Practical AI
Security Bootcamp
Understand, Attack, Defend, and Monitor AI Systems
Tools and Frameworks
& so many more tools...
WHY INFOSEC TRAIN
Why Attend?
AI is rapidly becoming part of business applications, security workflows, automation systems, and decision-making processes. But as AI adoption grows, so does the attack surface across data, models, prompts, APIs, agents, pipelines, and infrastructure.
This bootcamp helps participants understand how AI systems work, how attackers target ML, LLM, and agentic AI systems, and how defenders can apply practical controls across the AI security lifecycle. Through trainer-led demonstrations, real-world scenarios, and framework-based explanations, participants will see how AI attacks, threat modelling, defenses, monitoring, and incident response work in practice.
AI Security Lifecycle
Learn fundamentals, attacks, defense, monitoring, and response
Trainer-Led Demonstrations
Practical walkthroughs across ML, LLM, and agentic systems
ML, LLM & Agent Coverage
Understand risks across modern AI systems
Framework-Based Approach
Map risks using MITRE ATLAS, STRIDE, and OWASP
Practitioner Tool Exposure
Explore tools for attacks, guardrails, monitoring, and defense
AI Security Career Edge
Earn 8 CPEs and strengthen role readiness
What's included
- 8 CPE Credits
- 4 Live Sessions · 8 Hours Total
- 60-Day Recording Access
- Trainer-Led Demos & Labs
- Certificate of Attendance
Secure Your Seat
Comprehensive 8-Hour Bootcamp
LEAD INSTRUCTOR
Meet the Expert
Experience
Avnish
AI Security | AI Governance | Cloud Security | Data Security | Consultant & Trainer
Avnish is an Information Security and Cloud Security Consultant with 7+ years of experience in AI security, cloud security, data security, AI-assisted threat detection, and securing AI/ML pipelines. For this bootcamp, he brings practical expertise in AI attacks, threat modelling, defensive controls, monitoring, and governance through real-world examples and trainer-led demonstrations.
Specializations:
- AI security, AI governance, and model risk assessment
- Securing AI/ML pipelines, models, and data flows
- AI-assisted threat detection and SOC automation
- AWS, Azure, ELK, Splunk, and cloud security monitoring
- Threat modelling, incident response, and vulnerability remediation
- NIST AI RMF, AI governance, and model risk validation
CURRICULUM OUTLINE
Agenda
Deep-dive training sessions across 4 days
- Understanding AI: Components of an AI system, and Predictive vs Generative vs Agentic AI
- How Machines Learn: Machine Learning vs Deep Learning, Model Development Lifecycle
- Transformer Architecture: Understanding how LLMs work: tokens, embeddings, attention, context window, system vs user prompts
- Agentic AI Basics: planning, memory, tool use
- Understanding: how AI Systems vary from Traditional Software
- The AI Attack Surface: prompt, data, model, API, pipeline, and infrastructure risk layers
- Limitations of AI: hallucination, bias, explainability gaps, overreliance
Map the architecture of a sample AI application and identify its assets, entry points, and data flows
- OWASP: ML Security Top 10 overview
- Poisoning attacks: data and model poisoning concepts
- Evasion attacks: crafting inputs to bypass ML-based detectors
- Privacy attacks: Extraction and inference
- Attack tooling: Adversarial Robustness Toolbox (ART)
Run a pre-built evasion attack against a sample network attack detector
- OWASP LLM: Top 10 and OWASP Agentic AI Top 10 overview
- Direct and indirect prompt injection: system prompt leakage
- Jailbreaking techniques
- Context-window: overflow and vector store poisoning
- Tool misuse: excessive agency, and MCP-specific attack vectors
- Automated testing tools: Garak, PyRIT, Promptfoo
Run an automated vulnerability scan against a sample AI application and identify exploitable weaknesses
- Threat modelling fundamentals: threat actors, assets, abuse cases, entry points, data flows, existing controls
- Applying MITRE ATLAS and STRIDE to AI applications
- Trust boundaries in AI systems: data, model, pipeline, infrastructure, and (for LLM/agentic systems) tool and memory boundaries
Produce a trust-boundary/data-flow diagram and threat model for a sample AI application
- Defense-in-Depth for AI: the layered security model across data, model, application, and infrastructure
- Data Security: data classification, leakage patterns (prompt, output, log, vector DB), sanitization and de-identification techniques
- Data protection tooling: Microsoft Presidio (PII detection/masking), Gitleaks/TruffleHog (secrets detection)
- Model Security for ML/DL: adversarial training, artifact integrity, model registry protection
- Model registry tooling: MLflow (versioning, model cards)
- Model Security for LLMs & Agents: input/output guardrails, guardrail approaches (rule-based, classifier-based, LLM-as-judge)
- Guardrail tooling: LLM-Guard, LlamaGuard, GuardrailsAI
- AI Application Security: API authentication/authorization, rate limiting, and cost budgeting
- Gateway tooling: LiteLLM (rate limiting, authentication, request logging)
Apply input/output guardrails to a vulnerable LLM application and verify they block malicious prompts
- AI security monitoring overview:prompt, API usage, model query, and tool/agent action monitoring
- Monitoring/observability tooling: Evidently AI (drift reports), Arize Phoenix, Langfuse/Langsmith (LLM tracing)
- Abuse detection: high-volume querying, repeated extraction, and repeated jailbreak attempts
- AI incident scenarios: prompt injection, sensitive data exposure, model extraction, poisoned dataset
- AI Supply Chain Security: open-source model and dataset risk, third-party API risk, model provenance verification
- Supply chain tooling: SBOM/AI-BOM concepts
Trace LLM calls and tool invocations using Arize Phoenix, and identify repeated jailbreak/extraction attempts from the trace data
*Note: Participants will have access to session recordings for a period of 60 days.
What's included
- 8 CPE Credits
- 4 Live Sessions · 8 Hours Total
- 60-Day Recording Access
- Trainer-Led Demos & Labs
- Certificate of Attendance
Secure Your Seat
Comprehensive 8-Hour Bootcamp
PARTICIPANTS
Who Is This For?
Who Should Attend?
This bootcamp is suitable for:
- Security professionals exploring AI security as an emerging specialization
- SOC analysts, penetration testers, AppSec engineers, and security engineers
- IT and security managers evaluating AI risks before approving AI initiatives
- Developers, ML engineers, and data professionals who want to understand AI attack and defense concepts
- GRC, risk, audit, and compliance professionals seeking a technical foundation before moving into AI governance frameworks
Prerequisites
Participants should have:
- Basic understanding of IT and cybersecurity fundamentals, such as networks, applications, and common attack types
- Familiarity with AI/ML concepts is helpful but not mandatory, as Day 1 covers the required fundamentals
- Comfort reading code and scripts conceptually, as walkthroughs will include code and CLI examples
- No prior AI security experience is required
Participant Resources
Participants will receive:
- Practical walkthrough documents for all demonstrations covered in the bootcamp
- Reference materials for key AI security topics, tools, frameworks, and attack/defense concepts
- Guidance to replicate the demonstrated practicals at their own pace after the bootcamp
BOOTCAMP BENEFITS
Key Takeaways
-
Understand AI, ML, LLM, and agentic AI risks
-
Identify attack surfaces across models, prompts, APIs, and data
-
Explore AI attacks through trainer-led practical demonstrations
-
Map AI threats using OWASP, MITRE ATLAS, and STRIDE
-
Learn defenses for guardrails, data, models, and applications
-
Earn 8 CPEs and build practical AI security readiness
Secure Your Seat
Join us for 4 days of intensive AI security training.
- 60-day recording access
- Certificate & 8 CPEs