Threat briefs

A Reflected Cross-site Scripting (XSS) vulnerability exists in the Autodesk Fusion desktop application, triggered by a maliciously crafted HTML payload in the documentName parameter of a Fusion protocol URL.

A vulnerability exists in desmondmorris node-tesseract up to 0.2.7 where the process function in lib/tesseract.js is susceptible to OS command injection via the image argument.

Ajenti versions prior to 2.2.13 contain an improper access control vulnerability that allows an unauthenticated remote attacker to execute arbitrary code on the host server.

Unauthenticated attackers can exploit a path traversal vulnerability in the Rembg /api/remove endpoint to download malicious ONNX files, leading to arbitrary ONNX computation and file disclosure.

Better Auth is vulnerable to an authentication bypass where OAuth sign-in state can be used as a valid Magic Link token, allowing unauthenticated attackers to hijack user accounts or create unauthorized verified accounts.

The msgpack-python C extension contains a memory leak and a reference count corruption vulnerability triggered during deserialization of malformed data, enabling denial of service and potential memory safety issues.

A credential phishing campaign uses fake project bid invitations and password-protected PDFs to direct recipients to fraudulent document-sharing platforms and fake Microsoft login pages.

Multiple vulnerabilities in Firefox for iOS versions prior to 156.1, detailed in MFSA2026-117, allow for data integrity compromise and security policy bypass.

Nextcloud disclosed multiple vulnerabilities across its server platform and various applications, including Tables, Whiteboard, Files Confidential, Collectives, and Richdocuments, which could allow attackers to bypass security controls, manipulate data, or cause Denial-of-Service.

Multiple vulnerabilities in GLPI allow remote attackers to perform SQL injection, cross-site scripting, and security policy bypass, resulting in data integrity loss.

Multiple vulnerabilities in the Squid web proxy cache allow remote unauthenticated attackers to cause a denial of service or perform information disclosure.

An authenticated remote attacker can exploit a vulnerability in Moxa NPort Wireless Device Servers to cause a denial of service condition.

A vulnerability in NoMachine allows a remote, unauthenticated attacker to execute arbitrary code on the target system.

A vulnerability in the go-git library integrated into Red Hat OpenShift GitOps may allow an unauthenticated remote attacker to manipulate files or disclose sensitive information.

Multiple vulnerabilities in the Qt framework allow a remote, unauthenticated attacker to cause a denial-of-service condition and manipulate or disclose sensitive data.

Multiple unspecified vulnerabilities have been identified in Mattermost Server, requiring immediate review of vendor security bulletins and application of necessary patches.

OpenBao contains multiple undisclosed vulnerabilities that allow an attacker to bypass security controls.

CVE-2026-114148 allows local privilege escalation on Linux when an administrator extracts a malicious .NET tarball containing root-owned SUID binaries.

NVIDIA TensorRT-LLM contains an improper tokenizer loading vulnerability that could allow an attacker to achieve remote code execution, information disclosure, data tampering, or denial of service by influencing the control sphere.

Electrum versions prior to 4.8.3 fail to verify signatures of wallet-owned inputs in unconfirmed transactions, allowing malicious servers to trick users into signing fraudulent fee-replacement transactions.

hagent through commit 761eec6 is vulnerable to OS command injection via the args parameter in hagent_arduino and hagent_esp32 tools.

Kali-GPT versions through commit a608c74 are vulnerable to OS command injection in the execute_tool() function in kali_gpt/tools/mcp_server.py.

The mcp-ffmpeg package through 1.0.5 contains an OS command injection vulnerability in the resize-video, extract-audio, and get-video-info tools.

The sf-mcp package through version 1.3.2 is vulnerable to OS command injection due to insufficient sanitization of Salesforce CLI flag values in the escapeValue function, allowing remote code execution.

The ATXP Gemini CLI extension through version 1.0.0 is vulnerable to OS command injection in the runAtxp() function due to unsanitized MCP tool parameters, allowing arbitrary code execution.

DroidMind through 0.4.0 contains a command injection vulnerability that allows MCP clients to bypass the shell command allowlist by using $(...) or backtick substitution.

The ai-sdlc plugin MCP server through 0.31.2 is vulnerable to OS command injection via the repo argument, allowing execution of commands by supplying shell metacharacters.

Umbraco CMS contains an authentication bypass vulnerability involving Host header poisoning that allows unauthenticated attackers to hijack back-office sessions and execute code via Razor templates.

Multiple vulnerabilities in Apple iOS and iPadOS allow unauthorized access, privilege escalation, or system crashes via malicious files, network traffic, or web pages.

A vulnerability in Microsoft Authenticator for Android (CVE-2026-80097) allows malicious applications to impersonate trusted services and intercept sign-in tokens for victim work accounts.