Passwords alone are not enough to protect something as valuable as your AWS account. In this subchapter, we’ll look at two mechanisms that greatly enhance security: Multi-Factor Authentication (MFA) and temporary credentials (STS). They are the difference between a vulnerable account and a well-protected one.
MFA: Adding a Second Barrier
MFA stands for Multi-Factor Authentication. The idea: to log in, it’s not enough to know something (the password); you also have to have something (a code that changes every few seconds on your phone).
The “factors” of authentication are of three types:
- Something you know: a password.
- Something you have: your phone with a code app, or a physical key.
- Something you are: your fingerprint, your face (biometrics).
MFA combines at least two of these factors. The most common is password (know) + phone code (have).
Analogy: It’s like an ATM. To withdraw money you need the card (something you have) and the PIN (something you know). With only one of the two, you can’t. If someone steals your card but doesn’t know the PIN, they can’t withdraw money.
Why MFA is So Effective
The big advantage: even if an attacker steals your password, they can’t get in without the second factor, which is physically on your phone.
Real example: An attacker gets your AWS password through a phishing email. They try to log in… but AWS asks for the MFA code, which only appears on your phone. The attacker is locked out. The stolen password alone is useless.
According to security studies, enabling MFA blocks the vast majority of automated account theft attacks. It’s one of the most effective and cheapest security measures out there.
How It’s Used in AWS
- You can use an authenticator app (Google Authenticator, Authy, Microsoft Authenticator…) that generates 6-digit codes that change every 30 seconds.
- Or a physical security key (like a YubiKey) for maximum protection.
Absolute golden rule: ALWAYS enable MFA on the root user, no exceptions. The root has total power over your account; protecting it with MFA is the first thing you should do when creating an account. And enable it for all users with important permissions. We’ll review this in subchapter 7.5.
The Problem with Permanent Credentials
Before talking about STS, let’s understand the problem it solves.
An IAM user can have permanent access keys to program against AWS. The problem: those keys never expire. If they leak (uploaded to GitHub by mistake, stolen from a laptop…), an attacker can use them indefinitely until someone notices and disables them.
Permanent credentials are like a key that never expires: convenient, but dangerous if lost.
STS: Temporary Credentials That Expire Automatically
STS stands for Security Token Service. Its function: to generate temporary credentials that expire automatically after a while (from a few minutes to a few hours).
Analogy: STS credentials are like a hotel key card. It gives you access to your room, but stops working the day you leave. You don’t have to return it or worry: it expires on its own. If you lose it, the damage is limited because it will soon stop working.
Why Temporary Credentials Are More Secure
| Permanent Credentials | Temporary Credentials (STS) | |
|---|---|---|
| Do they expire? | No (until someone deletes them) | Yes, automatically |
| Risk if leaked | High (valid forever) | Low (stop working soon) |
| Need to rotate manually | Yes | No (they regenerate automatically) |
The fact that they expire on their own drastically reduces the risk: a stolen credential that stops working in an hour is much less dangerous than one that lasts forever.
The Connection with Roles
Remember roles from subchapter 7.1? Here’s the magic: when someone (or something) “assumes a role,” STS gives them temporary credentials with the permissions of that role.
This is exactly what happens when:
- An EC2 instance assumes a role to read S3 (STS gives it temporary credentials behind the scenes).
- A Lambda function accesses a database (same mechanism).
- A user from another account assumes a role to access your resources (cross-account).
[EC2 / Lambda / User] ──assumes a role──► [STS]
│
delivers temporary credentials
(with the role’s permissions, expire automatically)
▼
[Secure access to the resource]Why this is huge for security: With roles + STS, there are no permanent keys stored anywhere. The EC2 instance doesn’t have a file with keys someone could steal; it gets temporary credentials that renew and expire automatically. That’s why the best practice from subchapter 7.1 (“use roles, not keys on the server”) is so important.
The Modern Best Practice
The professional and secure way to work:
- People: log in with user + password + MFA. For advanced tasks, assume roles (temporary credentials).
- Services and applications (EC2, Lambda, etc.): use roles, never embedded permanent keys.
- Permanent access keys: avoided whenever possible. If used, rotate them frequently.
Modern tools like AWS IAM Identity Center (formerly AWS SSO) make it easy to give people access with temporary credentials and MFA centrally, without permanent keys.
What You Should Remember
- MFA adds a second factor (a code on your phone) in addition to the password: even if your password is stolen, they can’t get in. Always enable it on root and important users.
- Permanent credentials (access keys) are dangerous because they don’t expire: if leaked, they work indefinitely.
- STS generates temporary credentials that expire automatically, greatly reducing risk (like a hotel key card).
- When assuming a role, STS delivers those temporary credentials. That’s why roles + STS allow services like EC2 or Lambda to access resources without storing permanent keys.
