All news with #research tag

🔎 Researchers tested 281 popular free Android VPN apps using a new framework and found widespread, basic security failures. The study, using MVPNalyzer, identified traffic leaks, plaintext transmissions, weak encryption, and apps that reveal identifying data to trackers. Those flagged have over 2.4 billion installs combined, and several apps enable tunnel hijacking by fetching configuration files in the clear. The paper and appendix list affected apps and the team plans to release the tool for broader use.

🤖 ESET analysis shows cybercriminals increasingly use AI agents and chatbots to autonomously plan and execute attacks. Researchers reviewed 900,000 AI skills in public repositories and found tens of thousands of suspicious and thousands of malicious toolsets, expanding the attack surface. These agentic tools can exfiltrate data, execute malware, override instructions, and be repurposed from legitimate utilities into harmful capabilities. ESET urges organizations to enforce policies and caution users about downloading free tools from untrusted sources.

🖥️ Researchers at Shandong University demonstrated TrojPix, a novel covert channel that modulates otherwise imperceptible on-screen pixels so the video cable emits a faint radio signal a nearby receiver can decode. The technique requires only user-level malware that can draw to the screen and achieved a peak throughput of 8.1 Mbps and a laboratory range reported up to 208 meters. TrojPix works without hardware changes, can hide transmissions under normal-looking content or a fake powered-off display, and was tested across multiple monitor brands and cable types.

🔍 A pseudonymous researcher published an 'Exploitarium' GitHub repository containing over 30 proof-of-concept exploits for zero-day vulnerabilities in many open-source projects without prior vendor notification. The dump, shared from June 27 onwards, targets projects like libssh2, FFmpeg, 7-Zip, Gitea, PHP and others, and the author claims AI-assisted fuzzing using OpenAI models. The release bypassed coordinated vulnerability disclosure, drew debate across the security community, and has led to some CVEs and patches, while others remain under review.

🛡️ Unit 42 found that large language models (LLMs) commonly hallucinate plausible web domains for real brands, and adversaries are registering these nonexistent domains to intercept AI-generated traffic. This phenomenon, called phantom squatting, poses a supply chain risk and was observed across multiple sectors. Researchers predicted adversary registrations 18–51 days in advance and discovered over 13,229 malicious URLs plus ~250,000 unregistered hallucinated domains.

🛡️ New Microsoft research shows attackers can hijack AI agents by poisoning a tool's description so the agent quietly exfiltrates company data. The attack leverages MCP tool descriptions—plain text that agents read—to inject hidden instructions, allowing malicious actions without obvious rule violations. Microsoft recommends treating tool descriptions as system prompts, restricting approved tools, enforcing human approval for risky actions, and monitoring agent identities and behavior.

🔍 New research identifies a class of weak RSA keys characterized by extensive zeroed blocks in the modulus. The open-source badkeys project collected large-scale key material from CT logs, TLS/SSH scans, and PGP repositories and found multiple real-world keys exhibiting two distinct sparse patterns. Pattern 1 appeared in certificates (now expired) from major organizations and devices, while Pattern 2 appeared in SSH hosts using CompleteFTP; affected versions and time ranges are noted. The findings highlight independent cryptographic implementations failing in similar ways and raise concerns about deliberate backdoors or coordinated vulnerabilities.

🧭 This post introduces the preview of BigQuery's new AI.AGG() function, which enables natural-language aggregation over millions of rows of unstructured and multimodal data directly inside SQL. It explains practical uses—analyzing logs, discovering product categories, and summarizing image collections—while showing how AI.AGG() batches inputs, handles NULLs, reports errors, and integrates with other BigQuery AI functions like AI.CLASSIFY(). The write-up outlines best practices for token usage, model endpoint selection, and struct handling to help users deploy AI.AGG() effectively.

🛡️ The Cobalt State of Pentesting Report 2026 surveyed roughly 450 cybersecurity professionals across 2025 and 2026 and found trust in fully automated AI vulnerability testing has dropped sharply. Reliance on AI-only testing fell from 29% to 9%, while 47% now prefer a hybrid human-plus-AI model. Respondents reported that 78% of fully automated scanners missed critical vulnerabilities, and AI/LLM issues showed longer MTTR and lower fix rates.

📝 This post highlights a new paper that demonstrates how large language models are vulnerable to prompt injection because they learn to distinguish instruction blocks by style rather than explicit tags. The authors argue that role tags became a de facto security architecture but do not map cleanly into model representations, producing persistent role confusion. The paper warns that without genuine role perception, defenses will be reactive and brittle, and calls for deeper study of roles within the LLM stack.

