Threat briefs

A command injection vulnerability in TRENDnet Router 1.1.02b01 allows remote, authenticated attackers to execute arbitrary commands by manipulating the wan_type parameter.

An authenticated administrator can execute arbitrary code on LibreNMS servers by injecting commands into the Signal Alert Transport configuration fields, triggering unsafe system exec calls.

This brief provides detection logic to identify potential webshell activity on Linux systems by monitoring for suspicious child processes spawned by common web server applications.

Detection of adversarial reconnaissance activities leveraging the Get-AppLockerPolicy PowerShell cmdlet to map host-based application execution restrictions.

Roundup of Oracle security advisories published in August 2026.

Adversaries may use Active Directory Service Interfaces (ADSI) within PowerShell to create local or domain accounts, effectively bypassing standard monitoring for typical user-creation commands.

The @rhinostone/swig template engine (CVE-2023-25345) contains a path traversal vulnerability in its filesystem loader, allowing unauthenticated attackers to read arbitrary local files via include or extends tags.

The RabbitMQ Java client (amqp-client) contains a pre-authentication vulnerability allowing remote attackers to trigger a StackOverflowError by sending deeply nested AMQP table structures.

The RabbitMQ Java client library is vulnerable to arbitrary class loading and static initializer execution via unvalidated input in the JSON-RPC ProcedureDescription, which can lead to remote code execution.

GeoLens versions prior to 1.2.3 contain multiple authorization bypass vulnerabilities allowing unauthenticated or low-privileged users to access private dataset metadata, schema, rows, and raster/vector tile data.

MeshCentral versions prior to 1.1.60 are vulnerable to stored cross-site scripting (XSS) due to insufficient sanitization of data fields sent from MeshAgents, allowing attackers to execute arbitrary JavaScript in administrative browser sessions.

The MONAI library contains a command injection vulnerability where unsanitized configuration values in YAML files are passed to shell execution, allowing arbitrary code execution.

The MONAI library contains a hardcoded insecure deserialization vulnerability in NumpyReader, allowing arbitrary code execution when processing malicious .npy or .npz files.

MineAdmin versions before 3.2.0-alpha.2 contain a path traversal vulnerability in the app-store plugin service allowing authenticated attackers to interact with arbitrary file system directories.

An authorization bypass vulnerability in Lemur allows authenticated users to revoke arbitrary certificates by creating duplicate certificate records and bypassing ownership and endpoint-attached safeguards.

A vulnerability in the Kobako gem allows guest mruby scripts to escape the sandbox and execute arbitrary Ruby code on the host process via improper method dispatch.

The mtdowling/jmespath.php library contains a critical code injection vulnerability, CVE-2026-54133, allowing attackers to execute arbitrary PHP code when untrusted JMESPath expressions are processed by the CompilerRuntime.

Froxlor API endpoints in the Customers, Admins, and Ftps classes leak password hashes and TOTP 2FA secret seeds, enabling offline credential cracking and MFA bypass.

Mattermost has released critical security patches for multiple versions to address vulnerabilities tracked under CVE-2026-9816.

A threat actor exploited SQL injection in a public-facing application to achieve OS-level remote code execution by abusing Oracle Java Source to deploy the custom 'khunt' post-exploitation toolkit.

An Akira ransomware affiliate gained initial access via a SonicWall VPN and attempted to evade security controls by rebooting the host into Safe Mode, an anti-EDR tactic that ultimately caused the ransomware to crash.

MacSync Stealer is a macOS-based information stealer that evades detection through rapid domain rotation while maintaining consistent behavioral pivots in its payload retrieval, C2 communication, and chunked exfiltration patterns.

This detection identifies potential prompt injection or jailbreak attempts against Amazon Bedrock Claude models by monitoring for linguistic patterns designed to override system instructions or bypass safety guardrails.

Detection of potential cross-region inference abuse in Amazon Bedrock Claude models, which may indicate attempts to bypass regional security controls or perform unauthorized data processing.

The loading of tainted Linux kernel modules may indicate the presence of rootkits or malicious persistence mechanisms used to bypass security controls and intercept system calls.

Adversaries are leveraging unauthorized device registration in Microsoft Entra ID to obtain Primary Refresh Tokens (PRT) and maintain persistent, authenticated access to cloud environments.

Medusa ransomware affiliates target critical infrastructure and healthcare entities using rapid exploitation of newly disclosed vulnerabilities and abuse of legitimate RMM software for persistence and exfiltration.

A path traversal vulnerability in Wazuh versions 4.0.0 through 4.14.5 allows unauthenticated remote attackers to trigger a denial of service by sending a specially crafted agent enrollment request.

The Submariner operator exposes long-lived service account tokens within Custom Resource specifications, allowing attackers with RBAC access to gain full control over mesh network resources.

Kraken agents are vulnerable to cache poisoning because they rely on CRC32 checksums instead of SHA-256 digests to verify peer-to-peer blobs, enabling malicious peers to inject unauthorized container images.