Data Protection & Resilience
AWS SAA SAA-C03 · Design Secure Architectures
Encryption: protecting data at rest and in transit
Encryption is the exam's most reliable security topic, and the rules are consistent across services. Data should be encrypted at rest (stored on disk) and in transit (moving across a network). AWS provides the building blocks: AWS Key Management Service (KMS) creates and manages encryption keys, and most AWS services integrate with it. A KMS key (formerly called a customer master key) can be AWS managed (created and managed by AWS for a service), customer managed (you create and control it, including its key policy and rotation), or AWS owned (used internally by AWS). Customer managed keys give you control over who can use the key and let you audit every use through CloudTrail. Automatic key rotation is available for customer managed keys and is a best practice the exam expects you to know.
Encryption at rest is available — and often enabled by default — across the major services: S3 encrypts new objects by default (SSE-S3 with Amazon-managed keys, or SSE-KMS with keys you control), EBS volumes can be encrypted with encryption by default at the Region level, RDS databases and snapshots support encryption at rest, and DynamoDB encrypts tables by default. Encryption in transit uses TLS: HTTPS for web traffic and API calls, TLS for database connections. The exam's favorite encryption scenario is a compliance requirement: data must be encrypted at rest with keys the company controls and audits. The answer is SSE-KMS (or the equivalent KMS integration) with a customer managed key, because only KMS gives you key policies, audit trails, and rotation control.
Secrets and credentials management
Hard-coding passwords, API keys, and database credentials in application code or configuration files is one of the most common security failures, and the exam tests the alternatives. AWS Secrets Manager stores secrets (database credentials, API keys, OAuth tokens) and can automatically rotate them on a schedule — including rotating RDS, Aurora, Redshift, and DocumentDB credentials without application changes. AWS Systems Manager Parameter Store stores configuration values and secrets (as SecureString parameters encrypted with KMS) at no additional charge for standard parameters, but it does not rotate secrets automatically. The exam's rule: if the question mentions automatic rotation, the answer is Secrets Manager; if it just needs secure storage of a configuration value, Parameter Store is the cheaper fit. Both are far better than embedding credentials in code, in AMIs, or in environment variables on instances.
Protecting the edge: WAF, Shield, and network security
AWS WAF is a web application firewall that filters HTTP(S) traffic in front of CloudFront distributions, Application Load Balancers, and API Gateway APIs. It blocks common attacks (SQL injection, cross-site scripting), filters by IP, geography, or request patterns, and supports rate-based rules that block clients sending too many requests. Managed rule groups from AWS and third parties cover common threats without writing custom rules. AWS Shield protects against DDoS attacks: Shield Standard is automatically enabled for all AWS customers and defends against common network and transport layer attacks, while Shield Advanced adds 24/7 DDoS response team support, advanced detection, cost protection against DDoS-related scaling, and WAF integration. The exam's distinction: WAF filters application-layer requests; Shield defends against volumetric DDoS attacks. They complement each other and are often deployed together.
Network-layer protection inside the VPC comes from security groups and NACLs (covered in the VPC guide), plus AWS Network Firewall for stateful inspection of VPC traffic, and Amazon GuardDuty for threat detection. For private connectivity, VPC endpoints and PrivateLink keep traffic off the public internet, and Direct Connect provides a private path from on-premises. The exam rewards designs that layer these controls: WAF at the edge, security groups at the instance, encryption everywhere, and monitoring to catch what slips through.
Backup and resilience: designing for failure
Resilience is the ability to keep working when things fail, and the exam tests it as a set of patterns rather than a single service. AWS Backup centralizes backup management across services (EC2, EBS, RDS, DynamoDB, EFS, and others) with backup plans, retention policies, and cross-Region copies — the answer when a question asks for centralized, policy-driven backups. Multi-AZ deployment protects against the failure of one Availability Zone; multi-Region deployment protects against the failure of an entire Region. Health checks and automatic failover (Route 53, ELB, RDS Multi-AZ) remove failed components from service without human intervention. Decoupling with SQS and SNS means one component's failure does not cascade to others. Stateless application tiers behind a load balancer mean any instance can fail or be replaced without losing user sessions — session state belongs in ElastiCache or DynamoDB, not on the instance.
