OneNym.idDiscuss deployment
English

THE HARDENED STACK / Appendix A

The IT Security Hardening Checklist

A long list of security controls is only useful if a small IT team can work through it in order and keep returning to it. This appendix gathers the book's actions into one audit you can run across the whole estate.

This chapter is published in full.

This is the consolidated checklist the rest of the book builds toward. It is meant to be worked through, not admired. Each item is a one-line action; where the book devotes a whole chapter to the reasoning behind an item, I will not repeat the reasoning here. Treat this as an audit you can run across the whole estate in a single pass, ticking what is done, flagging what is not, and turning the gaps into tickets.

A word on how to use it. Do not try to do everything at once; a two-person team that attempts a hundred changes in a week will break something and lose the room. Work domain by domain, and within each domain do the high-leverage items first: the ones that close the widest gap for the least effort. If you want a sequenced path rather than a flat list, Appendix B turns this checklist into a phased plan. And remember that "done" for most of these is not a one-time state but a recurring one; a stale checklist is worse than none, because it tells you that you are covered when you are not.

A note on scope. No small team does all of this, and the checklist is deliberately longer than any one organisation needs so that you can choose. If you are overwhelmed by its length, use the shortcut the whole book keeps returning to: of everything here, the handful that prevents the most harm for the least effort is universal MFA, reliable and tested backups with one copy out of reach, prompt patching of what is exposed, removal of standing administrative privilege, and enough logging to investigate. If you did only those and did them well, you would be ahead of most organisations your size. Everything else on this list extends and deepens that core; none of it replaces it.

Identity and access

  • Require multi-factor authentication for every user, with no standing exceptions.
  • Move administrators and other high-value accounts to phishing-resistant MFA (passkeys or FIDO2 security keys), not SMS.
  • Block or retire legacy authentication protocols that cannot enforce MFA.
  • Configure conditional access so that risk, device state, and location shape every sign-in, not just the password.
  • Create separate, named administrative accounts; never administer from a day-to-day user account.
  • Establish break-glass accounts, exclude them from conditional access deliberately, store their credentials offline, and alert on their use.
  • Enforce least privilege: grant the minimum role needed, and review grants on a schedule.
  • Turn on just-in-time privileged access (PIM or equivalent) so admin rights are activated for a window, not held permanently.
  • Inventory every service account and non-human identity, and record an owner for each.
  • Replace long-lived service-account passwords and API keys with managed identities or short-lived credentials wherever the platform allows.
  • Define a joiner-mover-leaver process and automate the leaver step so access dies the day employment does.
  • Run a recurring review for stale accounts, orphaned accounts, and over-privileged accounts, and disable before deleting.
  • Set a sane password policy aligned to current guidance: length over forced complexity, screen against known-breached passwords, and drop mandatory periodic rotation for user passwords.
  • Protect self-service password reset and MFA registration so an attacker cannot simply enrol their own second factor.
  • Log all privileged activity and alert on new admin assignments.
  • Require re-authentication for sensitive administrative actions rather than relying on a long-lived session.
  • Monitor for new MFA method registrations and mailbox forwarding rules, both common signs of account takeover.

Endpoints

  • Enrol every laptop, desktop, and mobile device in management (Intune or equivalent); an unmanaged device is an unknown device.
  • Apply a hardening baseline to each platform rather than shipping defaults.
  • Patch operating systems on a defined cadence, and define a separate, faster lane for emergency patches.
  • Patch third-party software too; the browser, the PDF reader, and the runtime are as exposed as the OS.
  • Deploy endpoint detection and response (EDR), and confirm it is actually reporting, not merely installed.
  • Remove local administrator rights from standard users, and manage the local admin account with a rotation tool such as LAPS.
  • Turn on full-disk encryption everywhere and escrow the recovery keys centrally.
  • Enable application control or allow-listing on your most exposed and most sensitive machines first.
  • Configure the host firewall and disable services and ports you do not use.
  • Set screen-lock timeouts and require re-authentication after idle.
  • Separate BYOD from corporate data using managed app policies rather than full device control where you cannot own the device.
  • Build golden images or configuration profiles so new devices arrive hardened, not hardened later.
  • Track patch and compliance status on a dashboard, and chase the machines that fall out of compliance.
  • Control removable media and USB storage where your data sensitivity warrants it.
  • Harden browsers and office applications through policy: govern extensions, block internet-sourced macros, and keep both on automatic update.

