Weekly Security Brief: Critical WordPress RCE, SonicWall Zero‑Days, AI Service Exploits, SharePoint Vulnerability & More

0
122

Key Takeaways

  • A pre‑authenticated WordPress core flaw (CVE‑2026-63030 + CVE‑2026-60137) enables anonymous remote code execution; attackers are already weaponising it with AI‑assisted tools.
  • SonicWall SMA VPN appliances were hit by two zero‑days (CVE‑2026-15409, CVE‑2026-15410) before patches were released, allowing full device takeover.
  • OpenSSL’s “HollowByte” DoS bug can be triggered with just 11 bytes, exhausting server memory; fixes are now in multiple OpenSSL branches.
  • CISA added a critical SharePoint deserialization RCE (CVE‑2026-58644) to its KEV catalog, confirming it was exploited as a zero‑day prior to Patch Tuesday.
  • New malware frameworks (OkoBot, NadMesh) illustrate attackers’ shift toward modular, multi‑purpose tooling that steals crypto keys, AI service credentials, and Kubernetes tokens.
  • Defensive guidance remains the same: patch urgently, validate that no backdoors remain after patching, and monitor exposed services and AI‑related assets.

WordPress Core Pre‑Auth RCE Chain

Searchlight Cyber disclosed a two‑part vulnerability affecting vanilla WordPress installations. CVE‑2026-63030 is a REST API batch‑route confusion issue, while CVE‑2026-60137 is a SQL injection in core code. Chaining them lets an unauthenticated attacker send a single request that executes arbitrary PHP code without any plugins or special configurations. watchTowr reported that proof‑of‑concept exploits are already circulating and that early signs of in‑the‑wild use are emerging. The firm urges immediate patching, followed by thorough forensic checks for backdoors that may have been dropped before the fix.

SonicWall SMA Zero‑Days Exploited Before Disclosure

Volexity linked a previously unknown threat actor, UTA0533, to the exploitation of two SonicWall Secure Mobile Access (SMA) 1000 series zero‑days. CVE‑2026-15409 (CVSS 10.0) and CVE‑2026-15410 (CVSS 7.2) were combined to achieve arbitrary command execution and full control of the VPN appliances. The attacks began on or before June 22, 2026, and were only discovered during an incident response investigation. SonicWall released patches the prior week; organizations are advised to apply them immediately and audit logs for signs of compromise.

OpenSSL “HollowByte” Denial‑of‑Service Flaw

Okta’s Red Team detailed a minuscule DoS vulnerability in OpenSSL, dubbed HollowByte. By transmitting a carefully crafted 11‑byte payload before any TLS handshake, an unauthenticated remote attacker can cause OpenSSL to reserve up to 128 KB of heap memory for a message that never arrives. Repeated abuse quickly exhausts server RAM, triggering a denial‑of‑service condition. The flaw has been patched in OpenSSL releases 4.0.1, 3.6.3, 3.5.7, 3.4.6, and 3.0.21, which now grow buffers only as actual data arrives, eliminating the memory‑amplification effect.

CISA Adds SharePoint RCE Zero‑Day to KEV Catalog

Following Microsoft’s Patch Tuesday on July 14, 2026, CISA inserted CVE‑2026-58644 (CVSS 9.8) into its Known Exploited Vulnerabilities list. The vulnerability is a deserialization flaw in SharePoint Server that permits unauthenticated remote code execution. Microsoft confirmed the bug was exploited in the wild before the patch, classifying it as a zero‑day. Federal agencies must apply the fix by July 19, 2026, and all administrators should prioritize patching and verify that no malicious artifacts remain.

OkoBot Malware Framework Targets Crypto Wallets

Kaspersky described OkoBot, an evolved version of the TookPS downloader, which delivers more than 20 malicious modules via an SSH tunnel. The framework steals cryptocurrency seed phrases by injecting implants into Trezor Suite, Ledger Wallet, and Ledger Live, logs keystrokes and clipboard data, captures screenshots, and can even harvest video streams from the victim’s application window. Initial infection uses ClickFix and GitHub‑hosted masquerading software, with a browser‑extension loader that deploys the Rilide stealer. The campaign is active across >25 countries, peaking in Brazil, Vietnam, Canada, Mexico, and Türkiye, and remains unattributed.

NadMesh Botnet Hunts Exposed AI Services

QiAnXin XLab reported a Go‑based botnet, NadMesh, that scans for publicly accessible AI services such as ComfyUI, Ollama, n8n, Open WebUI, Langflow, and Gradio. Once a target is found, NadMesh extracts AWS keys and Kubernetes tokens, establishes persistence through SSH authorized‑keys, hidden files in /dev/shm, /var/tmp, /tmp, and cron watchdogs, and exfiltrates credentials. The botnet illustrates how attackers are repurposing AI‑oriented infrastructure as a stepping stone for broader cloud‑credential theft.

