CYBER SECURITY • CLOUD AND MODERN INFRASTRUCTURE SECURITY

Shared Responsibility Model — Explain shared responsibility model conceptually

Understanding who secures what in cloud environments prevents critical gaps in defense.

Historical Context & Motivation

Before cloud computing became ubiquitous, organizations owned and operated every layer of their IT infrastructure—from the physical servers in their data centers to the applications running on those servers. Security was conceptually straightforward: you owned everything, so you secured everything. When enterprises began migrating workloads to cloud platforms, a fundamental question emerged: who is responsible for securing which components? The traditional perimeter-based security model, in which an organization protected its own network boundary, became insufficient because infrastructure was now distributed across third-party data centers managed by providers like Amazon, Microsoft, and Google.

The Shared Responsibility Model arose as a conceptual framework to delineate security obligations between cloud service providers (CSPs) and their customers. Without such a framework, organizations frequently assumed the cloud provider handled all security concerns—a dangerous misconception that led to misconfigurations, data breaches, and compliance failures. The model clarifies that while a CSP secures the underlying infrastructure (the "security of the cloud"), the customer must secure their own data, identities, and application configurations ("security in the cloud").

2006
AWS Launches EC2
Amazon Web Services introduces Elastic Compute Cloud, marking the mainstream arrival of Infrastructure as a Service (IaaS). Organizations begin renting virtual machines, raising immediate questions about security jurisdiction over shared hardware.
2010
CSA Guidance v2.1
The Cloud Security Alliance publishes comprehensive guidance for cloud security, formalizing the idea that security responsibilities are split between providers and consumers depending on the service model (IaaS, PaaS, SaaS).
2013
NIST SP 500-292
The National Institute of Standards and Technology publishes its Cloud Computing Reference Architecture, explicitly defining roles and responsibilities across cloud actors, codifying the shared responsibility concept in government standards.
2017–2019
Major Breach Wave
High-profile breaches at Capital One and other enterprises are traced to misconfigured cloud resources. These incidents demonstrate the real-world consequences of misunderstanding shared responsibility—customers failed to secure what was clearly their domain.
2020+
Zero Trust & Cloud-Native Security
The shared responsibility model evolves to incorporate zero-trust architectures, container orchestration (Kubernetes), and serverless paradigms, further complicating the division of duties and demanding more granular frameworks.

The central gap that the Shared Responsibility Model addresses is the ambiguity of ownership in multi-tenant, abstracted environments. When you no longer rack your own servers, patch your own hypervisors, or wire your own switches, the security perimeter fragments—and without a clear framework, critical layers fall between the cracks.

Core Principles & Definitions

The Shared Responsibility Model rests on several foundational ideas that every cloud practitioner must internalize. These principles dictate how security obligations are partitioned, how accountability is established, and how compliance frameworks map onto cloud deployments. Understanding them is prerequisite to correctly implementing security controls in any cloud environment.

1

Division of Control

Security responsibility follows control. Whichever party can configure, patch, or modify a component bears the obligation to secure it. The CSP controls the physical infrastructure and hypervisor; the customer controls their OS instances, data, and access policies.
2

Service Model Dependency

The boundary between provider and customer responsibility shifts depending on the service model. In IaaS, the customer manages more layers (OS, middleware, runtime). In SaaS, the provider manages nearly everything except user data and access management.
3

Security OF vs. IN the Cloud

AWS coined this distinction: the CSP secures the cloud infrastructure itself (hardware, networking, facilities), while the customer secures everything they place within it—data encryption, IAM policies, firewall rules, and application-level controls.
4

Inherited vs. Customer Controls

Some controls are inherited from the provider (e.g., physical security, redundant power). Others are entirely the customer's responsibility (e.g., data classification, endpoint protection). A third category—shared controls—requires both parties to act (e.g., patch management, awareness training).
5

Non-Transferable Accountability

While responsibility for implementing controls can be distributed, regulatory and legal accountability for data protection typically remains with the data owner—the customer. You can outsource operations but not compliance liability.
KEY TAKEAWAY
Think of the Shared Responsibility Model like renting an apartment. The landlord (CSP) is responsible for the building's structural integrity, plumbing, and electrical wiring. The tenant (customer) is responsible for locking the front door, not leaving the stove on, and safeguarding their valuables. If someone breaks in because you left a window open, the landlord is not liable—even though the building has a security system.

Visual Explanation — The Responsibility Stack

The IaaS responsibility stack shows how layers are distributed. Violet layers (C) are the customer's responsibility, amber (S) indicates shared controls, and cyan (P) belongs to the cloud service provider. Notice that the customer manages the majority of the stack in an IaaS model.

