Security, Cost & Well-Architected Frameworks
Cloud Security Architecture
Security is non-negotiable in cloud architecture. Interviewers assess your ability to design secure systems while maintaining usability and performance.
Shared Responsibility Model
Understanding who secures what is fundamental.
AWS Shared Responsibility
| AWS Responsibility | Customer Responsibility |
|---|---|
| Physical security | IAM policies |
| Network infrastructure | Security groups/NACLs |
| Hypervisor | OS patching (EC2) |
| Managed service security | Data encryption |
| Global infrastructure | Application security |
Service-Specific Boundaries
| Service Type | AWS Manages | Customer Manages |
|---|---|---|
| IaaS (EC2) | Hardware, hypervisor | OS, apps, data |
| PaaS (RDS) | OS, DB engine | Data, access |
| SaaS (S3) | All infrastructure | Data, permissions |
Identity and Access Management
IAM Best Practices
-
Least Privilege Principle
- Start with zero permissions
- Add only what's needed
- Use IAM Access Analyzer
-
Role-Based Access
- Prefer roles over users for services
- Use cross-account roles
- Implement temporary credentials
-
MFA Everywhere
- Root account: Hardware MFA
- Admin users: Virtual MFA minimum
- Programmatic: STS with MFA
Interview Question: IAM Design
Q: "Design an IAM strategy for a multi-account AWS organization."
A: Implement hub-and-spoke model:
Organization Root
├── Security Account (hub)
│ ├── IAM Identity Center
│ ├── CloudTrail aggregation
│ └── Security tooling
├── Production Account
│ └── Cross-account roles → Security
├── Development Account
│ └── Cross-account roles → Security
└── Sandbox Account
└── Limited cross-account access
Key Decisions:
- Centralized identity in Security account
- Permission sets managed via IAM Identity Center
- Break-glass procedures for emergencies
- Service Control Policies (SCPs) at OU level
Network Security
Defense in Depth
Internet
↓
[WAF] - Layer 7 filtering
↓
[CloudFront] - DDoS protection
↓
[ALB in Public Subnet] - SSL termination
↓
[Security Group] - Stateful firewall
↓
[Private Subnet] - Application tier
↓
[NACL] - Stateless subnet firewall
↓
[Security Group] - Database access
↓
[Isolated Subnet] - Data tier
VPC Security Design
| Component | Purpose | Best Practice |
|---|---|---|
| Security Groups | Instance firewall | Least privilege, reference by SG ID |
| NACLs | Subnet firewall | Explicit deny rules, logging |
| VPC Flow Logs | Network monitoring | Enable for all subnets |
| PrivateLink | Service connectivity | Avoid public internet |
| Transit Gateway | Multi-VPC routing | Centralized inspection |
Interview Question: Zero Trust Architecture
Q: "How would you implement Zero Trust in AWS?"
A: Zero Trust principles in AWS:
-
Verify Explicitly
- IAM authentication for all requests
- VPC endpoints for service access
- mTLS for service-to-service
-
Least Privilege Access
- Fine-grained IAM policies
- Security group microsegmentation
- Time-limited credentials (STS)
-
Assume Breach
- Network segmentation
- Encryption everywhere (TLS, KMS)
- Continuous monitoring (GuardDuty)
Encryption Strategy
Encryption at Rest
| Service | Default | Recommended | Key Management |
|---|---|---|---|
| S3 | None | SSE-KMS | Customer CMK |
| EBS | None | KMS encrypted | Customer CMK |
| RDS | Optional | Encryption enabled | Customer CMK |
| DynamoDB | None | KMS encryption | AWS or Customer CMK |
Encryption in Transit
- External traffic: TLS 1.2+ mandatory
- Internal traffic: TLS between services
- Database: SSL connections enforced
- S3: Bucket policy requiring TLS
KMS Best Practices
-
Key Hierarchy
- Separate keys per environment
- Cross-account key sharing via policies
- Automatic key rotation (1 year)
-
Key Policies
- Explicit allow for key administrators
- Separate usage from administration
- Audit all key usage
Security Monitoring
Detection Services
| Service | Detection Type | Use Case |
|---|---|---|
| GuardDuty | Threat detection | Malicious activity |
| Inspector | Vulnerability scanning | EC2, Lambda, ECR |
| Macie | Data discovery | PII in S3 |
| Security Hub | Aggregation | Unified dashboard |
| Config | Compliance | Resource configuration |
Incident Response
Preparation:
- Runbooks for common incidents
- IAM roles for incident responders
- Forensic account for evidence
Detection → Analysis → Containment → Eradication → Recovery
Key Insight: Security in the cloud is about automation. Manual security processes don't scale—use AWS Config rules, automated remediation, and infrastructure as code.
Next, we'll explore cost optimization strategies. :::