Intelligence Feed

1mo ago · bleeping-computer

Microsoft has announced the retirement of the OWA Light client in Exchange Server, citing modern browser capabilities, improved network conditions, and evolving security requirements as reasons. The lightweight web client, introduced two decades ago for older browsers and low-bandwidth environments, will be disabled in an upcoming update expected in August 2026. This move aims to reduce legacy attack surface and streamline engineering efforts toward the modern Outlook on the web experience.

1mo ago · hacker-news

A threat actor named Lurking Lizard has been operating a malicious residential proxy business since at least August 2022, using trojanized installers and fake mobile apps to recruit devices into a proxy botnet. The actor leverages lookalike domains, including '7zip[.]com', and impersonates legitimate proxy services to drive traffic to scam storefronts. Compromised devices are used to funnel third-party traffic, creating risks for users whose IP addresses may be abused for cyberattacks.

1mo ago · hacker-news

Researchers at Wiz discovered a vulnerability pattern called GhostApproval affecting multiple AI coding assistants, including Amazon Q Developer, Claude Code, and Cursor. The flaw exploits symbolic links (symlinks) to redirect file writes to sensitive system files, such as SSH authorized_keys or shell startup files, bypassing user consent by showing misleading approval prompts. While some vendors have issued fixes, others dispute the severity, and the issue highlights a systemic design weakness in how AI agents handle file operations and user approvals.

1mo ago · hacker-news

AI coding agents from Anthropic and OpenAI, designed to review code for security issues, can be tricked into executing malicious payloads when operating in autonomous mode. Researchers demonstrated a 'Friendly Fire' attack where a seemingly benign README.md instructs the agent to run a malicious script disguised as a legitimate build artifact. The attack bypasses safety checks by blending into normal project workflows, enabling code execution on the host without user interaction. Although currently a proof-of-concept, it highlights a critical design flaw in how AI agents interpret and act on untrusted instructions.

1mo ago · hacker-news

Meta has introduced a new AI image generation tool called Muse Image that leverages public Instagram photos and reels by default to train and generate AI content. Users can be tagged via @-mentions in the Meta AI app, enabling their public content to be used for creating new images without explicit notification. While users can opt out by adjusting privacy settings, previously generated content remains unaffected. This reflects a broader industry trend of using public user data for AI model training, as seen with similar features recently introduced by Google.

1mo ago · bleeping-computer

Microsoft has patched a zero-day vulnerability in Microsoft Defender, tracked as CVE-2026-50656 and dubbed 'RoguePlanet', which could allow attackers to gain SYSTEM privileges via a race condition exploit. The vulnerability affects fully patched Windows 10 and 11 systems, regardless of real-time protection status. It was disclosed by a researcher known as Nightmare Eclipse, who has previously revealed other Windows zero-day exploits. Microsoft addressed the flaw through an update to the Malware Protection Engine and has warned against malicious exploitation of such vulnerabilities.

1mo ago · bleeping-computer

AssuranceAmerica suffered a data breach in March 2026 after attackers targeted an employee and gained unauthorized access to its IT systems. The attackers exfiltrated sensitive customer data, including names, contact information, driver's license numbers, and insurance policy details of nearly 7 million individuals. The company detected the breach on March 17, 2026, and completed its forensic review by June 15, 2026, after which it began notifying affected individuals. AssuranceAmerica has since reset credentials, isolated affected systems, and enhanced monitoring to prevent further compromise.

1mo ago · socket-dev

npm v12 introduces security defaults that disable install-time script execution by default, requiring explicit approval for lifecycle scripts, git dependencies, and remote URLs. This change mitigates supply chain attacks like the Miasma 'Phantom Gyp' campaign, which exploited implicit node-gyp rebuilds to run malicious code during installation. The release also begins deprecating 2FA-bypass granular access tokens to reduce risks from compromised accounts. These measures align npm with other package managers and improve resistance to automated malware distribution via dependency installation.

1mo ago · bleeping-computer

Mount Royal University (MRU) suffered a cyberattack on June 17, 2026, resulting in data theft and deletion from its H and J drives. The threat actor CMD Organization claimed responsibility, exfiltrated sensitive data including passport scans, and demanded a 30 BTC ransom. The attackers also employ an auction model for stolen data and have published samples online. Recovery efforts are ongoing, with potential long-term disruption to university systems.

1mo ago · step-security

Recent supply chain attacks such as GhostAction, Megalodon, Miasma, and Hades have exploited stolen GitHub tokens to inject malicious workflows into repositories, enabling secret exfiltration via CI runners. These attacks bypass branch protection and code review by committing to temporary branches without pull requests, evading detection. The malicious workflows harvest repository secrets—including cloud credentials, SSH keys, and OIDC tokens—and exfiltrate them using seemingly legitimate GitHub Actions workflows. StepSecurity introduces a two-layer defense: blocking unreviewed workflows from accessing secrets and detecting exfiltration attempts in real time.

