Identity is the new perimeter
The chapter opens with the root cause of the fictional company's breach: a marketing contractor's account with a reused password and no second factor. The attacker signed in normally and reached customer data through a reporting dashboard. The security lead summarises it as "We did not get hacked. We got logged into."
With people working remotely, data in cloud services and applications reached over the internet, identity is what decides each access request. Three separate questions matter. Authentication asks whether you are who you claim to be. Authorisation asks what you may do. Accounting records what the identity actually did. They fail independently, and the contractor incident was a failure of both the first and the second.
MFA everywhere, and phishing-resistant where it counts
If the reader does only one thing from the book, the author asks that it be MFA on every account, with any exception justified and documented. Everywhere means closing the paths that bypass the second factor: legacy authentication protocols, APIs that accept only a password, and service accounts exempted long ago.
Factors differ in strength. SMS codes are better than nothing but can be intercepted or socially engineered. Authenticator apps with number matching are better and defeat MFA fatigue attacks, where a user is flooded with prompts until they approve one. The strongest option is phishing-resistant MFA based on FIDO2, such as security keys and passkeys, which are cryptographically bound to the genuine site's origin and will not sign in to a look-alike domain.
The recommended path is MFA for everyone now using authenticator apps, then phishing-resistant factors for administrators and high-risk users first. The expert section adds two cautions: session tokens can still be stolen after sign-in, so pair strong MFA with short sessions and device compliance, and harden the account recovery path, because attackers will target a weaker fallback.
SSO and conditional access
Single sign-on through a central identity provider such as Entra ID means fewer passwords, one place to enforce MFA, one place to disable a leaver, and one place that sees every sign-in. The same thinking applies to Google Workspace, Okta, JumpCloud or Ping under different names.
Conditional access evaluates the context of each sign-in, including user, application, device, location and risk, and then allows, requires more proof, demands a compliant device or blocks. A small, well-understood set of policies carries most of the value: MFA for everyone, legacy authentication blocked, compliant devices for sensitive applications, and stricter handling of risky sign-ins. Policies should be kept as reviewed configuration, tested in report-only mode, rolled out in stages and always exclude break-glass accounts. In the book's example, report-only logs revealed that shared factory PCs would have been locked out at shift change, so they were enrolled first.
Least privilege, RBAC and privileged access
Role-based access control makes least privilege manageable: define roles for jobs, assign people to roles, and prefer narrow built-in or tightly scoped custom roles over broad owner or administrator roles. The chapter shows a custom Azure role limited to backup operations.
Administrative rights should be just-in-time rather than standing, with elevation requested for a limited window and logged, as Microsoft's Privileged Identity Management provides. Break-glass accounts are a small number of emergency administrators excluded from conditional access, protected by long unique passwords stored offline, split or vaulted so no one person holds them, and alerted on any use. The author stresses reducing friction so that secure elevation is easy; otherwise people quietly restore standing admin rights.
The account lifecycle and non-human identities
Joiners should get access from a role, not by copying a colleague. Movers must lose old access when they gain new access, or privilege creep follows. Leavers should be disabled promptly, ideally driven automatically from the HR system, by disabling the account and revoking sessions rather than deleting it. Contractor and vendor access should carry an expiry date from the start.
Service accounts, app registrations, API keys and pipeline credentials are often the most over-privileged and least watched identities. Scope them tightly, prefer managed identities or workload roles that use short-lived credentials, keep unavoidable secrets in a secret manager and rotate them. Regularly review accounts with no sign-in for 90 days or more and the members of every privileged directory role.
KEY POINTS
Key points
- Confirm MFA is required for every account and close every exemption and legacy authentication path.
- Give administrators phishing-resistant MFA first, then high-risk users, then everyone.
- Deploy a small set of conditional access policies in report-only mode, read the logs, then enforce in stages.
- Replace standing global admin rights with just-in-time elevation and create monitored break-glass accounts.
- Automate leaver offboarding from HR so accounts are disabled within hours, not deleted.
- Run a stale-account query for 90 days without sign-in and review every privileged role membership.
QUESTIONS
Common questions
- What is phishing-resistant MFA and why is it better?
- Phishing-resistant MFA uses FIDO2 security keys or passkeys whose cryptographic keys are bound to the genuine website's origin. Unlike SMS codes or app approvals, which a user can be tricked into relaying through a fake page, a passkey will not produce a valid sign-in for a look-alike domain. The Hardened Stack recommends rolling it out to administrators and high-risk users first.
- Which conditional access policies should a small organisation start with?
- The Hardened Stack recommends a small baseline: require MFA for all users, block or tightly restrict legacy authentication, require compliant or managed devices for sensitive applications, and raise requirements or block sign-ins flagged as risky. Run each policy in report-only mode first, enforce in stages, and always exclude break-glass accounts so a mistake cannot lock you out.
- How do I protect against MFA fatigue attacks?
- MFA fatigue is when an attacker with a valid password sends repeated approval prompts until the user accepts one. Turning on number matching, where the user must type a number shown on the sign-in screen, stops accidental approvals, and phishing-resistant factors such as passkeys avoid the problem entirely.
- What is a break-glass account and how should it be protected?
- A break-glass account is a small number of emergency administrator accounts kept for when normal sign-in fails, for example when conditional access misfires. They are excluded from conditional access, so they need very long unique passwords stored offline, credentials split or vaulted so no single person holds them, and an immediate alert on any use.
- What is the right way to offboard a leaver in Entra ID?
- Disable the account rather than delete it, revoke its sign-in sessions so existing tokens stop working, and log the action for the audit trail. Ideally this is triggered automatically from the HR system within hours of the person leaving. Contractor and vendor access should be given an expiry date when it is created so it cannot be forgotten.
- How should service accounts be secured?
- Scope each service account to the minimum it needs, prefer platform-managed identities such as Azure managed identities or AWS roles that issue short-lived credentials, and keep any unavoidable secret in a secret manager with rotation. Inventory and review non-human identities like any other asset, including checking whether any have crept into privileged roles.