AI Red Teaming Course and AI Security Expert Certification (Live)

Core concepts of artificial intelligence and modern generative AI systems.

• Introduction to AI, machine learning and deep learning
• Introduction to generative AI
• Supervised, unsupervised and reinforcement learning
• Large language model fundamentals
• Neural networks, weights, biases and training
• Transformers, tokens, embeddings, context windows
• Training vs inference
• Prompting basics and model behavior
• Model inputs, outputs and prediction pipelines
• LLM APIs, inference endpoints and parameters

The foundations of AI red teaming and how attackers analyze AI systems.

• Introduction to AI security and AI red teaming
• OWASP Top 10 for LLM Applications
• AI safety vs AI security vs AI ethics
• Identifying AI components in web apps
• AI threat landscape and real world attacks
• Mapping AI trust boundaries
• Understanding AI attack surfaces
• AI red team lab setup
• MITRE ATLAS framework overview
• Overview of AI red teaming tools

Gathering information about AI systems before launching attacks.

• Identifying AI powered features in applications
• Mapping data sources used by AI systems
• Fingerprinting LLMs and ML models
• Identifying inputs, tools, plugins, integrations
• Discovering hidden AI endpoints and APIs
• Understanding AI permissions and access boundaries
• Extracting system prompts and hidden instructions
• Preparing attack paths for AI exploitation
• Analyzing model behavior through prompt testing

Offensive techniques used against classical machine learning models.

• Introduction to adversarial machine learning
• Model extraction and model stealing
• Data poisoning attacks
• Query based model replication
• Label flipping and training data manipulation
• Surrogate model creation
• Backdoor attacks in ML models
• Membership inference attacks
• Dataset contamination and integrity attacks
• Model inversion attacks
• Evasion attacks and adversarial examples
• Training data and PII extraction
• White box vs black box attacks
• ML supply chain and malicious pretrained models
• FGSM and PGD attack techniques
• Pickle deserialization and model file exploitation
• Fooling image, text and tabular models

Attacking LLMs, chatbots and prompt based AI applications. This is the core offensive module of the program.

• Introduction to LLM vulnerabilities
• Multilingual prompt attacks
• Direct prompt injection
• Indirect prompt injection
• Instruction override attacks
• Hidden instructions in documents
• System prompt manipulation
• Poisoned web content attacks
• Context hijacking
• Email and file based prompt injection
• Bypassing basic AI restrictions
• System prompt leakage
• Advanced jailbreaking techniques
• Training data memorization attacks
• Persona based jailbreaks
• Sensitive data extraction from LLMs
• Role playing exploits
• Exfiltration via markdown, links, tool output
• Encoding bypasses with Base64, ROT13, hex
• LLM red teaming exercises and challenges
• Payload splitting and multi step jailbreaks

Attacks against modern AI architectures such as RAG systems, AI agents and multimodal AI.

• Retrieval augmented generation architecture
• Remote code execution through agent capabilities
• Red teaming RAG applications
• Resource exhaustion and infinite loop attacks
• Document poisoning
• Multimodal AI red teaming
• Vector database manipulation
• Vision language model attacks
• Retrieval hijacking
• Hidden text inside images
• Access control bypass in RAG systems
• Adversarial image prompts
• Cross tenant data leakage in AI apps
• Audio based injection attacks
• Red teaming AI agents
• Fine tuning and alignment exploitation
• Tool use exploitation
• Safety alignment bypass and reward hacking
• Function calling abuse
• LoRA adapter security risks
• Command injection through AI agents
• Session and memory abuse in AI apps
• SSRF attacks via AI tools

Scaling AI red teaming and running structured offensive security assessments.

• Automated AI vulnerability scanning
• Kubernetes and container risks for AI
• Prompt fuzzing techniques
• Offensive use of AI in cyberattacks
• Automated jailbreak testing
• AI powered phishing
• Using Garak for LLM security testing
• Deepfake based social engineering
• Using Promptfoo for prompt evaluation
• Voice cloning attacks
• Custom Python scripts for AI attacks
• AI assisted malware generation
• Chaining AI with web vulnerabilities
• Scoping AI red team engagements
• XSS through AI outputs
• Rules of engagement
• SQL injection via AI generated queries
• Legal and ethical considerations
• SSRF through AI connected tools
• Threat actor emulation
• MLOps and AI infrastructure attacks
• Multi stage AI attack simulation
• ML pipeline exploitation
• Exploit chaining and risk prioritization
• Model registry attacks
• Evidence collection and proof of concept
• CI/CD poisoning

Turning technical attacks into professional red team reports.

• AI red team reporting structure
• Severity scoring for AI vulnerabilities
• Executive summary creation
• Business impact analysis
• Technical vulnerability documentation
• Mapping findings to OWASP and MITRE ATLAS
• Attack scenario explanation
• Remediation recommendations
• Proof of concept presentation
• Presenting AI security risks to stakeholders

Defensive security measures to protect AI systems.

• Introduction to AI blue teaming
• Adversarial training basics
• Input validation for AI applications
• Safety alignment and model hardening
• Output filtering and content controls
• AI logging and audit trails
• LLM firewalls and guardrails
• Detecting prompt injection attempts
• Secure prompt design
• Monitoring abnormal model behavior
• System prompt protection techniques
• Drift detection and anomaly detection
• Access control for AI systems
• AI incident response
• Secure RAG architecture
• Building an AI security monitoring strategy
• Secure AI agent design