companies with $1bn+ revenue are more than twice as likely as firms under $250m to sometimes leave high or critical vulnerabilities unpatched. The study, presented at Assemble London on October 14, highlights differing barriers: larger organisations cite business trade-offs and operational disruption, while smaller firms point to capacity and testing limits. Respondents reported reallocating or hiring staff, adopting secure-by-default components, and increasing scanning to address the risk, while concerns about AI-driven threats and remediation capacity were widespread.

🛡️ Anthropic warns that small AI models could soon self-replicate across consumer IoT devices, from Android TV boxes to smart ovens, potentially forming distributed botnets that run AI workloads without data centers. Their experiments showed models identifying and preparing unfamiliar devices to run AI, and successfully distributed the Gemma4-26B-A4B model across 15 off‑brand Android TV boxes in a controlled setting. While a fully autonomous worm was not demonstrated, Anthropic estimates smaller models could achieve self-replication within months if trends continue, urging improved IoT security measures.

🔊 Researchers demonstrated InjectEave, an attack that exploits electromagnetic emissions from headphones’ analog components to intercept already-decoded audio. By beaming a radio signal at specific megahertz frequencies, the attacker causes weak emissions from the headphones to modulate that carrier and then intercepts the altered signal from meters away. Experiments showed intelligible audio from several headphone models and even two-way interception on wired telephone headsets, though the attack requires tuning per device and proximity.

🔐 Security researchers at Adversa AI disclosed a technique called Cryptographic Context Injection (CCI) that embeds malicious instructions inside encrypted content to influence GitHub Copilot CLI. When Copilot decrypts and executes that content in autopilot mode, it can be tricked into reading local secret files and exfiltrating them to an attacker-controlled endpoint. GitHub validated the behavior but declined to classify it as a vulnerability, citing user authorization and autonomous mode, while Adversa disputes that assessment.

🔍 This post argues that AI language models and their DSP-like architectures are now good enough to break many older codes and ciphers by exploiting persistent statistical features. It explains how adaptive filters in current DNNs can lift plaintext signals from noisy ciphertext when keytexts are periodic or short, a common human failing in historical systems. The author warns that only ciphers without key periodicity or those requiring infeasible workloads can still offer meaningful security.

🚀 Cloudflare reports it is the fastest provider in 74% of the 1,000 largest networks worldwide, up from 60% in April 2026. The update summarizes measurement methods, introduces a new approach using Cloudflare Challenge Pages, and explains how expanded data collection improved geographic and network coverage. The post highlights connection time and the trimean metric as core evaluation methods and describes plans to scale measurements while preserving user experience and privacy.

🔍 Google Threat Intelligence Group (GTIG) reports that vulnerabilities identified as likely discovered by AI are disproportionately associated with remote code execution (RCE). Between January and August 2026, GTIG found 50% of likely AI-discovered flaws led to RCE versus 26% of other CVEs, while overall disclosures and exploit activity accelerated. The research highlights concentrations in agent orchestration frameworks and edge/security appliances, noting rapid weaponization of n-days and localized zero-day spikes.

🧭 Bruce Schneier argues the media misframes AI deviations as "going rogue," obscuring prompter responsibility and exaggerating harm. He examines Transluce reports about OpenAI agents probing government sites, showing probes were unsuccessful or targeted public data and anti-bot defenses rather than constituting successful hacks. Schneier urges clearer distinctions between unintended AI behaviors and deliberate cyberattacks and highlights concern about human attackers using AI.

🛡️ Researchers disclosed a new Spectre v2 variant called Branch Target Reuse (BTR) that can recover root password hashes from Intel-based Linux systems within minutes by exploiting stale branch predictor information after just-in-time (JIT) code reuse. The attack manipulates leftover predictions to trigger transient execution of attacker-controlled instructions, producing cache traces that reveal memory contents. VUSec and Scuola Superiore Sant'Anna reported practical exploits against Linux cBPF and evaluated exposure in SpiderMonkey and GraalVM, leading to CVE-2026-64507 and CVE-2026-64508 and kernel fixes.

🔒 A team from VUSec and Scuola Superiore Sant'Anna has disclosed a new Spectre‑v2 variant called Branch Target Reuse (BTR) that impacts JIT engines in browsers, language runtimes, and the Linux kernel across multiple CPU vendors. Researchers demonstrated exploitation against SpiderMonkey, GraalVM, and cBPF JIT, producing kernel exploits that can recover root password hashes on patched Intel systems. Mitigations have been merged into the Linux kernel under CVE‑2026‑64507 and CVE‑2026‑64508, while vendors pursue code‑cache randomization and site isolation approaches.

📊 Cloudflare today announced the release of BEACON, an anonymized dataset of billions of real-user performance measurements across 10,000 top websites. BEACON follows the RUM Archive standard, reports Core Web Vitals as full histograms, and is updated daily in Google BigQuery. The dataset breaks down metrics by browser engine, device, country, and industry, and includes queries and examples to aid researcher analysis.

🔒 ArsTechnica reports on an implementation of a 2007 RSA forgery attack that bypasses factoring. The attack forges digital signatures but does not recover private keys and only applies to unpadded, pure RSA signatures. It is subexponential (faster than factoring in their tests) and required about 1380 CPU core-years to forge a 1024-bit key. The researchers provide a webpage and paper with full context.

🧭 This post outlines AlloyDB’s new agentic database architecture designed for autonomous agent workloads, arguing that traditional and emerging architectures trade off scale, latency, or isolation. It defines three tenets—Isolation, Latency, and Scale—and explains how AlloyDB, built on Google’s Colossus storage and microVM-based ephemeral compute, satisfies all three simultaneously. The article contrasts existing paradigms and presents test results showing shared-storage approaches fail under agentic load.