- Best practice: people with MFA, services with roles, avoid permanent keys.
In the last IAM subchapter, we’ll bring everything together in a list of security best practices you should always apply.
Cloud, AWS & Terraform — From Zero to Expert
Chapter 1 · What is cloud computing
- 1.1 The traditional client-server model
- 1.2 Problems the cloud came to solve
- 1.3 On-premise vs cloud vs hybrid
- 1.4 The three service models: IaaS, PaaS, SaaS
- 1.5 The five pillars of cloud (according to NIST)
- 1.6 Real advantages: elasticity, pay-as-you-go, global availability
Chapter 2 · The cloud market and major providers
- 2.1 AWS, Azure and GCP: differences and market share
- 2.2 Why learn AWS first
- 2.3 Concepts that are universal among providers
Chapter 3 · Regions, availability zones and edge
- 3.1 What is an AWS region and how to choose it
- 3.2 Availability Zones: high availability by design
- 3.3 Edge locations and CloudFront
- 3.4 Latency, resilience and data sovereignty
Chapter 4 · Compute: EC2
- 4.1 Instances: types, families and when to choose each
- 4.2 AMIs, key pairs and Security Groups
- 4.3 Instance lifecycle
- 4.4 Elastic IPs and Placement Groups
- 4.5 Savings Plans vs Reserved vs On-Demand vs Spot
Chapter 5 · Storage: S3
- 5.1 Buckets, objects and keys
- 5.2 Storage classes (Standard, IA, Glacier…)
- 5.3 Versioning and object lifecycle
- 5.4 Bucket policies and ACLs
- 5.5 Static website hosting
Chapter 6 · Networking: VPC
- 6.1 What is a VPC and why you need it
- 6.2 Public and private subnets
- 6.3 Internet Gateway and NAT Gateway
- 6.4 Route Tables and Network ACLs
- 6.5 VPC Peering and endpoints
Chapter 7 · Identity and access: IAM
- 7.1 Users, groups, roles and policies
- 7.2 The principle of least privilege
- 7.3 Identity-based vs resource-based policies
- 7.4 MFA and temporary credentials (STS)
- 7.5 IAM security best practices
Chapter 8 · Managed databases
- 8.1 RDS: engines, Multi-AZ and read replicas
- 8.2 Aurora and its advantages over vanilla RDS
- 8.3 DynamoDB: key-value / document model
- 8.4 ElastiCache for in-memory cache
- 8.5 When to use each type of database
Chapter 9 · Why Infrastructure as Code
- 9.1 Problems with manual provisioning
- 9.2 Declarative vs imperative IaC
- 9.3 Terraform vs CloudFormation vs Pulumi vs CDK
- 9.4 The plan → apply → destroy cycle
Chapter 10 · HCL: the Terraform language
- 10.1 Resource, variable, output, locals blocks
- 10.2 Data types: string, number, bool, list, map, object
- 10.3 Expressions, references and built-in functions
- 10.4 Conditionals and loops (count, for_each, for)
Chapter 11 · Providers and state
- 11.1 How the AWS provider works
- 11.2 The terraform.tfstate file and its importance
- 11.3 Local state vs remote state (S3 + DynamoDB)
- 11.4 Essential commands: init, plan, apply, destroy, fmt, validate
Chapter 12 · Your first real infrastructure in Terraform
- 12.1 Create a VPC with subnets from scratch
- 12.2 Launch a public EC2 instance
- 12.3 Associate a Security Group and an Elastic IP
- 12.4 Outputs and references between resources
- 12.5 Team workflow: PR review of plans
Chapter 13 · Load balancing and auto scaling
- 13.1 Application Load Balancer vs Network Load Balancer
- 13.2 Target Groups, listeners and rules
- 13.3 Auto Scaling Groups: policies and metrics
- 13.4 Warm pools and lifecycle hooks
Chapter 14 · Serverless with Lambda
- 14.1 The Lambda execution model
- 14.2 Triggers: API Gateway, S3, DynamoDB Streams, SQS
- 14.3 Dependency management and layers
- 14.4 Cold starts and strategies to reduce them
- 14.5 Limits and anti-patterns
Chapter 15 · Messaging and events
- 15.1 SQS: standard vs FIFO queues, DLQ
- 15.2 SNS: topics, subscriptions, fan-out
- 15.3 EventBridge: event buses and rules
- 15.4 Patterns: pub/sub, decoupling, saga
Chapter 16 · Content delivery and DNS
- 16.1 Route 53: record types and routing policies
- 16.2 CloudFront: distributions, caches and origins
- 16.3 ACM: free SSL/TLS certificates
- 16.4 WAF integrated with CloudFront
Chapter 17 · Containers on AWS
- 17.1 Docker: quick review of key concepts
- 17.2 ECR: private image registry
- 17.3 ECS: task definitions, services, Fargate vs EC2
