All news in category “Security Advisory and Patch Watch”

🚨 The U.S. Cybersecurity and Infrastructure Security Agency has issued an urgent directive requiring federal agencies to mitigate actively exploited vulnerabilities in IBM Langflow, N-central, and Apache Tomcat within three days. The most severe issue, tracked as CVE-2026-9198, impacts Langflow and permits unauthenticated remote code execution via chained API endpoints. Vendors have released patches and a hotfix, but incomplete fixes and public proof-of-concept exploits have enabled ongoing attacks. CISA added all three flaws to its Known Exploited Vulnerabilities catalog and urged immediate remediation.

🛡️ Two critical vulnerabilities in the open-source AI control plane Paperclip allow attackers to import a malicious agent and trigger command execution on either network-accessible servers or local developer machines; a third flaw exposes sensitive control-plane details via inadequately guarded API routes. Vendors have released fixes in the source tagged v2026.416.0, which enforces stricter import permissions and hostname validation, and operators are urged to upgrade and review deployment exposure.

🔒 New research from Oasis Security disclosed three vulnerabilities in Paperclip, an open-source AI agent orchestration control plane, that exposed sensitive data and allowed unauthenticated command execution on servers and developer machines. Two issues were rated critical and one carried a CVSS score of 10.0. The flaws include self-registration and CLI authorization weaknesses, missing access checks, and a DNS rebinding risk in local development mode, all of which have been patched.

🔒 Vendors HashiCorp, Veeam, and the Django Software Foundation have released fixes for 11 vulnerabilities affecting Terraform MCP Server, Veeam Service Provider Console, and Django. The most severe include an unauthenticated credential-exposure bug in Veeam (9.5), a cross-tenant token-reuse issue in Terraform MCP (10.0), and a GeoDjango spatial lookup flaw that can write files or trigger code execution. Operators are advised to upgrade to Terraform MCP Server 1.1.0+, Veeam 9.3.0.35057, and Django 6.0.8 / 5.2.17; exposure depends on configuration and none of the flaws show public exploitation as of August 5, 2026.

🛡️ Security researchers disclosed multiple vulnerabilities in the open-source AI agent platform Paperclip, including an authorization bypass, exposed APIs, and a DNS rebinding weakness that could lead to remote code execution and developer-machine compromise. Oasis Security detailed how default registration and import behaviors allowed attackers to escalate privileges and execute arbitrary commands by uploading malicious agent configurations. Patches were released in versions 2026.416.0 and 0.3.1 to harden authorization, validate hostnames, and restrict risky imports.

🔒 A memory corruption vulnerability in the Linux kernel's Open vSwitch datapath, tracked as CVE-2026-64531 and dubbed OVSwrap, allows ordinary local users to escalate to root on many default-configured distributions. Disclosed on July 28, 2026 by Asim Manizada, the bug exploits a 16-bit length wrap in nested Netlink attributes and ships with a public PoC containing records for ~800 kernel builds. Upstream fixes were released on July 24; mitigations include blocking the openvswitch module, unloading it, rebooting, or disabling unprivileged user namespaces.

🔒 An unauthenticated attacker could read any file the Gitea service account can access in versions 1.22.1–1.27.0 by posting crafted Org-mode markup to the markup endpoint. The issue, tracked as CVE-2026-59774 and rated Critical (CVSS 9.8), was fixed in Gitea 1.27.1. Self-hosted admins should upgrade immediately and rotate exposed credentials if the endpoint was reached.

🛡️ CISA on August 5, 2026, added three actively exploited flaws to its Known Exploited Vulnerabilities (KEV) catalog, including a critical Langflow RCE (CVE-2026-9198) and an Apache Tomcat encryption bypass (CVE-2026-34486). The advisory also includes an N-able N-central authentication bypass (CVE-2026-18556) and a related incomplete fix tracked as CVE-2026-18577. Agencies must apply available patches and mitigations promptly to prevent ongoing exploitation.

🛡️ A critical CVE-2026-66066 in Ruby on Rails’ Active Storage can let unauthenticated attackers read sensitive files or escalate to RCE by abusing image processing via libvips. Fixed in Active Storage versions 7.2.3.2, 8.0.5.1 and 8.1.3.1, the flaw affects apps that accept untrusted uploads and use libvips; admins should update Rails, ensure libvips ≥ 8.13, rotate secret_key_base, and audit uploads and logs.

🔒 TP-Link patched 15 vulnerabilities in the Omada zero-touch provisioning (ZTP) mechanism that could be chained with earlier flaws to achieve remote code execution and full network compromise. Forescout’s Vedere Labs disclosed the issues at Black Hat USA, noting impacts across Omada controllers, gateways, switches, access points, cloud services, mobile apps, and various TP-Link devices. The flaws include hard-coded keys, information disclosure, device hijacking, client-side code execution, and interception of encrypted communications. Administrators are urged to apply firmware updates, use strong unique credentials, enable MFA, rotate secrets if compromise is suspected, and monitor for suspicious activity.

