AWS Certified AI Practitioner (AIF-C01) Domain 5

Security, Compliance, and Governance for AI Solutions

Official Exam Guide: AWS Certified AI Practitioner Exam Guide

---

Domain Overview

Domain Weight: 14% of the exam

This domain tests your understanding of security, compliance, and governance practices specific to AI/ML workloads on AWS.

---

Key Concepts

1. Data Security for AI/ML

Why: AI/ML workloads process large amounts of data, often sensitive. Protecting this data is critical.

Security Measures:

• Encryption at rest: Amazon S3, EBS encryption
• Encryption in transit: TLS/SSL
• Access controls: IAM policies and roles
• Data classification: Identify sensitive data
• Secure data pipelines: Encrypted data flow

AWS Services:

• AWS KMS (Key Management Service)
• AWS Secrets Manager
• Amazon Macie (data discovery and classification)
• IAM (access management)

AWS Documentation:

• Security in SageMaker
• Amazon Bedrock Security

2. Model Security

Why: ML models themselves can be targets for attacks or contain sensitive information.

Threats:

• Model theft
• Model inversion attacks
• Adversarial examples
• Data poisoning

Protection Measures:

• Access controls for models
• Model encryption
• Endpoint security
• Input validation
• Output filtering

3. Compliance and Regulations

Why: AI/ML systems must comply with industry regulations and data protection laws.

Key Regulations:

• GDPR (data protection in EU)
• HIPAA (healthcare data in US)
• PCI DSS (payment card data)
• SOC 2 (service organization controls)
• Industry-specific regulations

AWS Compliance:

• AWS Artifact (compliance reports)
• HIPAA eligible services (SageMaker, Bedrock)
• PCI DSS compliant infrastructure
• SOC reports available

AWS Documentation:

• AWS Compliance Programs
• SageMaker Compliance

4. Governance and Monitoring

Why: Proper governance ensures AI systems are used appropriately and monitored for issues.

Governance Practices:

• AI use case approval processes
• Model validation and testing
• Change management
• Access controls and audit trails
• Regular reviews

Monitoring:

• Model performance monitoring
• Data quality monitoring
• Security monitoring
• Usage tracking
• Cost monitoring

AWS Services:

• Amazon SageMaker Model Monitor
• AWS CloudTrail (audit logging)
• Amazon CloudWatch (monitoring and alerting)
• AWS Config (configuration management)

AWS Documentation:

• SageMaker Model Monitor
• CloudTrail for ML

5. IAM for AI/ML Services

Why: Proper access controls prevent unauthorized use of AI services and data.

Best Practices:

• Least privilege principle
• Role-based access control
• Service-specific permissions
• Temporary credentials
• MFA for sensitive operations

Common Patterns:

• IAM roles for SageMaker notebooks
• IAM roles for Bedrock API access
• IAM policies for S3 data access
• Cross-account access for ML pipelines

AWS Documentation:

• IAM for SageMaker
• IAM for Bedrock

6. Network Security for AI/ML

Why: Isolating AI/ML workloads in secure networks prevents unauthorized access.

Security Measures:

• VPC deployment
• Private subnets
• VPC endpoints
• Security groups
• Network ACLs

AWS Services:

• Amazon VPC
• VPC Endpoints for AWS services
• AWS PrivateLink

AWS Documentation:

• SageMaker in VPC

---

AWS Services for AI/ML Security

Core Security Services

1. AWS Identity and Access Management (IAM)
   • User and role management
   • Fine-grained permissions
   • IAM Documentation
2. AWS Key Management Service (KMS)
   • Encryption key management
   • Data encryption
   • KMS Documentation
3. AWS CloudTrail
   • API call logging
   • Audit trails
   • CloudTrail Documentation
4. Amazon CloudWatch
   • Monitoring and alerting
   • Log aggregation
   • CloudWatch Documentation
5. AWS Config
   • Resource configuration tracking
   • Compliance monitoring
   • Config Documentation
6. Amazon Macie
   • Sensitive data discovery
   • Data classification
   • Macie Documentation

---

Shared Responsibility Model for AI/ML

AWS Responsibilities:

• Physical security of infrastructure
• Hardware and network security
• Service availability
• Foundation model security (for Bedrock)

Customer Responsibilities:

• Data security and encryption
• Access management (IAM)
• Network configuration (VPC)
• Model security
• Application-level security
• Compliance with regulations

---

Best Practices Summary

1. Encrypt data at rest and in transit
2. Use IAM roles instead of access keys
3. Enable CloudTrail for audit logging
4. Implement least privilege access
5. Monitor models in production
6. Classify data and protect sensitive information
7. Use VPC for network isolation
8. Regular security reviews and audits
9. Compliance validation for regulated industries
10. Incident response plan for security events

---

AWS Documentation

• AWS Security Best Practices
• Machine Learning Security
• SageMaker Security
• Bedrock Security

---

Final Thoughts on Domain 5

Security, compliance, and governance are critical for production AI/ML systems. Understand both AWS security services and AI-specific security considerations!