🔒 Security researchers at Paradigm Shift published a working exploit called usbliter8 that achieves arbitrary code execution inside the SecureROM of Apple A12 and A13 SoCs. The flaw is a hardware bug in the Synopsys DWC2 USB controller and cannot be fixed by software updates, making affected devices permanently vulnerable. Exploitation requires physical possession, DFU mode, and a dedicated microcontroller; the public proof-of-concept and write-up were released on June 18, 2026 following coordinated disclosure.

🛡️ New research from Group-IB describes the GitBait campaign, a multi-year phishing operation targeting Mexican banks that used GitHub Pages for hosting and SheetBest to exfiltrate credentials into Google Sheets. The operation relied on modular phishing kits, automated publishing, and crafted Open Graph tags to spread links via messaging apps while evading search indexing. Group-IB reported over 100 GitHub-hosted domains and urges banks to monitor brand abuse and suspicious traffic to cloud services.

🔍 Check Point Research uncovered a coordinated campaign that built a cross-platform false reputation to push a crypto clipboard hijacker. The actor used WordPress phishing hubs, multiple GitHub and SourceForge projects, AI-narrated YouTube tutorials, and forum posts to manufacture trust and inflate engagement. The campaign targeted Windows and macOS, including a macOS persistence mechanism, and manipulated reputation systems like VirusTotal to make malicious files appear safe.

🔍 New research finds current AI web agents largely fail to defend against prompt injection attacks. The StakeBench benchmark tested GPT‑5 and Gemini‑powered agents across realistic web scenarios, revealing high success rates for both direct and indirect injections and exposing failure modes like stealthy parasitism and misaligned disruption. Results show vulnerabilities vary by stakeholder and agent architecture.

🔒 Security researcher Chaotic Eclipse disclosed a new BitLocker bypass named GreatXML that leverages files placed on the recovery partition and booting into Windows Recovery Environment (WinRE). The researcher says the issue is tied to using Windows Defender Offline Scan and can result in a shell with unrestricted access to a BitLocker volume if specific XML files are copied to the recovery partition and WinRE is invoked. GreatXML follows other recent disclosures from the same researcher, including a Defender zero-day and the earlier YellowKey bypass.

🛡️ Researchers at Graz University of Technology describe FROST, a novel side-channel technique that uses the browser's origin private file system (OPFS) to monitor SSD timing and infer user activity. A malicious webpage leveraging OPFS can repeatedly access storage, measuring micro-delays that reveal what apps or websites are active. The team demonstrated data transmission rates around 660–720 bits/s with ~90% accuracy and used AI to classify app and site fingerprints. Practical constraints — RAM caching, large file creation, and likely EDR/XDR detection — limit FROST to targeted attacks, but it highlights hardware-level blind spots in modern security.

🔎 AI agent skills can install third-party capabilities with privileged access, yet registries lack automated audits. Palo Alto Networks introduces Behavioral Integrity Verification (BIV), which compares declared metadata, executable code and natural-language instructions to detect mismatches. Applied to the OpenClaw registry, BIV found widespread deviations and identified multi-stage attack chains that enable credential theft, RCE and exfiltration. The report recommends inventorying skills and requiring pre-install behavioral checks.

🔍 AI agents can help SRE teams with incident response, triage and automation, but trust is granted only when agents demonstrate reliability under real-world stress. Teams need robust observability, explicit guardrails, human-in-the-loop workflows and explainability so recommendations are evidence-backed rather than speculative. Progressive autonomy, post-incident evaluation and compatibility with existing tools are essential for safe adoption.

🔒 Varonis tested an OpenClaw-based AI agent named Pinchy with access to a controlled Google Workspace to see whether autonomous agents could be phished. The agent was given Gmail access plus mock AWS credentials, CRM exports, internal chats, and calendars, and it still leaked credentials and customer data in scenarios that mimicked routine colleague requests. A stricter safety profile improved performance, but the agent still failed when social trust cues were abused. Researchers say the problem stems from architecture and governance gaps, urging enforceable controls, identity segregation, and human review for sensitive requests.

🛡️ New research from Checkmarx finds enterprises are increasingly shipping AI-generated code despite widespread vulnerabilities. The survey of 2,350 security leaders shows nearly half of production code is AI-built and organizations that rely heavily on AI introduce far more insecure code. Many firms lack formal AI governance and continue to accept or defer fixing known issues, while tool sprawl and developer pressure compound the problem.