When a question describes a failure scenario, the design answer follows a checklist: Is the data replicated (Multi-AZ, cross-Region replication, backups)? Is the compute spread across AZs with automatic replacement (Auto Scaling)? Is traffic routed only to healthy components (load balancer health checks, Route 53 failover)? Is the application decoupled so failures do not cascade (queues)? The best answer is usually the one that checks the most boxes at the lowest cost — and the exam's wrong answers usually fail one of them, such as putting everything in a single AZ or storing session state on the instance.
Defense in depth and incident response
No single control makes an architecture secure; the exam expects defense in depth — layers of controls so that a failure in one layer is caught by another. At the edge, CloudFront and WAF filter requests; at the network, security groups and NACLs restrict traffic; at the identity layer, IAM roles and least privilege limit what any credential can do; at the data layer, encryption protects information even if access controls fail; and across everything, CloudTrail, GuardDuty, and Config watch for the unexpected. When a question asks for the “most secure” design, the answer is usually the one that adds a layer the other options lack.
When something does go wrong, AWS provides the tools for a structured incident response. The standard steps are: contain the affected resource (change its security group to isolate it, or detach it from the load balancer), preserve evidence (snapshot the EBS volumes before making changes, so forensic analysis is possible), eradicate the cause (revoke compromised credentials, remove malicious software, close the vulnerability), and recover (restore from a known-good snapshot or redeploy from a clean AMI). Amazon Macie helps on the data side: it uses machine learning to discover sensitive data — such as personally identifiable information or credentials — stored in S3, and alerts when buckets are accessed unusually. For compliance evidence, AWS Artifact provides on-demand access to AWS's security and compliance reports and certifications, which auditors and regulators accept as evidence of the platform's controls. The exam's incident questions usually test the order of operations: isolate first, preserve evidence, then fix — never delete the evidence by terminating the instance before snapshotting it.
Exam traps to watch for
- Using SSE-S3 when the requirement is keys the company controls and audits. SSE-KMS with a customer managed key is the answer when key control and auditability matter.
- Storing database credentials in code or Parameter Store when automatic rotation is required. Secrets Manager is the service that rotates secrets.
- Confusing WAF with Shield: WAF filters application-layer web requests; Shield defends against DDoS attacks. They solve different problems.
- Storing session state on EC2 instances in an Auto Scaling group. When an instance is replaced, the sessions die with it — put state in ElastiCache or DynamoDB.
Practice questions
Test yourself on this guide. Answers are checked instantly, with a full explanation for each question.
Q1.A healthcare company stores patient records in S3. Compliance rules require that data be encrypted at rest using encryption keys that the company controls, with an audit trail of every key use. Which encryption option meets these requirements?
Explanation: SSE-KMS with a customer managed key gives the company control over the key (key policy, rotation, disabling) and logs every key use to CloudTrail for auditing — exactly the compliance requirement. Option B encrypts the data but the keys are managed by Amazon, with no customer key policy or per-use audit trail. Option C puts keys in code, where they can be exposed and cannot be rotated or audited properly. Option D ignores the encryption requirement entirely.
Exam objective: SAA-C03 Task Statement 1.3: Determine appropriate data security controls
Q2.A company stores database credentials in its application configuration files. The security team wants the credentials stored securely and rotated automatically every 30 days without application code changes. Which AWS service meets these requirements?
Explanation: Secrets Manager stores secrets and rotates them automatically on a schedule, including native rotation support for RDS, Aurora, Redshift, and DocumentDB credentials. Option B stores secrets securely (as SecureString parameters) but does not rotate them automatically. Option C manages encryption keys, not application secrets or rotation. Option A is object storage and is not a secrets management service.
Exam objective: SAA-C03 Task Statement 1.3: Determine appropriate data security controls
Q3.A public-facing web application is experiencing a large volumetric DDoS attack that is overwhelming the Application Load Balancer with traffic. Which AWS service provides advanced DDoS protection with access to a 24/7 response team?