The diagram above illustrates the responsibility stack for an IaaS deployment such as AWS EC2 or Azure Virtual Machines. Observe that the customer controls the upper layers—data encryption, application security, identity management, and OS patching—while the provider secures the lower layers including the physical data center, hardware, and hypervisor. The network configuration layer occupies an interesting middle ground: the provider ensures the underlying network fabric is resilient and secure, but the customer must configure their virtual private clouds, security groups, and firewall rules correctly. This layered view is the mental model every cloud security professional should carry.

How the Model Works Across Service Models

The boundary between provider and customer responsibility is not static—it shifts based on the cloud service model in use. The NIST cloud computing reference architecture defines three primary models: Infrastructure as a Service (IaaS), Platform as a Service (PaaS), and Software as a Service (SaaS). As you move from IaaS to SaaS, the provider assumes control over progressively more layers, and the customer's security surface area contracts accordingly. However, the customer's responsibility for data protection, identity governance, and access control persists across all three models.

IaaS — Maximum Customer Responsibility

In an IaaS model (e.g., AWS EC2, Google Compute Engine), the provider delivers raw compute, storage, and networking resources. The customer installs and manages the operating system, middleware, runtime environment, application code, and data. This model provides maximum flexibility but demands that the customer handle OS patching, firewall configuration, intrusion detection, encryption key management, and application-layer defenses. The provider's obligation ends at the hypervisor and the physical network infrastructure.

PaaS — Balanced Responsibility

In a PaaS model (e.g., AWS Elastic Beanstalk, Azure App Service, Heroku), the provider manages the operating system, runtime, and middleware in addition to the infrastructure. The customer focuses on application code, data, and identity/access management. While this reduces the customer's attack surface, it introduces a dependency on the provider's patching cadence for the managed components. Misconfigured application settings, insecure API endpoints, and poor secret management remain squarely in the customer's domain.

SaaS — Minimal Customer Surface

In a SaaS model (e.g., Google Workspace, Salesforce, Microsoft 365), the provider manages nearly everything: infrastructure, platform, application code, and often default security configurations. The customer's residual responsibilities include user provisioning, access control policy, data classification and loss prevention, and ensuring compliance with their organization's regulatory requirements. Even in the most managed model, a customer who grants excessive permissions or fails to enable multi-factor authentication creates exploitable attack surfaces.

Side-by-side comparison showing how the responsibility boundary shifts across IaaS, PaaS, and SaaS. As you move from left to right, the provider's domain expands while the customer's shrinks. Data and IAM remain the customer's obligation in every model.

Detailed Breakdown — Controls by Category

AWS documentation formally categorizes security controls into three types: inherited controls, shared controls, and customer-specific controls. This taxonomy is useful beyond AWS—it applies conceptually to any CSP's implementation of the shared responsibility model. Understanding these categories helps security architects perform gap analyses and map their compliance obligations to the correct party.

Three categories of controls in the shared responsibility framework
Control CategoryDescriptionExamples
Inherited ControlsControls that the customer fully inherits from the CSP. The customer does not need to implement or manage these.Physical and environmental controls (biometric access, fire suppression, power redundancy, hardware decommissioning, facility surveillance)
Shared ControlsControls where the CSP provides the infrastructure-level implementation and the customer must implement their own layer. Both parties contribute.Patch management (CSP patches infrastructure OS; customer patches guest OS). Configuration management (CSP configures infra; customer configures own instances). Security awareness (CSP trains its employees; customer trains theirs).
Customer-Specific ControlsControls that are entirely the customer's responsibility, determined by the application or data requirements.Data encryption (at rest and in transit), service and communications protection (e.g., TLS termination), network firewall rules, zone security, data classification, backup strategies, DLP policies
⚠️ Common Misconfiguration Pitfall
The 2019 Capital One breach involved a misconfigured WAF (Web Application Firewall) on AWS EC2 instances. AWS correctly secured its infrastructure layer, but the customer's firewall configuration allowed a Server-Side Request Forgery (SSRF) attack to access instance metadata and exfiltrate over 100 million records. This is a textbook example of a customer-side failure within the shared responsibility model.

A useful heuristic for determining who is responsible: if you can see it, configure it, or access it through the cloud console or API, it is your responsibility to secure it. If it is abstracted away and invisible to you—like the hypervisor firmware or the physical rack layout—then it falls under the provider's domain. This heuristic covers approximately 90% of cases, though edge cases exist in managed services where the provider exposes partial configuration while retaining operational control.

