Visão Geral

Este curso apresenta os fundamentos de segurança de Large Language Models, abordando arquitetura, riscos, proteção de dados, ataques contra modelos, segurança de aplicações e controles defensivos.

Objetivo

Após realizar este curso, você será capaz de:

• Compreender riscos de segurança em LLMs
• Identificar principais vetores de ataque
• Aplicar controles defensivos
• Avaliar segurança de aplicações com LLM

Publico Alvo

• Cybersecurity Professionals
• AI Security Professionals
• Developers
• Security Architects

Pre-Requisitos

• Conhecimentos básicos de Cybersecurity
• Familiaridade com IA Generativa
• Noções de APIs
• Conhecimentos básicos de aplicações web
• Familiaridade com LLMs é recomendada

Conteúdo Programatico

Module 1: LLM Security Fundamentals

1. Large Language Models
2. LLM architecture
3. Training and inference
4. Tokens
5. Context windows
6. Model interfaces
7. LLM application architecture
8. Security attack surface
9. LLM threat landscape
10. LLM security principles

Module 2: LLM Threats

1. Prompt Injection
2. Jailbreaking
3. Data leakage
4. Sensitive information disclosure
5. Model manipulation
6. Insecure output handling
7. Excessive agency
8. Supply chain risks
9. Model availability attacks
10. LLM threat modeling

Module 3: Application Security

1. LLM application architecture
2. System prompts
3. User prompts
4. Context management
5. Retrieval-Augmented Generation
6. Tool calling
7. External integrations
8. Application permissions
9. Output validation
10. Secure LLM application design

Module 4: Data Security

1. Training data
2. Context data
3. Sensitive information
4. Data leakage
5. Privacy
6. Data minimization
7. Access controls
8. Data isolation
9. Logging and monitoring
10. Data security controls

Module 5: LLM Access and Authorization

1. User authentication
2. Application authentication
3. API security
4. Tool permissions
5. Role-based access
6. Least privilege
7. Service identities
8. Authorization boundaries
9. Privileged operations
10. Access security

Module 6: LLM Security Monitoring

1. Prompt monitoring
2. Output monitoring
3. Tool usage monitoring
4. Anomaly detection
5. Security logging
6. Abuse detection
7. Incident response
8. Security metrics
9. LLM security testing
10. Monitoring architecture

Module 7: LLM Security Governance

1. AI governance
2. Security policies
3. Risk management
4. Threat modeling
5. Security assessments
6. Model evaluation
7. Third-party model risks
8. Supply chain security
9. Compliance considerations
10. LLM security governance

Module 8: Practical LLM Security Assessment

1. LLM threat modeling
2. Application security assessment
3. Prompt security assessment
4. Data security assessment
5. Access control review
6. Tool security assessment
7. Monitoring validation
8. Risk assessment
9. Security recommendations
10. LLM security case studies