🔒 cPanel issued a targeted security release addressing a database privilege escalation flaw (CVE-2026-58048) that allowed an authenticated cPanel account with MySQL/MariaDB access to execute SQL in the administrative database context. The update also patches an HTTP request-smuggling issue in cpsrvd (CVE-2026-58047) and multiple Exim vulnerabilities; temporary workarounds are available for systems that cannot immediately upgrade. Administrators should apply the listed builds or revoke MySQL access until patched.

🔒 CISA added a high‑severity vulnerability affecting N‑able N‑central to its Known Exploited Vulnerabilities (KEV) list after reports of active exploitation. Tracked as CVE-2026-18577, the flaw is an incomplete patch for CVE-2026-18556 and permits authentication bypass and account takeover; it is fixed in version 2026.3 HF1. N‑able and researchers note indicators such as a malicious "svchost.exe" in user documents, a service named "Cloudflared," and inbound connections from several VPN exit node IPs linked to Mullvad and NordVPN.

🔒 Security researchers from Palo Alto Networks' Unit 42 disclosed three related attacks, collectively dubbed "Pass-ta-key," that let malware on compromised Windows devices abuse Google Password Manager's synced passkeys in Chrome on TPM-equipped machines. The techniques — Pass-ta-key, Silver Pass-ta-key, and Golden Pass-ta-key — exploit weaknesses in device trust, onboarding, recovery, and synced credential handling rather than breaking passkey cryptography. While the attacks require existing malware on the victim's device, they can bypass or subvert user verification and even extract the master key that encrypts synced passkeys, enabling account takeover and future key decryption. Unit 42 reported findings to Google and affected services; some issues, such as eBay's validation, have been fixed.

🔒 N‑able has issued a hotfix (2026.3.1.7) after detecting active exploitation of an authentication bypass vulnerability, CVE-2026-18577, affecting hosted and on-premises N-central servers. The vendor disclosed the issue on August 1 and released an update the following day, urging immediate upgrade to versions 2026.3 or later. Hosted deployments were updated automatically; on-premises customers must install the patch manually. Indicators of compromise and mitigation guidance are available on the hotfix download page.

🔒 Unit 42 describes three post-compromise attacks against Chrome's Google Password Manager cloud authenticator—Pass-ta-key, Silver Pass-ta-key and Golden Pass-ta-key—that let malware on Windows obtain valid authentication assertions or extract the master secret without user interaction. The techniques exploit how Chrome stores and reloads TPM-wrapped keys, allows deferred user-verification key creation during re-enrollment, and exposes the 32-byte Security Domain Secret (SDS) in process memory. The research is limited to Windows with TPM and starts from a compromised endpoint; it does not claim cryptographic failure and has no CVEs listed as of August 3, 2026.

🔒 Thermo Fisher Scientific issued a July 31 security bulletin addressing a vulnerability in select Applied Biosystems human identification software that could allow .fsa and .hid files to be modified before analysis. The issue is tracked as CVE-2026-17583 with a High severity (CVSS v4.0 8.2). Five supported product lines received updates that add digital signatures, while three end-of-life products will receive no fixes. The vendor urges customers to install updates or follow recommended controls for file custody, access, and network restrictions.

🛡️ Three high‑severity vulnerabilities in Hugging Face's Diffusers library, collectively named FaceHugger, can let crafted model repositories execute arbitrary code on machines that load them. Zafran Labs attributes the issues to TOCTOU race conditions that bypass the trust_remote_code safeguard during model loading. The flaws were fixed in Diffusers 0.38.0; users are advised to patch or follow recommended mitigation steps when loading custom pipelines.

🛡️ The Rails project patched a critical Active Storage vulnerability (CVE-2026-66066) that can let unauthenticated attackers read arbitrary files and potentially achieve remote code execution when libvips is used. The flaw affects multiple Rails branches before specified patch releases and requires accepting uploads from untrusted users. Administrators should upgrade libvips to 8.13+, apply Rails updates, and rotate exposed secrets. ImageMagick users are not affected by this vector.

🛡️ Adobe released updates for Campaign Classic (ACC) to patch a maximum-severity authorization vulnerability (CVE-2026-48449, CVSS 10.0) that could enable arbitrary code execution without user interaction. The fixes, delivered in ACC v7.4.3 build 9398 for Windows and Linux, also address a high-severity SQL injection (CVE-2026-48448, CVSS 8.6) enabling arbitrary file reads. Adobe additionally remediated eight critical-rated flaws in Adobe Bridge that could lead to privilege escalation and code execution, crediting multiple external researchers. Users are urged to apply the updates promptly for protection.

🔒 Broadcom has released patches addressing five vulnerabilities across multiple VMware products, with three rated as critical. Affected products include vCenter, ESX, Workstation, Fusion, and various cloud and telco platforms. Issues range from authentication bypass and out-of-bounds write enabling remote code execution to syslog-related arbitrary code execution. Administrators should apply fixes from Broadcom's advisory promptly.