Network

  • Segment the network so a foothold in one zone does not reach everything; isolate production, corporate, guest, and management traffic.
  • Put management interfaces on a dedicated, tightly restricted segment, never on the general LAN.
  • Replace broad allow rules with least-privilege firewall rules, and document why each rule exists.
  • Control east-west traffic between internal zones, not just north-south at the perimeter.
  • Move remote access from flat VPN toward identity-aware access (ZTNA) where you can, and require MFA on whatever remote access remains.
  • Filter DNS to block known-malicious domains, and log DNS queries for later investigation.
  • Disable unused wireless features, separate guest Wi-Fi from corporate, and use strong enterprise authentication on corporate wireless.
  • Change default credentials on every network device and appliance, without exception.
  • Keep network device firmware patched on a schedule.
  • Monitor for new or unexpected devices appearing on the network.

Cloud (Azure and AWS)

  • Turn on the cloud provider's security posture and threat-detection services at the account or tenant level.
  • Enable audit logging (CloudTrail, Azure Activity, and resource-level diagnostic logs) and ship it somewhere tamper-resistant.
  • Remove long-lived access keys in favour of roles and workload identity federation (OIDC).
  • Enforce MFA on the root or global-administrator accounts and lock those credentials away.
  • Apply least-privilege IAM policies and review them for wildcards and unused permissions.
  • Use organisation-level guardrails (service control policies, Azure Policy) to prevent whole classes of misconfiguration.
  • Scan continuously for public storage, open security groups, and unencrypted resources, and remediate.
  • Encrypt data at rest with managed keys, and record which key protects what.
  • Tag every resource with owner, environment, and data sensitivity so you know what you are protecting.
  • Separate production from non-production into different accounts or subscriptions.
  • Define infrastructure as code and run policy-as-code checks in the pipeline before deployment.
  • Set budget and anomaly alerts; a sudden spend spike is sometimes the first sign of compromise.

Microsoft 365 and Entra ID

  • Use Secure Score as a map, not a target, and work the highest-impact recommendations first.
  • Disable legacy authentication across the tenant.
  • Apply a conditional-access baseline covering MFA, device compliance, and risky sign-ins.
  • Put admin roles behind PIM and keep the number of permanent global admins in the low single digits.
  • Restrict external and guest sharing to what the business genuinely needs, and review guest accounts regularly.
  • Control application consent so users cannot grant risky OAuth permissions to arbitrary apps; require admin review.
  • Configure Defender policies for email, endpoints, and cloud apps at a practical, maintainable level.
  • Turn on unified audit logging and set a retention period that matches your investigation needs.
  • Apply basic data-loss-prevention and sensitivity labelling to your most sensitive content.
  • Alert on risky sign-ins, impossible travel, and new mail-forwarding rules.

Email

  • Publish an SPF record that lists only your legitimate senders and ends in a hard fail.
  • Sign outbound mail with DKIM for every sending domain and subdomain.
  • Roll out DMARC and progress it deliberately from monitoring to quarantine to reject.
  • Read your DMARC aggregate reports and fix legitimate senders before you tighten enforcement.
  • Enable anti-phishing and anti-spoofing protection, including impersonation protection for your senior people and key domains.
  • Turn on attachment and link protection (detonation and time-of-click rewriting) if your platform offers it.
  • Give users a one-click way to report suspicious mail, and make sure those reports reach someone.
  • Lock down automatic external mail forwarding, a common sign of a compromised mailbox.
  • Treat inbound filtering as one layer, not the whole defence; pair it with user awareness (see Book 3).

