Skip to content

Independent marketplace for verified cloud accounts

Multi-Factor Authentication Across Cloud Providers

MFA is the single highest-value control on a cloud login, but only if recovery is planned. This guide compares factor types, enrolment order and what to do when a device is lost.

If you do one thing to a new cloud account before putting anything on it, make it this. A second authentication factor on the console login defeats the overwhelming majority of credential attacks, because those attacks run on passwords harvested elsewhere and have nothing else to offer.

The part that gets skipped is recovery. Enrolling a factor takes two minutes. Deciding what happens when the phone holding that factor falls in a river takes longer, and the teams that skip it discover the gap at the worst possible moment.

Why a password alone fails here

Cloud console credentials have a property that makes them unusually attractive: they are reachable from anywhere, they control resources that cost money to run, and compromise is profitable immediately. An attacker does not need to understand your business to monetise an account, because compute itself is the prize.

Password reuse is the main delivery mechanism. A breach at an unrelated service produces an address and a password, automation tries that pair across cloud providers, and the only thing that stops it is a factor the attacker does not hold. Password complexity rules do not help, because the password was not guessed.

The factor types and how they differ

Not all second factors are equal, and the differences matter more than vendors imply.

Factor Resists phishing Main weakness
SMS code No Interceptable through number porting and relayed to a fake site in real time
Email code No Only as strong as the mailbox, which is often the account recovery channel too
Authenticator app code No A convincing fake login page can collect and replay the code within its window
Push approval Partly Fatigue: repeated prompts eventually get approved
Hardware security key Yes Physical loss, and a second key is needed as backup
Platform authenticator or passkey Yes Tied to a device or ecosystem, and recovery varies

The phishing resistance column is the one that separates the list. Hardware keys and passkeys bind the authentication to the domain the browser is actually talking to, so a lookalike site receives nothing it can reuse. Codes of any kind can be read aloud to an attacker by a convincing caller, and regularly are.

That said, an authenticator app on every login is dramatically better than a hardware key on one login and nothing elsewhere. Coverage first, then strength.

Which login to enrol first

Start with the account owner or root identity, the one tied to billing. It is the identity that can undo every other control, and it is usually the one still protected by a password alone because nobody signs in with it day to day.

Enrol a hardware key on it if you can, and a second hardware key stored in a different physical location. Then work outward: administrators next, then anyone who can modify infrastructure, then everyone else. Service identities and automation are a separate problem entirely, since they do not have a human to present a factor, and they belong in the model described in IAM fundamentals covering users, roles and least privilege.

If the account arrived from someone else rather than being created by you, the ordering changes. Assume any existing factor could belong to a previous holder. Remove every enrolled device you did not personally register, enrol your own, and only then continue. The full sequence is in the post-handover security checklist for a cloud account.

Recovery codes and the backup factor problem

Nearly every provider issues a set of single-use recovery codes at enrolment. These are the bypass for the control you just enabled, which makes them as sensitive as the password itself.

Print them or write them down, and store them physically where the people who would need them can reach them without the person who enrolled. A photograph in a phone gallery, a note in a chat thread or a file on the desktop all defeat the purpose. If your organisation uses a password manager with a shared vault, that is an acceptable home as long as the vault itself is not protected by the same factor you are backing up.

Two rules save a lot of grief. Never let one person be the only holder of a factor for a shared account. And test recovery before you need it, by deliberately using one code in a controlled moment and confirming the process works and regenerating the set afterwards.

What to do when a device is lost

Move fast, in this order. Revoke the lost factor from the account first, using a second administrative identity rather than the recovery codes if one exists. Terminate active sessions, because revoking a factor does not sign out a browser that is already authenticated. Enrol a replacement. Then regenerate recovery codes, since you cannot be sure whether the old set is still private.

After that, check the audit trail for the period between the loss and the revocation. A sign-in from an unfamiliar location or an unusual action during that window changes this from an inconvenience into an incident, and the response is covered in detecting and responding to unexpected cloud activity.

The reason a second administrative identity matters is this exact scenario. An account with one administrator and one factor has a single point of failure that a lost phone is enough to trigger.