Worked Example — Mapping Responsibility for a Web Application

Consider a startup deploying a three-tier web application on AWS. The architecture uses EC2 instances (IaaS) for compute, RDS (managed database, PaaS-like) for data persistence, and S3 (managed storage) for static assets. Let us walk through a systematic process to determine who is responsible for each security control.

Security Responsibility Mapping for a Three-Tier Web App on AWS
1
Step 1 — Identify the Service Models in UseEnumerate each cloud service and classify it. EC2 is IaaS—the customer manages the guest OS, middleware, and application. RDS is a managed service (PaaS-adjacent)—AWS manages the database engine patching and OS, but the customer manages database access credentials, security groups, encryption settings, and backup retention policies. S3 is a managed storage service—AWS secures the underlying storage infrastructure, but the customer must configure bucket policies, access control lists (ACLs), and enable encryption.
Three distinct service types → three different responsibility profiles.
2
Step 2 — Map Security Layers to Each ServiceFor the EC2 instances: the customer must harden the AMI (Amazon Machine Image), configure the OS firewall (iptables/nftables), install and maintain intrusion detection (e.g., OSSEC), keep the OS patched, deploy the application securely, and manage TLS certificates. For RDS: the customer must configure security groups to restrict inbound connections, enable encryption at rest (KMS), enforce SSL for connections in transit, set strong master credentials, and manage IAM database authentication. For S3: the customer must disable public access by default, enable versioning and MFA delete for critical buckets, configure server-side encryption (SSE-S3 or SSE-KMS), and set up access logging.
Identified 15+ customer-owned security controls across three services.
3
Step 3 — Identify Inherited ControlsAcross all three services, the customer inherits AWS's physical security, environmental controls, hardware lifecycle management, and network backbone security. These are documented in AWS SOC 2 Type II reports and can be reviewed through AWS Artifact. No action is required from the customer for these controls, but they should verify compliance certifications when onboarding regulated workloads.
Physical security, facility controls, and hardware management → fully inherited from AWS.
4
Step 4 — Address Shared ControlsPatch management is shared: AWS patches the RDS database engine and underlying OS, but the customer must patch the EC2 guest OS and any software installed on it. Configuration management is shared: AWS maintains its infrastructure configuration, but the customer must use tools like AWS Config or Terraform to enforce their own resource configurations. Awareness training is shared: AWS trains its staff, but the customer's developers and operations team need cloud security training.
Three shared control domains identified; customer must implement their portion independently.
5
Step 5 — Document and ValidateCreate a Responsibility Assignment Matrix (RACI) mapping each control to either Provider, Customer, or Shared. Use AWS Well-Architected Tool or a CSPM (Cloud Security Posture Management) tool like Prisma Cloud or AWS Security Hub to continuously validate that customer-side controls are properly configured. Schedule quarterly reviews to account for new services, architecture changes, or updated compliance requirements.
A living responsibility matrix ensures no security gap falls between the cracks as the architecture evolves.

Strengths, Limitations & Cross-Provider Comparison

Strengths vs. limitations of the shared responsibility model
StrengthsLimitations
Provides clear delineation of duties, reducing ambiguity in security ownership.Boundaries can be fuzzy for managed services that blur IaaS/PaaS lines (e.g., AWS Lambda, Azure Functions).
Leverages CSP expertise for infrastructure-layer security that most organizations cannot replicate internally.Customers often overestimate provider coverage—assuming SaaS vendors handle all security when IAM and data governance remain their duty.
Maps well to compliance frameworks (SOC 2, ISO 27001, FedRAMP) that require explicit control ownership.Multi-cloud and hybrid environments create overlapping models that are difficult to harmonize.
Scalable model that works across service types and organizational sizes.The model is descriptive, not prescriptive—it tells you what to secure, not how to secure it.
Cross-provider comparison of shared responsibility implementations
AspectAWSAzureGCP
Terminology"Security OF / IN the cloud""Shared responsibility model" with explicit layered chart per service"Shared fate" — emphasizes collaborative partnership over rigid boundaries
DocumentationWell-Architected Framework, Shared Responsibility webpageMicrosoft Learn documentation, Trust CenterGoogle Cloud Architecture Framework, Shared Fate whitepaper
Unique FeatureAWS Artifact for compliance report access; Security Hub for posture managementAzure Defender integrated natively; Compliance Manager for regulatory mappingGoogle positions itself as an active partner with built-in guardrails (e.g., BeyondCorp zero trust)
KEY TAKEAWAY
Google Cloud's evolution from "shared responsibility" to "shared fate" reflects an industry trend toward providers taking a more active role in helping customers implement their side of the model—through opinionated defaults, embedded security tooling, and prescriptive guidance. The underlying division of control, however, remains fundamentally the same across all three hyperscalers.