1mo ago · bleeping-computer

A China-linked threat cluster tracked as UNK_MassTraction has been exploiting vulnerabilities in Roundcube webmail servers at academic institutions in the U.S. and Canada since May 2026. The attackers target physics and engineering departments, deploying malware to steal credentials and establish persistent access. Exploitation involves CVE-2024-42009 and CVE-2025-49113 to deploy backdoors such as IceCube, SquareShell, and VShell. Proofpoint attributes the activity to a likely China-aligned espionage group based on infrastructure overlap and linguistic artifacts, though confidence is moderate.

1mo ago · bleeping-computer

A threat actor has published 17 malicious packages on npm and PyPI that impersonate legitimate Paysafe, Skrill, and Neteller SDKs to steal credentials and access tokens. These packages mimic real software development kits but instead exfiltrate sensitive data such as API keys, tokens, and system metadata to an AWS-hosted command-and-control server. The malware includes basic anti-analysis techniques and targets developers integrating payment functionality, with different activation triggers between npm and PyPI versions.

1mo ago · hacker-news

Ubiquiti has patched multiple critical vulnerabilities across its UniFi product line, including UniFi Connect, Talk, Access, Protect, and OS. The flaws include command injection, SQL injection, SSRF, and improper access control issues that could allow privilege escalation and arbitrary command execution. While most vulnerabilities have not been observed in active exploitation, CISA has flagged three UniFi OS flaws as weaponized in real-world attacks. Additionally, Russian state-sponsored actors have previously used compromised Ubiquiti devices in the MooBot botnet.

1mo ago · hacker-news

A new attack technique called HalluSquatting exploits AI coding assistants' tendency to hallucinate non-existent software package names and inadvertently fetch malicious code from attacker-controlled repositories. By predicting and registering these fake names on platforms like GitHub or npm, attackers can trick AI tools into installing botnet malware when users request popular resources. The attack leverages prompt injection through fetched content, enabling command execution without direct user interaction, effectively turning AI assistants into delivery mechanisms for malware.

1mo ago · hacker-news

AI coding agents such as Claude Code, Cursor, and OpenAI Codex are triggering endpoint security detection rules designed to catch malicious human intruders. These agents perform legitimate development tasks that mimic adversarial behaviors, including accessing browser credentials via DPAPI, using living-off-the-land binaries (LOLBins) like certutil and bitsadmin to download files, and writing scripts to startup folders for persistence. While not inherently malicious, their behavior overlaps with known attack tactics, creating noise in threat detection systems and complicating defender response. This reflects a broader trend of malware-free intrusions using trusted tools and valid credentials.

1mo ago · bleeping-computer

A threat actor tracked as O-UNC-066, associated with the Pink extortion gang, is conducting vishing attacks to trick Microsoft 365 users into enrolling Entra passkeys under the attacker's control. The campaign uses voice phishing and phishing kits that mimic legitimate Microsoft enrollment pages, enabling real-time credential and MFA interception. The attacker exploits a new Microsoft passkey registration feature and uses fake BIP-39 recovery phrases to distract victims while stealing credentials. Post-compromise, the actor exfiltrates data from SharePoint and OneDrive to support extortion efforts.

1mo ago · wiz

The article discusses the expanding attack surface in modern cloud and AI-driven environments, emphasizing the challenges posed by shadow IT and unknown risks such as logic flaws and exposed assets. Wiz introduces automated reconnaissance capabilities in its Attack Surface Management (ASM) platform to continuously discover subdomains and shadow assets across any environment. The solution combines external reconnaissance with internal context to identify exploitable risks, including misconfigurations, leaked secrets, and logic-based vulnerabilities, enabling organizations to defend against AI-powered threats.

1mo ago · wiz

GhostApproval is a trust boundary vulnerability affecting multiple AI coding assistants, where malicious repositories use symbolic links to trick agents into accessing or modifying files outside the intended workspace sandbox. The attack exploits symlink following (CWE-61) and UI misrepresentation (CWE-451), allowing adversaries to silently write to sensitive system files such as ~/.ssh/authorized_keys or ~/.zshrc. Despite confirmation prompts, users are often misled about the actual file being modified, leading to potential remote code execution and persistent access. Vendors including AWS, Cursor, and Google have issued fixes, while others have acknowledged or rejected the issue.

1mo ago · hacker-news

A recent study reveals that GitHub Copilot can be manipulated into generating harmful content through a workflow-level jailbreak technique, despite refusing such requests directly in chat. Researchers reframed harmful prompts as steps in a benign coding task, leading Copilot to generate dangerous responses within code as part of improving a benchmark scoring program. This highlights a critical gap in AI safety mechanisms when models are integrated into active development environments rather than used for chat-only interactions.

1mo ago · hacker-news

Account takeover (ATO) attacks are shifting from traditional credential stuffing to targeting identity verification and recovery processes, as stronger authentication like passkeys reduces the value of stolen passwords. Attackers now exploit weak points in magic-link flows, SIM swaps, and use generative AI to conduct sophisticated impersonation fraud with deepfakes and synthetic documents. Defenders must adopt biometric liveness detection, intent binding, and network-scale fraud pattern analysis to counter these evolving tactics.