Team members, contractors and shared access

Every person gets their own identity and their own factor. This sounds obvious and is routinely violated, usually by a small team sharing one login because setting up individual access felt like overhead.

Shared logins break MFA in a specific way: the factor has to live somewhere, so it ends up on one person’s phone, and that person becomes a bottleneck or the code gets forwarded in a chat message. Both outcomes are worse than no MFA at all, because they create the appearance of control.

For contractors, enrol a factor on their own identity and set a deliberate end date for the engagement. When the engagement finishes, disable the identity rather than only removing the factor. The guide to handing a cloud account over to a client covers the reverse direction, where you are the one leaving.

Where MFA does not help

MFA protects an interactive login. It does nothing for several other paths into your account, and believing otherwise produces false confidence.

  • API keys and access tokens. They authenticate without a factor by design. A leaked key is a full bypass, which is why long-lived keys should be rare and short-lived credentials should be the default.
  • Session hijacking. A stolen session cookie represents an already-authenticated session. Shorter session lifetimes reduce the window.
  • Insider misuse. A legitimate user with a valid factor doing something they should not is an authorisation problem, not an authentication one.
  • Billing and support channels. Some account changes can be made through support processes that verify identity differently. Know what yours allows.

This is why MFA sits alongside audit logging rather than replacing it. If you want to know what an authenticated identity did, you need the record, which is the argument in audit logging and activity monitoring from day one.

Differences between providers worth planning around

The underlying standards are shared, so an authenticator app or a hardware key works across platforms. What differs is enforcement and administration.

Some providers let an organisation administrator require a factor for every identity and block access without one. Others leave enrolment to the individual, which means your policy is only as good as your follow-up. Some support multiple hardware keys per identity, some allow only one, and that single detail determines whether you can hold a backup key at all. Recovery processes vary widely in how much human review they involve.

Check these specifics in the provider’s current documentation before you design a rollout, because they change. Platforms with a strong enterprise identity heritage tend to offer the most administrative control, which the AWS vs Azure comparison explores in the wider context of their identity models. If centralised directory control is what you need, a Buy Azure Account listing fits that pattern, while a Buy Gsuite Account is often the identity anchor for teams already working in that ecosystem.

A rollout order that does not lock anyone out

Announce the change with a date and a short guide. Enrol administrators first and confirm every one of them has a working backup method. Make enrolment available to everyone else and leave it optional for a defined window, tracking who has done it. Then enforce it, with a documented path for anyone who is locked out on the first morning, because there will be someone.

Keep one break-glass identity with a factor held separately from the rest, monitored so that any use of it raises an alert. Review enrolment status at the same cadence as your access review, since new joiners and new devices quietly erode coverage over time.

Related comparisons

Azure vs UpCloud
vs

Azure vs UpCloud

UpCloud is a Finnish provider built around fast storage and a small, clear product line. Azure is a Windows centric…

Azure vs Vultr
vs

Azure vs Vultr

Vultr sells raw capacity in many locations with almost no ceremony. Azure sells governance, policy and scale. Here is how…

AWS vs Azure
vs

AWS vs Azure

Azure is usually chosen through identity and licensing rather than through infrastructure. This comparison covers enterprise agreements, hybrid estates, hierarchy…

Azure vs Hetzner
vs

Azure vs Hetzner

Hetzner sells capacity at a price Azure does not try to match. This comparison covers what the gap actually buys…

Accounts discussed in this guide

Buy Azure Account
Buy Azure Account
Microsoft Azure

Buy Azure Account

Microsoft Azure has firmly established itself as one of the world's leading cloud computing platforms, trusted by enterprises, government agencies, and developers across the…

From$25 13 options

Keep reading

What Cloud Credits Cover, and What They Never Cover
Cost Management 7 min read

What Cloud Credits Cover, and What They Never Cover

A credit is a conditional, time limited, provider controlled entitlement, not money. What a balance actually buys, what it refuses to buy,…

Need the account, not just the theory?

Browse documented configurations across every provider in the catalog, or ask us which one fits your workload.

Telegram