Explanation: Shield Advanced is the DDoS protection service: it adds advanced detection, cost protection, and access to the AWS DDoS Response Team (DRT) on top of the automatic Shield Standard protection. Option D filters application-layer web requests (SQL injection, XSS) but is not a volumetric DDoS defense. Option B detects threats but does not mitigate an ongoing DDoS attack. Option C inspects VPC network traffic and is not a DDoS mitigation service for public endpoints.
Exam objective: SAA-C03 Task Statement 1.2: Design secure workloads and applications
Q4.A company runs a stateless web application on EC2 instances in an Auto Scaling group. User sessions are currently stored in memory on each instance, and users lose their sessions whenever an instance is replaced during scaling events. What is the BEST solution?
Explanation: Moving session state to a shared store (ElastiCache for speed, DynamoDB for durability) makes the application tier truly stateless: any instance can serve any user, and replacing instances loses nothing. Option C gives up the resilience and scaling benefits of the ASG. Option A (sticky sessions) reduces the problem but does not solve it — when the instance a user is stuck to is terminated, the session is still lost. Option D is unrelated: the problem is where state lives, not how much memory the instance has.
Exam objective: SAA-C03 Task Statement 2.1: Design scalable and loosely coupled architectures
Q5.A company needs a centralized way to manage backups across EC2, RDS, DynamoDB, and EFS, with backup plans, retention rules, and copies in a second Region. Which AWS service meets this requirement?
Explanation: AWS Backup is the centralized backup service: it manages backup plans, schedules, retention, and cross-Region copies across EC2, EBS, RDS, DynamoDB, EFS, and other services from one place. Option D manages the lifecycle of EBS snapshots only, not backups across all these services. Option A audits API activity. Option B manages S3 object storage classes and expiration, not backups of databases and file systems.
Exam objective: SAA-C03 Task Statement 2.2: Design highly available and/or fault-tolerant architectures
Q6.A web application behind an Application Load Balancer is receiving SQL injection attempts in HTTP requests. The company wants to block these requests before they reach the application servers. Which solution meets this requirement?
Explanation: AWS WAF inspects HTTP requests at the load balancer and blocks those matching SQL injection rules — the managed rule groups cover this without custom rule writing. Option B protects against DDoS attacks, not application-layer injection. Option A logs network traffic but blocks nothing. Option D is a network-layer control that cannot inspect HTTP request content, and blocking individual IPs does not scale against distributed attacks.
Exam objective: SAA-C03 Task Statement 1.2: Design secure workloads and applications
Q7.A company encrypts its EBS volumes with a customer managed KMS key. The security policy requires the key material to be changed automatically every year without any application downtime. Which KMS feature meets this requirement?
Explanation: Automatic key rotation generates new key material for a customer managed key every year, and AWS services keep using the key alias transparently — old data encrypted with previous key material remains decryptable because KMS retains the old material, so there is no downtime and no re-encryption. Option A is wrong: automatic rotation does not require re-encrypting data. Option D is the manual, disruptive approach that automatic rotation exists to avoid. Option B is wrong on both counts: AWS managed keys rotate automatically too, and switching to one would give up the customer control the company wants.
Exam objective: SAA-C03 Task Statement 1.3: Determine appropriate data security controls
Q8.A company runs a critical application in a single Availability Zone. The application must survive the failure of that Availability Zone with minimal downtime. Which combination of changes BEST achieves this?
Explanation: Surviving an AZ failure requires the whole stack to span AZs: application instances in two AZs behind a load balancer that routes only to healthy targets, and a database with a synchronous standby in the second AZ (RDS Multi-AZ) for automatic failover. Option B helps recovery after a failure but the application still goes down and must be rebuilt. Option D improves performance and visibility but does nothing for AZ resilience. Option C protects S3 objects against overwrites, which is unrelated to AZ failure.
Exam objective: SAA-C03 Task Statement 2.2: Design highly available and/or fault-tolerant architectures