- 17.4 EKS: when Kubernetes and when not
Chapter 18 · Modules: reuse and composition
- 18.1 Anatomy of a Terraform module
- 18.2 Input variables, outputs and dependencies
- 18.3 Local modules vs Terraform Registry modules
- 18.4 Module versioning with Git tags
- 18.5 Design of generic vs domain-specific modules
Chapter 19 · Workspaces and environment management
- 19.1 Terraform workspaces: use cases and limitations
- 19.2 Directory strategy per environment (dev/stg/prod)
- 19.3 Terragrunt: DRY for environment configurations
- 19.4 Environment variables and .tfvars files
Chapter 20 · Remote backends and locking
- 20.1 Configure S3 + DynamoDB as backend
- 20.2 State locking: avoiding team corruption
- 20.3 State migration between backends
- 20.4 terraform import: bring existing resources into state
Chapter 21 · Infrastructure testing
- 21.1 Terraform validate and fmt in CI
- 21.2 Checkov and tfsec: static security analysis
- 21.3 Terratest: integration tests in Go
- 21.4 Contract testing between modules
Chapter 22 · Terraform in CI/CD
- 22.1 Basic pipeline: lint → plan → apply in GitHub Actions
- 22.2 Atlantis: GitOps for Terraform
- 22.3 Terraform Cloud / HCP Terraform
- 22.4 Drift detection and automatic reconciliation
Chapter 23 · Defense in depth
- 23.1 AWS Organizations and Service Control Policies
- 23.2 AWS Config: continuous compliance
- 23.3 GuardDuty: threat detection
- 23.4 Security Hub: centralized view
- 23.5 KMS: key management and rotation
- 23.6 Secrets Manager vs Parameter Store
Chapter 24 · Observability: logs, metrics and traces
- 24.1 CloudWatch Logs, metrics and alarms
- 24.2 CloudWatch Dashboards and Contributor Insights
- 24.3 X-Ray: distributed tracing
- 24.4 OpenTelemetry on AWS
- 24.5 Managed Grafana and Managed Prometheus
Chapter 25 · Cost optimization
- 25.1 AWS Cost Explorer and budgets with alerts
- 25.2 Trusted Advisor and Compute Optimizer
- 25.3 Rightsizing: how to detect overprovisioning
- 25.4 Savings Plans vs Reserved Instances: strategic decision
- 25.5 FinOps: culture and processes to control spending
Chapter 26 · High availability and disaster recovery
- 26.1 RTO and RPO: defining objectives
- 26.2 Strategies: backup/restore, pilot light, warm standby, multi-site
- 26.3 Route 53 health checks and automatic failover
- 26.4 AWS Backup: centralized backup policy
Chapter 27 · AWS Well-Architected Framework
- 27.1 The six pillars: operational excellence, security, reliability, performance efficiency, cost optimization, sustainability
- 27.2 Well-Architected Tool: formal reviews
- 27.3 How to apply the framework in design decisions
Chapter 28 · Serverless architectures at scale
- 28.1 Event-driven architecture with Lambda + EventBridge
- 28.2 Saga pattern for distributed transactions
- 28.3 Step Functions: orchestration of complex workflows
- 28.4 Lambda@Edge and CloudFront Functions
Chapter 29 · Data platforms on AWS
- 29.1 Data Lake with S3, Glue and Athena
- 29.2 Kinesis Data Streams and Firehose for streaming
- 29.3 Redshift: data warehousing at scale
- 29.4 Lake Formation: data governance
Chapter 30 · Multi-account and landing zones
- 30.1 Why separate workloads into different accounts
- 30.2 AWS Control Tower and Account Factory
- 30.3 Centralized log and security management
- 30.4 Terraform at multi-account scale with shared modules
Chapter 31 · Platform Engineering and Internal Developer Platform
- 31.1 Golden paths and abstractions over Terraform
- 31.2 AWS Service Catalog
- 31.3 Backstage as a developer portal
- 31.4 Terraform modules as internal product
Chapter 32 · Relevant AWS certifications
- 32.1 Cloud Practitioner: is it worth it?
- 32.2 Solutions Architect Associate → Professional
- 32.3 DevOps Engineer Professional
- 32.4 Specialty: Security, Database, Networking
- 32.5 HashiCorp Terraform Associate
Chapter 33 · Projects to consolidate what you've learned
- 33.1 Project 1: serverless blog (S3 + CloudFront + Lambda + DynamoDB)
- 33.2 Project 2: REST API with ECS Fargate + RDS + ALB
- 33.3 Project 3: data platform with Glue + Athena + Redshift
- 33.4 Project 4: multi-account landing zone with Terraform and Control Tower
