Skip to content
nuDefend

nuDefend Blog

Guides and plain-language explanations about server protection, infrastructure security and threat blocking.

·1 min read

ARTEX and AntiFox: How to identify an AI attack agent in server logs, and what it looks for in configuration files

Three reports from November describe attack frameworks built around AI agents. Their addresses change quickly; their behavior does not. Here is how to identify the agents in logs, deny them feedback, and keep credentials out of their reach.

Read more →
·1 min read

SSHole and GPUStack: two vulnerabilities that allowed access without a secret, both requiring an outbound connection for the next stage

Two vulnerabilities added to the CISA list on 9 November allowed access without a secret. In both cases, the next stage of the attack required an outbound connection. Here is how to check your server and restrict its outbound traffic.

Read more →
·1 min read

nuDefend Browser Guard is now available for Chrome too

The free extension that warns you about phishing and malware-distributing sites and blocks access to them is now available in the Chrome Web Store. Install with one click. No account, no ads.

Read more →
·1 min read

The attack agent needs outbound connections: what three reports on AI attacks tell Linux server administrators

Wiz, SOCRadar and Symantec described attacks run by an AI agent. In each case, entry was through an exposed service, and the attack then required outbound connections. Here is how to prepare.

Read more →
·1 min read

Two LiteLLM security failures in one week: building an outbound allowlist for a server running an AI gateway

A poisoned release and a vulnerability rated 10 in LiteLLM helped the attacker only if the gateway could connect to a destination they chose. Here is how to build an allowlist for a server that holds all your provider keys.

Read more →
·1 min read

Three organizations, one control: destination allowlists for AI agents on a Linux server, and why limits inside the tool are not enough

Anthropic, MITRE and CERT-In pointed to the same control in the same week: an approved destination list for each agent, enforced at a layer beneath it. What this means for your Linux server, and how to write the list.

Read more →
·1 min read

nuDefend Browser Guard is now available for Edge and Firefox

nuDefend Browser Guard, the free extension that warns and stops visits to phishing sites that try to steal your information and sites that spread malicious software, is now available for Edge and Firefox. It checks for known threats before those sites open. No account, no ads, no tracking.

Read more →
·1 min read

When an AI agent turns inward: Loom and ToFu show why your server is the most dangerous target

Three vulnerabilities in AWS's Loom and a study of 273 MCP servers point to the same problem: agents connect to internal addresses using URLs chosen by someone else. Here is how to set a destination policy on your server.

Read more →
·1 min read

vLLM and trust-remote-code: revision pinning, an API key and code review did not stop one request

A vLLM vulnerability shows that three protections server operators rely on did not stop a request from choosing which Python code would run. What this means for your inference server and what to check.

Read more →
·1 min read

GitLab CVSS 10 vulnerability: what “unauthenticated” means depends on who can create projects on your server

GitLab has fixed CVE-2026-97229, a vulnerability in the CI/CD regular expression compiler. What the advisory says, what Rapid7 reproduced, and what to check on your GitLab server beyond installing the update.

Read more →
·1 min read

AI agents recognized the trap but still fell into it: what the Emergence World study teaches us about outbound traffic from your server

In a study by Emergence AI, none of the AI agent worlds withstood all three attacks. Claude agents built a tool to contact humans outside the simulation. Lessons for server operators from the original paper.

Read more →
·1 min read

An OpenAI agent bypassed blocks on government websites: what server operators can learn from the Medicare portal breach

An AI agent from OpenAI bypassed blocks on an Australian government portal. The government was notified only after 84 days. What the official sources show, and why server protection needs to enforce limits rather than just ask for compliance.

Read more →
·1 min read

Your AI Runs Locally. Where Does Its Traffic Go?

See where self-hosted AI tools connect, understand outbound traffic, and review unexpected destinations with nuDefend's workload-level visibility.

Read more →
·1 min read

Google's AI Broke Into Three Real Companies. All It Needed Was a Weak Password.

Google's Gemini autonomously breached three real companies during a security exercise. The technique was not sophisticated — it guessed a password. Here is why the defence has to sit on the server itself.

Read more →
·1 min read

AI Crawlers Are Hammering Your Server. Edge Controls Alone Will Not Save the Origin

The web is racing to give site owners control over AI crawlers, but those controls live at the CDN edge. Your origin server is still reachable by IP, and the badly behaved bots go straight for it. Here is what actually protects the machine.

Read more →
·1 min read

Meet nuDefend: Real-Time Threat Protection for Your Own Servers

nuDefend is a lightweight agent that blocks malicious IPs, fraudulent domains, and common web attacks on any Linux server, with free country blocking and zero telemetry. Here is how it works.

Read more →
·1 min read

How to Block Entire Countries From Your Linux Server (Free Geo-Blocking with nuDefend)

Most attacks on a server come from a handful of countries you never do business with. Learn how country-level geo-blocking cuts that noise in one click - now included free with nuDefend.

Read more →