Briefs

An anonymous user attempts to enumerate Kubernetes API endpoints, resulting in a series of failed API requests across multiple endpoints, potentially revealing the cluster's exposed surface.

Detection of interactive access to a Kubernetes service account token or certificate followed by a Kubernetes API request, potentially indicating credential theft and lateral movement within the cluster.

Adversaries attempt to clear Linux system logs using utilities like rm, rmdir, shred, and unlink to conceal malicious activity and evade detection.

A SQL injection vulnerability (CVE-2026-7060) exists in liyupi yu-picture versions up to a053632c41340152bf75b66b3c543d129123d8ec, allowing a remote attacker to execute arbitrary SQL commands by manipulating the sortField argument in the PageRequest function of PictureServiceImpl.java.

A heap-based buffer overflow vulnerability exists in llama.cpp due to an integer overflow in the `ggml_nbytes` function, allowing attackers to potentially achieve Remote Code Execution (RCE) by crafting malicious GGUF files.

Attackers may disable Protected Process Light (PPL) protection for the LSASS process by modifying specific registry keys, allowing for credential dumping and other malicious activities.

The lupa library's attribute_filter is not consistently applied when attributes are accessed through built-in functions like getattr and setattr, leading to a sandbox escape and arbitrary code execution.

An unauthenticated Server-Side Request Forgery (SSRF) vulnerability in MagicMirror² allows remote attackers to force the server to perform arbitrary HTTP requests, exfiltrate environment variables, and potentially compromise cloud instances or internal networks.

MsBuild.exe making outbound network connections may indicate adversarial activity as attackers leverage MsBuild to execute code and evade detection.

A vulnerability exists in NATS servers configured to accept leafnode connections where a malicious remote NATS server can trigger a server panic by exploiting compression negotiation on the leafnode port.

Network-AI is vulnerable to missing authentication on the MCP HTTP endpoint, allowing unauthenticated privileged tool calls that could lead to configuration changes and agent manipulation.

A stored same-origin XSS vulnerability in Note Mark allows an authenticated user to upload malicious HTML, SVG, or XHTML files as note assets, leading to arbitrary code execution in a victim's browser due to missing content type restrictions and resulting in potential access to private notes, books, and authenticated API actions.

OAuth2 Proxy versions 7.5.0 through 7.15.1 are vulnerable to an authentication bypass (CVE-2026-41059) due to improper handling of URL fragments in conjunction with `skip_auth_routes` or `skip_auth_regex`, potentially allowing unauthenticated access to protected resources.

A machine learning job has identified an unusual spike in Okta group lifecycle change events, indicating potential privilege escalation activity, where adversaries may be altering group structures to escalate privileges, maintain persistence, or facilitate lateral movement within an organization’s identity management system.

This analytic identifies instances where the same Okta user logs in from different cities within a 24-hour period, potentially indicating a compromised account and leading to account takeovers and data breaches.

This detection identifies unusual network patterns and connection problems within Ollama, encompassing unauthorized API access attempts beyond localhost and warning-level network errors like DNS lookup failures, TCP connection issues, or host resolution problems, which can signal network-based attacks, unauthorized access, or infrastructure reconnaissance.

A vulnerability in OpenClaw versions 2026.4.21 and earlier allows a non-owner loopback client to spoof the owner context by manipulating request headers, potentially gaining unauthorized access to owner-gated operations.

OpenMage LTS version 20.16.0 and earlier has a critical vulnerability in the XML-RPC/SOAP API session ID generation, which uses a predictable MD5 hash of time-derived inputs, allowing attackers to brute-force and hijack active API sessions for data exfiltration, order fraud, and supply chain manipulation.

The OpenRemote IoT platform is vulnerable to expression injection, allowing remote code execution due to an unsandboxed Nashorn JavaScript engine and an inactive Groovy sandbox, leading to full server compromise.

A remote code execution vulnerability exists in OpenTelemetry Java agent versions prior to 2.26.1 due to unsafe deserialization in the RMI instrumentation, potentially allowing attackers with network access to execute arbitrary code on vulnerable systems.

Oxia exposes the full bearer token, including JWT header, payload, and signature, in debug log messages when OIDC authentication fails, allowing attackers with log access to replay the tokens.

A race condition in Oxia's session heartbeat handling allows a remote client to trigger a denial-of-service by sending rapid KeepAlive requests during session expiration or closure, leading to a server crash.

Pachno 1.0.6 is vulnerable to stored cross-site scripting (XSS), allowing attackers to inject malicious HTML or scripts into POST parameters, which are then stored and executed in user browser sessions due to improper sanitization.

An unauthenticated attacker can achieve remote code execution on Paperclip instances by exploiting multiple vulnerabilities, including open signup, self-approval of CLI authentication challenges, and missing authorization checks in the company import endpoint, leading to arbitrary command execution as the server's OS user.

This detection rule identifies attempts to establish persistence on Windows systems by creating scheduled jobs in the Windows Tasks directory, excluding known legitimate jobs.

Unauthenticated SQL injection vulnerability exists in phpMyFAQ <= 4.1.1 due to improper handling of the User-Agent header in BuiltinCaptcha, allowing attackers to inject malicious SQL payloads and potentially gain complete control of the datastore.

This brief covers the abuse of application shimming in Windows via `sdbinst.exe` to achieve persistence and privilege escalation by executing arbitrary code within legitimate processes.

Adversaries may use recursive grep commands on Linux or macOS to discover credentials, secrets, keys, or tokens within files, indicating potential credential access and data exfiltration attempts.

The execution of a Linux shell process from a Java JAR application following an incoming network connection may indicate reverse shell activity.

PraisonAI Browser Server is vulnerable to unauthenticated WebSocket client hijacking due to exposing the browser bridge on 0.0.0.0 by default and accepting WebSocket clients that omit the Origin header, allowing unauthorized remote use of a connected browser automation session.