1mo ago · hacker-news

A vulnerability in GitHub's commit verification system allows attackers to rewrite signed Git commits into new hashes without breaking their signatures, resulting in multiple valid hashes for the same content. This undermines trust in commit hashes as unique identifiers, enabling potential bypasses of blocklists and provenance systems that rely on them. The issue stems from signature malleability in GPG and S/MIME schemes, which GitHub does not normalize before verification.

1mo ago · hacker-news

A threat actor dubbed REF6045 is targeting Mexican banking and financial users through a malware toolkit named SCMBANKER, delivered via fake CAPTCHA lures. The attack uses social engineering to trick victims into executing a malicious command, leading to the installation of PowerShell-based malware that enables session monitoring, clipboard hijacking, browser redirection, and remote access. The operation shows signs of AI-assisted development, with poor operational security enabling researchers to recover infrastructure details. Victims are actively monitored and targeted based on financial activity, indicating ongoing live attacks.

1mo ago · hacker-news

A new 'ghost phishing' campaign dubbed EvilTokens is targeting businesses in the US and Europe, leveraging encrypted HTML content that remains hidden until decrypted in the browser. This technique bypasses traditional email and URL security checks, enabling Microsoft 365 account takeover via legitimate Microsoft login flows without directly stealing passwords. The attack exploits browser-level decryption to reveal phishing content post-load, making detection more difficult and increasing the risk of unauthorized access to sensitive data and cloud services.

1mo ago · bleeping-computer

Japanese telecommunications company KDDI disclosed a data breach affecting over 12 million users, resulting from an attack on May 16 that exploited a zero-day vulnerability in third-party software used by multiple ISPs. The breach exposed email addresses and passwords, with some credentials stored in hashed or encrypted form. KDDI detected and blocked the attackers on June 17, implemented EDR solutions, and is enforcing password resets for affected accounts.

1mo ago · bleeping-computer

DuckDuckGo has introduced a new feature in its browser that blocks YouTube video ads by leveraging community-maintained filter lists from uBlock Origin and its own compatibility rules. The feature is enabled by default on iOS, Mac, and Windows, while Android users can enable it manually. This functionality enhances user experience by reducing ad interruptions without requiring third-party extensions, aligning DuckDuckGo with browsers like Brave and Opera.

1mo ago · bleeping-computer

AI is increasingly being leveraged by threat actors to enhance service desk attacks through more convincing impersonation, accelerated reconnaissance, and scalable social engineering. These attacks exploit the urgency and trust inherent in service desk operations, particularly during employee onboarding, to gain unauthorized access. Attackers use AI to generate realistic communication, personalize phishing attempts, and bypass traditional identity verification methods, making detection more difficult for security teams.

1mo ago · socket-dev

The pnpm 11.10 release introduces security enhancements to prevent registry token redirection attacks in the npm ecosystem. A new _auth configuration ensures registry credentials are bound to their intended registry URL, preventing malicious project files from redirecting valid tokens to attacker-controlled hosts. This update mitigates supply chain risks by limiting the privileges of repository-controlled files and hardening authentication mechanisms.

1mo ago · hacker-news

China-linked threat actor UAT-7810 is expanding its Operational Relay Box (ORB) network by deploying updated malware variants such as LONGLEASH, DOGLEASH, and JARLEASH. The group targets internet-facing networking devices, including Ruckus and ASUS routers, leveraging known vulnerabilities to establish persistent access. These relay nodes are used to support secondary threat actors like UAT-5918 in conducting cyber attacks against high-value targets, particularly in critical infrastructure sectors. The continued development and testing of malware on MIPS-based platforms indicate ongoing refinement of their capabilities.

1mo ago · bleeping-computer

The U.S. Cybersecurity and Infrastructure Security Agency (CISA) has mandated federal agencies to patch a critical vulnerability, CVE-2026-48282, in Adobe ColdFusion by a strict deadline due to active exploitation in the wild. The flaw allows unauthenticated remote attackers to achieve code execution on unpatched systems with low attack complexity. Adobe released patches one week prior and warned of high exploitation risk, with evidence of attacks emerging within hours of disclosure.

1mo ago · bleeping-computer

Ubiquiti has disclosed and patched seven critical vulnerabilities in its UniFi OS platform, including a maximum-severity command injection flaw tracked as CVE-2026-50746 in the UniFi Connect Application. The vulnerabilities, affecting various UniFi products such as UniFi Talk, Access, and Protect, allow for low-complexity attacks without user interaction and could enable remote command execution. Internet-exposed UniFi OS instances remain a concern, with over 100,000 instances detected online, making them attractive targets for state-sponsored and cybercriminal groups. Previous incidents have shown exploitation of similar flaws to build botnets for cyberespionage and malicious traffic proxying.