Connection to Advanced Frameworks & Emerging Paradigms

The Shared Responsibility Model provides a foundational framework, but modern cloud architectures introduce complexities that demand extensions and refinements. As organizations adopt containers, serverless functions, and multi-cloud strategies, the clean layer-by-layer model becomes more nuanced. Understanding these advanced contexts prepares you for real-world cloud security engineering.

Traditional vs. emerging responsibility paradigms
Traditional Shared ResponsibilityEmerging Paradigm
Static layer boundaries per service model (IaaS/PaaS/SaaS)Dynamic boundaries: Kubernetes shifts responsibility depending on whether you use managed control planes (EKS, GKE) vs. self-hosted
Customer manages all application-layer securityServerless (Lambda, Cloud Functions) abstracts the runtime; provider manages patching, scaling, and OS, but customer still owns function code, IAM, and data
Single-provider model with clear documentationMulti-cloud and hybrid deployments require harmonized responsibility matrices across providers with different terminologies and boundaries
Perimeter-based trust: trusted inside, untrusted outsideZero Trust Architecture: never trust, always verify — both provider and customer implement continuous verification, pushing identity-centric security to every layer
Manual compliance mapping via spreadsheets and auditsCSPM tools (Prisma Cloud, Wiz, Orca) automate continuous assessment of customer-side misconfigurations against the shared responsibility model

Looking forward, the rise of AI/ML workloads in the cloud introduces yet another dimension: who is responsible for model integrity, training data privacy, and inference endpoint security? Cloud providers are beginning to publish shared responsibility guidance specific to AI services (e.g., AWS SageMaker, Azure OpenAI Service), but this is an active area of development. As a cloud security professional, you should expect the shared responsibility model to continuously evolve as abstraction layers multiply and new service categories emerge.

Practice Problems

PROBLEM 1CONCEPTUAL
Explain the distinction between "security OF the cloud" and "security IN the cloud." For each phrase, provide two specific examples of security controls that fall under that category.
PROBLEM 2BASIC CALCULATION
A company uses 5 AWS services: EC2 (IaaS), RDS (managed PaaS), S3 (managed storage), Lambda (serverless), and SES (email SaaS). For each service, classify the "OS patching" control as Provider, Customer, or Not Applicable, and briefly justify your classification.
PROBLEM 3INTERMEDIATE
An organization deploys a Kubernetes cluster using Amazon EKS (managed control plane) with self-managed worker nodes running on EC2. Construct a responsibility matrix for the following controls: (a) Kubernetes API server patching, (b) worker node OS patching, (c) pod network policies, (d) container image vulnerability scanning, (e) etcd encryption. Indicate whether each is Provider, Customer, or Shared.
PROBLEM 4APPLIED
A healthcare startup stores Protected Health Information (PHI) in an Azure Blob Storage account. A security auditor finds that the storage account allows public anonymous access. The startup's CTO argues that Azure, as the cloud provider, should have prevented this misconfiguration. Evaluate the CTO's argument using the shared responsibility model and HIPAA compliance requirements. Who is at fault, and what controls should have been implemented?
PROBLEM 5CRITICAL THINKING
Google Cloud introduced the concept of "shared fate" as an evolution of the shared responsibility model. Critically analyze this framing: Does "shared fate" represent a genuine shift in how security responsibilities are distributed, or is it primarily a marketing rebranding? Support your argument with at least three specific technical or operational differences (or lack thereof) between traditional shared responsibility and shared fate.

Lesson Summary

The Shared Responsibility Model is the foundational framework for understanding security obligations in cloud computing. It establishes that the cloud service provider secures the underlying infrastructure—physical facilities, hardware, hypervisors, and the global network fabric ("security of the cloud")—while the customer secures their data, applications, identity and access management, and configurations ("security in the cloud"). The boundary between these domains shifts with the service model: IaaS places the most responsibility on the customer, PaaS distributes it more evenly, and SaaS leaves the customer responsible primarily for data and access policies.

Security controls are categorized as inherited (fully managed by the provider), shared (both parties contribute), or customer-specific (entirely the customer's domain). Emerging paradigms such as serverless computing, container orchestration, and zero trust architectures introduce more granular and dynamic responsibility boundaries. Regardless of how the model evolves, one principle endures: you can outsource operations, but you cannot outsource accountability.

Varsity Tutors • Cyber Security • Shared Responsibility Model