Explore our new Cybersecurity Terms Glossary, explained by our experts.

View the Glossary View the Glossary
PRIVIA

Cybersecurity Terms Glossary

What Are IaaS, PaaS and SaaS?

IaaS, PaaS, and SaaS describe how much of the stack a cloud provider operates. The model you choose decides which security controls remain yours and which you inherit.

Illustration for the cybersecurity term Cloud Service Models
Home / Blog / Cybersecurity Terms / What Are IaaS, PaaS and SaaS?
17 November 2026

What Are IaaS, PaaS and SaaS?

What Are IaaS, PaaS and SaaS?

Infrastructure as a service, platform as a service, and software as a service describe how much of the computing stack somebody else operates on your behalf. NIST Special Publication 800-145 defined the three in 2011 and the definitions have held up well. The security consequence is simple to state and easy to forget: the model determines which controls you inherit from the provider and which remain entirely yours.

The Three Models

Infrastructure as a service gives you virtual machines, storage, and networking. The provider runs the physical hardware, the facility, and the virtualisation layer. You own the operating system upwards: patching, hardening, access control, network rules, and the application itself. Amazon EC2, Azure Virtual Machines, and Google Compute Engine are the obvious examples.

Platform as a service adds the runtime. You deploy code or a database schema and the provider handles the operating system, the middleware, and the scaling. AWS Lambda, Azure App Service, Google App Engine, and managed database services belong here. You keep responsibility for the application, its dependencies, its data, and the identity configuration around it.

Software as a service is a finished application. Microsoft 365, Salesforce, Workday, and several thousand others. The provider runs everything; you configure it, decide who may use it, and remain responsible for the data you put into it.

Shared Responsibility in Practice

LayerIaaSPaaSSaaS
Data and classificationYouYouYou
Identity and accessYouYouYou
Application codeYouYouProvider
Operating system and runtimeYouProviderProvider
Virtualisation and hardwareProviderProviderProvider

Two rows never move. Data and identity are yours in every model, which is why the majority of cloud incidents involve a misconfigured permission or an exposed store rather than a provider failure. Under UK GDPR you remain the controller regardless of who operates the servers, and the ICO will ask what technical and organisational measures you applied—not what your supplier’s certificate says.

What Goes Wrong in Each

In IaaS, the classic failure is an unpatched or over-permissioned virtual machine. The 2019 Capital One breach began with a misconfigured firewall on an EC2 instance and an over-privileged role, as described in our entry on the IMDS attack. Storage left public is the other recurring theme.

In PaaS, the attack surface moves to the application and its dependencies. Serverless functions still execute your code, including the vulnerable library you have not updated, and they frequently run with permissions far wider than the task needs. Managed databases exposed to the internet with weak authentication remain depressingly common.

In SaaS, almost everything comes down to configuration and identity: external sharing left open, over-permissive OAuth grants to third-party applications, legacy authentication protocols that bypass multi-factor authentication, and administrator accounts without strong protection. The controls that help here are described in CASB and in our entry on Google Workspace security configuration.

Choosing a Model Is a Security Decision

Moving up the stack transfers operational burden to the provider, which generally improves security for organisations without large platform teams. A managed database is patched more reliably than one your team maintains between other priorities. The trade is control and visibility: you cannot install an agent on a serverless runtime, you get the logs the provider chooses to expose, and your exit options narrow, as discussed in cloud exit strategy.

Before adopting any model, ask for the shared responsibility matrix in writing, confirm which logs you can export and for how long, check how identity federates and whether conditional access applies, establish where data is stored and processed, and find out how you get your data back. Those five answers predict most of the security work ahead of you.

At Privia Security, we assess cloud architectures across all three models through our cloud security services.

FAQ – Frequently Asked Questions

Which Model Is Most Secure?

None of them inherently. SaaS and PaaS remove whole categories of work, such as operating system patching, which usually helps organisations without a dedicated platform team. They also reduce your visibility and control, and they concentrate risk in configuration and identity, where most cloud breaches actually occur. The secure choice is the model your team can operate well, properly configured and monitored.

Who Is Responsible if a SaaS Provider Is Breached?

Both of you, in different ways. The provider is responsible for the platform and, as a processor under UK GDPR, for the obligations set out in your contract. You remain the controller for the personal data involved, which means you must assess the breach, decide on ICO notification, and inform affected individuals where required. Due diligence and contractual terms agreed in advance determine how smoothly that works.

Is Serverless Computing IaaS or PaaS?

Functions as a service, such as AWS Lambda or Azure Functions, is usually classed as a specialised form of platform as a service. The provider runs the operating system, runtime, and scaling, and you supply code. The security implications follow: no host hardening to do, but application dependencies, function permissions, event source validation, and secrets handling are entirely yours.

Do I Need to Back Up SaaS Data?

Usually yes. Providers protect their platform and offer retention features, but those are designed for availability and compliance holds, not for recovering from malicious deletion or long-undetected corruption. If losing a year of documents, tickets, or records would be material to the business, arrange a backup you control and test a restore from it.

Related Terms

You May Be Interested In These

What Is Telecom Fraud and Revenue Leakage?

Cybersecurity Terms

17 November 2026

What Is Telecom Fraud and Revenue Leakage?

Read More
What Is Single Sign-On (SSO)?

Cybersecurity Terms

17 November 2026

What Is Single Sign-On (SSO)?

Read More
What Is a Proof-of-Concept Exploit?

Cybersecurity Terms

17 November 2026

What Is a Proof-of-Concept Exploit?

Read More