Trending CVEs of the Week

The article’s “Trending CVEs” section highlighted dozens of high‑severity issues currently under active exploitation or likely to be targeted soon. Notable entries include the WordPress chain (CVE‑2026-63030 / CVE‑2026-60137), SharePoint RCE (CVE‑2026-58644), Microsoft ADFS (CVE‑2026-56155), Zoom client flaws (CVE‑2026-53412), RabbitMQ bugs (CVE‑2026-57219 / CVE‑2026-57221), n8n issues (CVE‑2026-54305), and several driver‑related vulnerabilities (e.g., tdeio64.sys, CVE‑2026-14960/14961). Defenders are urged to prioritize patching those marked “urgent” and to verify exploitability in their environments.

Cybersecurity Webinars Highlight AI Risks

Two upcoming webinars address the growing tension between AI‑assisted development and security. The first, led by Okta’s Jeremy Kirk, demonstrates how agentic AI can leak credentials, bypass safety controls, and create new attack surfaces, offering practical steps to enforce least privilege and detect shadow AI. The second webinar provides a framework for governing the explosion of code generated by AI tools, advocating secure‑by‑default controls and coordinated release processes to keep risk manageable despite accelerated development velocity.

Around the Cyber World: Notable Incidents

  • Remcos RAT Campaign: Attackers spoofed Indian GST‑related government emails to distribute the Remcos remote access trojan, aiming to harvest sensitive taxpayer data.
  • Kudankulam Nuclear Plant Leak: A ransomware attack on Reliance Infra exposed ~14.3 GB of files, including ventilation schematics for the Kudankulam plant’s Units 3 and 4, leaked via a third‑party vendor (Yotta).
  • Blind Eagle Evolution: The threat actor migrated its VBScript delivery infrastructure to Russian bullet‑proof hosting (Proton66) and upgraded its AsyncRAT with WNF process injection, custom Base28 encoding, and a Hidden VNC banking‑fraud module.
  • Qilin Ransomware EDR Killer: Qilin now employs a kernel‑level driver‑based EDR disabler that maps physical memory to kernel space, allowing direct read/write operations to bypass endpoint defenses before deploying its ransomware payload.
  • DeFiTuna Exploit: A Solana‑based AMM lost $569,601 USDC after attackers created an illiquid TUNA/USDC pool, manipulated swap calculations, and withdrew funds through fraudulent liquidity positions.
  • Next.js Security Release Cadence: Vercel announced a monthly, pre‑announced security release schedule for Next.js to improve coordination with hosting providers and allow timely mitigations for unpatched installations.
  • Disguised Gambling Apps (Brazil): Over 60 iOS apps masquerading as games become betting platforms when accessed from Brazilian IPs, leveraging public GitHub repositories that guide Cursor agents to build the fronts.
  • Ransomware Q2 2026 Stats: The Gentlemen ransomware group topped the leaderboard with 300 victims, followed closely by Qilin (289). Overall, the top 11 groups accounted for 1,368 victims across 99 countries.
  • Chinese Money‑Laundering Charges: Two New York residents were charged with laundering at least $43 million derived from cyber investment fraud, using a network of shell companies and overseas transfers.
  • CISA Uses Anthropic’s Mythos AI: CISA is piloting the Mythos language model to audit government code for defects that could be exploited by foreign spies or cybercriminals.

Cybersecurity Tools Spotlight

  • VisionSec: An open‑source, self‑hosted threat‑intelligence platform combining domain monitoring, phishing detection, exposed‑service scanning, GitHub secret discovery, breach checks, email security assessments, and Telegram alerts in a modular Docker layout. Still early‑stage, with no formal releases yet.
  • owLSM: A Linux security agent that leverages eBPF LSM to run stateful Sigma rules inside the kernel, block malicious activity, correlate events across multiple probes, and supply rich context for monitoring and incident response. The authors caution that the tool lacks a formal audit and should be tested in a sandbox before production use.

Conclusion

The week’s events underscore a familiar pattern: attackers chain simple weaknesses—exposed services, insufficient input validation, outdated drivers, and clever social engineering—to achieve outsized impact, often outpacing defenders’ patch cycles. The rise of AI‑assisted vulnerability discovery and exploit generation further shrinks the window between disclosure and abuse. Organizations must therefore prioritize rapid patching, validate that no remnants remain after updates, maintain vigilant monitoring of externally facing assets (including AI‑related services), and adopt defense‑in‑depth strategies that anticipate multi‑step attack chains. By treating every public‑facing component as potentially already tested, security teams can close gaps before they are weaponized.

SignUpSignUp form

LEAVE A REPLY

Please enter your comment!
Please enter your name here