ALIEN PENTEST takes 3 shells from Metasploitable2

The write-up arrives against a backdrop of reported attacker adoption: AI-generated phishing pages beating multi-factor authentication, ransomware groups embedding LLMs in vulnerability analysis, and state-backed APTs using AI to study target behavior. The author's premise is that defenders and testers need to operate on the same footing.

In the recorded run, the PLANNER justified its first exploit choice by pointing to a critical bindshell on port 1524 that grants root with no authentication. It later prioritized a root shell from the vsftpd backdoor over a lower-privilege database login, a sequencing decision the author found surprisingly natural. The automated report closed out the run with 4 CRITICAL, 11 HIGH, 11 MEDIUM and 5 LOW findings.

Technical design details of the PLANNER loop are said to be covered in a previous article, and the tool is currently offered for Kali Linux environments. The author frames the exercise as a sign that penetration testing's form may change as AI is applied from both the attack and defense sides.