SEC535: Offensive AI - Attack Tools and Techniques™

Overview

The first day of SEC535 focuses on AI-driven reconnaissance and social engineering attacks, equipping you with the tools and techniques modern adversaries use to infiltrate organizations. We kick off with OSINT gathering using AI, leveraging powerful tools like Spiderfoot and Bbot to uncover valuable intelligence such as DNS records, employee emails, and internal phone numbers at the notional company “Meridian Systems.” From there, we explore how an AI Pentest Assistant powered by a Retrieval-Augmented Generation (RAG) database can streamline network enumeration and optimize vulnerability discovery to enhance penetration testing workflows.

Armed with this intelligence, we will transition to targeting the human attack surface by utilizing AI to craft hyper-personalized spear phishing campaigns and deploy deepfake-powered vishing attacks against employees of Meridian Systems. By the end of the day, you’ll have first-hand experience in weaponizing AI for reconnaissance and social engineering, gaining the skills necessary to outthink and outmaneuver real-world attackers.

Exercises

• AI Powered Recon: OSINT for Penetration Testing
• Using AI for Network Discovery
• Building PhishGPT
• Lure Lab: Phishing Mini-Range
• Audio Deepfake Development
• AI Dialed Deception: Vishing Mini-Range

Topics

• AI Pentest Assistants
  • Training Concepts
  • Assistant Architectures
  • The Angry Beaver Pentest Assistant
  • Shell GPT
• Gathering Open-Source Intelligence using AI
  • Identifying Key Targets
  • Using the Intelligence Cycle
  • Active vs Passive OSINT
  • What are Google Dorks, and Dorking with AI
  • Using Spiderfoot for Reconnaissance
  • Bbot Reimagined using AI
• Network Reconnaissance and Enumeration using AI
  • AI Powered Network Enumeration
  • Automated Vulnerability Prioritization using RAG databases
  • Adaptive Scanning and AI – Driven Enumeration Strategies
  • Intelligent Data Correlation for Asset Discovery
• Introduction to Social Engineering
  • Social Engineering Attack Surface
  • Psychology of Social Engineering
  • The Role of Social Engineering in Penetration Testing
• Creating AI Powered Phishing Emails
  • Tips for Writing Prompts
  • Why Limitations are Important
  • Phishing GPT
• Audio Deepfakes
  • Voice.ai
  • Speech Concepts
  • Components of Speech
  • Types of Audio Deepfake Technology
  • How Attackers leverage Audio Deepfakes
  • Case Study: CEO Fraud
• Visual Deepfakes
  • Face Swapping
  • Motion Transfer
  • Image-to-Image Translation
  • Lip Syncing and Audio Matching
  • Case Study: Finance Workers
  • HeyGen
  • Deep Live Cam

Overview

In this section we will explore how AI is redefining traditional exploitation techniques, from automating Metasploit workflows to uncovering vulnerabilities in web applications, memory corruption, and patch analysis.

We begin by reimagining Metasploit interactions by using AI to automate exploit selection, improve post-exploitation tactics, and rethink how we use persistence mechanisms. By integrating AI-driven Retrieval-Augmented Generation (RAG) architectures, we enhance vulnerability exploitation, allowing for real-time adaptation based on target responses.

Next, we dive into web-based exploitation, leveraging AI tools like ShellGPT to automate SQL injection, we will perform AI-assisted session hijacking, and use a GPT for automated web code analysis to identify and exploit vulnerabilities with shocking accuracy.

From web applications, we transition into memory corruption and stack smashing using AI-powered analysis to identify buffer overflow vulnerabilities, craft exploits, and evaluate modern mitigations like stack canaries, ASLR, and DEP.

Finally, we introduce AI-automated patch diffing, where AI-driven agents perform binary comparisons to detect security fixes, rank exploitability, and generate working exploits before vulnerabilities are publicly disclosed.

Exercises

• Metasploit Reengineered
• AInjection: Using AI for SQL Exploits
• SegfaultGPT: Writing Buffer Overflows with AI
• Patch Diffing with AI

Topics

• Reimagining Metasploit with Artificial Intelligence
  • Automated Exploit Selection
  • Post Exploitation and Persistence
  • RAG Architectures and Vulnerability Exploitation
• Web Based Exploits
  • Using ShellGPT for SQL Injection
  • Session Hijacking with AI
  • Web Based Code Analysis with AI
• Corrupting Memory and Smashing Stacks
  • Identifying Stack-Based Buffer Overflows with AI
  • Writing Buffer Overflows with AI
  • Evaluating Stack Canaries, ASLR, and DEP
• Automating Patch Diffing
  • Training an AI Agent for Patch Analysis
  • Performing Binary Comparisons with AI
  • Ranking Exploitable Changes with AI
  • Generating Exploits

Overview

In this section we will focus on how AI can be leveraged to create, enhance, and stealthily deploy malware, while also bypassing modern Endpoint Detection and Response (EDR) solutions. We will begin with AI-assisted malware development, exploring how to obtain and refine proof-of-concept exploits, implement quality-of-life features, and integrate stealth techniques to minimize detection. Additionally, we analyze anomalous behaviors generated by AI models like ChatGPT, discussing how attackers can manipulate AI-generated code while bypassing built-in safety restrictions.

With a solid foundation in AI-driven malware creation, we shift to EDR evasion strategies, utilizing AI to ghostwrite well known malware, such as meterpreter, abuse trusted binaries, and deploy AI-assisted masquerading and payload obfuscation to blend seamlessly into legitimate system activity.

By the end of this section, you'll have an in-depth understanding of how AI enhances malware development and facilitates stealth techniques, equipping you with advanced evasion strategies to bypass modern security defenses.

Exercises

• AI - Assisted Malware Development: Foundations
• AI - Assisted Malware Development: Advanced Techniques
• Subverting Trust: Abusing LOLBins and Trusted Tools for Evasion
• Undetectable: Using AI for Ghostwriting

Topics

• Creating Custom Malware with AI
  • Obtaining Proof-of-Concept
  • Quality of Life Features
  • Adding Stealth Features
  • Anomalous Behaviors from ChatGPT
  • Safety Features
• Evading EDR with AI
  • Ghostwriting Malware with AI
  • Abusing Trusted Binaries to Achieve Execution
  • AI-Assisted Masquerading and Payload Obfuscation