Defender Intermediate

Secure Coding for Large Language Model Applications

September 23, 2026 09:00 – 13:00 Auditorium 2 4 Hours · 6 Modules · 4 Labs
All Trainings

About This Training

AI-driven applications powered by Large Language Models are rapidly transforming products, developer workflows, and customer experiences. But these systems introduce unique security risks that traditional AppSec practices simply don't address.

This half-day hands-on course teaches developers, AppSec engineers, and architects how to design and build secure AI/LLM applications, with a focused deep-dive into five critical vulnerability areas from the OWASP Top 10 for LLM Applications 2025: Prompt Injection, Sensitive Information Disclosure, Excessive Agency, Improper Output Handling, and RAG/Data Poisoning.

Through hands-on labs and real-world case studies, attendees gain practical skills for deploying safe, trustworthy, and compliant AI capabilities at scale.

Your Trainer

Fabio Cerullo
Fabio Cerullo
Cycubix LTD · ISC2 Certified Instructor · CISSP, CSSLP, CCSP

Fabio Cerullo is a seasoned cybersecurity trainer and consultant with over 15 years of industry experience across financial services, government, startups, and software companies. He has delivered training to thousands of developers and security professionals worldwide, with a focus on application security, cloud security, and information security. Fabio is an official certified instructor for ISC2 and a Champion AWS Authorized Instructor, regularly teaching cloud architecture and security topics. He is a frequent speaker at international events hosted by ISC2, OWASP, and ISACA.

LinkedIn cycubix.com

Course Outline

1
30 min

Foundations of AI & LLM Security

  • Modern AI system architectures and threat surfaces
  • Why traditional AppSec models fail for AI
  • Key attack surfaces: model, context, tools, retrieval layers
  • Introduction to the OWASP Top 10 for LLM Applications 2025
2
45 min

Prompt Injection — LLM01:2025 🔬 Lab

  • Direct vs indirect prompt injection attack patterns
  • Jailbreaking techniques and adversarial prompts
  • Defenses: input validation, prompt hardening, sandboxing
  • Lab: Prompt Injection Leading to Sensitive Data Leakage
3
30 min

Sensitive Information Disclosure — LLM02:2025 🔬 Lab

  • How LLMs leak training data, system prompts, and user PII
  • System prompt extraction and output probing attacks
  • Data minimization and output filtering strategies
  • Lab: Client Data Leaks in Model
— Break (15 min) —
4
30 min

Excessive Agency — LLM06:2025 🔬 Lab

  • Risks of autonomous decision-making and over-permissioned agents
  • Principle of least privilege for AI agents
  • Human-in-the-loop controls and action confirmation patterns
  • Lab: Unrestricted Agent Functionality Leads to Command Injection
5
30 min

Improper Output Handling — LLM05:2025 🔬 Lab

  • Downstream injection via LLM-generated content (XSS, SQLi, RCE)
  • Output encoding, sanitization, and context isolation
  • Lab: Insecure HTML Handling Leading to XSS
6
30 min

RAG Security & Data Poisoning — LLM04:2025

  • RAG architecture attack surfaces and retrieval poisoning
  • Embedding manipulation and semantic privilege escalation
  • Indirect prompt injection via poisoned documents
  • Securing retrieval pipelines and knowledge base integrity
— Wrap-up & Q&A (15 min) —

Prerequisites

What You'll Take Home

📚
Digital Course Materials
Complimentary access to all materials via the Cycubix Academy
🏅
Certificate of Participation
Issued upon successful completion of the workshop
📖
Exclusive eBook
The Developer's Playbook for Large Language Model Security by Steve Wilson
🔬
Lab Environment Access
Secure AI/LLM lab environment provided during the workshop

Ready to Secure AI?

Seats are limited. Training ticket required — separate from the main conference ticket.

Book Your Seat