Data and backup

  • Classify data before you protect it, so effort flows to what matters most.
  • Encrypt sensitive data at rest and enforce encryption in transit everywhere.
  • Apply the 3-2-1-1-0 rule: three copies, two media, one off-site, one immutable or offline, and zero errors on restore test.
  • Make at least one backup copy immutable or air-gapped, so ransomware cannot encrypt or delete it.
  • Protect and monitor the backup system's own credentials; attackers target backups first.
  • Test restores on a schedule, not just backups; an untested backup is a hope, not a plan.
  • Set retention and legal-hold policies that meet your legal and business needs without hoarding forever.
  • Define RTO and RPO per system with the business, and size your backup design to meet them.
  • Document who can authorise a restore and how, so recovery is not blocked by a missing approval at 2 a.m.

Logging and monitoring

  • Centralise logs from identity, endpoints, cloud, network, and email into one searchable place.
  • Decide what to log based on what you would need to answer during an incident, and resist logging everything.
  • Protect log integrity and set retention long enough to investigate a breach you discover months later.
  • Build a small set of high-value detections (new admin, impossible travel, mass download, mailbox rule creation) rather than hundreds of noisy ones.
  • Tune alerts continuously so the signal survives; an alert nobody trusts is an alert nobody reads.
  • Ensure alerts reach a human who can act, around the clock within your realistic means.
  • Keep a simple dashboard of the handful of things you most want to know are healthy.
  • Review detections after every incident and add the one that would have caught it sooner.

Vulnerability and change

  • Run authenticated vulnerability scans across your own estate on a schedule, with authorisation.
  • Prioritise using exposure and real-world exploitation signals (EPSS, CISA KEV), not CVSS severity alone.
  • Set patch SLAs by risk tier and track whether you are meeting them.
  • Record exceptions with a compensating control and an expiry date, never open-ended.
  • Protect changes with review and a rollback plan; most self-inflicted outages are change-related.
  • Report progress in trends the business understands, not raw vulnerability counts.

People and process

  • Write down who to call, in what order, when something goes wrong, and keep an offline copy.
  • Keep incident runbooks current and rehearse at least one each year (Appendix C gives you the templates).
  • Keep an accurate asset inventory; you cannot protect what you cannot see.
  • Agree with the business on acceptable use, data handling, and reporting expectations.
  • Keep a short list of your suppliers and the access they hold, and review it when a relationship ends.
  • Carry cyber insurance appropriate to your size, and read what it actually requires of you before you need to claim.
  • Review this checklist on a recurring schedule and date your last pass, so the ticks still mean something.

This list is a reference, not a verdict. Date the day you last worked through it, note what you deferred and why, and come back to it on a schedule. An estate drifts; people leave, projects spin up cloud resources nobody tags, a baseline quietly loosens. The value of a checklist like this is not the single pass that makes you feel covered but the recurring pass that keeps the coverage real.

KEY POINTS

Key points

  • Run the checklist as one pass over the estate and turn every unticked item into a ticket.
  • Secure the short core first: MFA, tested backups with an out-of-reach copy, patching of exposed systems, no standing admin rights, and investigable logs.
  • Work domain by domain, highest-leverage items first, instead of changing everything in one week.
  • Set up break-glass accounts with offline credentials and an alert on every use.
  • Date each pass, record what you deferred and why, and schedule the next review.

QUESTIONS

Common questions

What should a small IT team do first from a long security hardening checklist?
Start with a short core that prevents the most harm for the least effort: MFA for every user, reliable and tested backups with at least one copy out of an attacker's reach, prompt patching of exposed systems, removal of standing administrative privilege, and enough logging to investigate an incident. The rest of a hardening checklist extends that core but does not replace it.
How should an IT security checklist be used without breaking things?
Work through it one domain at a time rather than changing everything at once, and within each domain take the items that close the widest gap for the least effort first. Mark what is done, flag what is not, and turn each gap into a ticket. A team that attempts too many changes in a short period is likely to cause outages and lose the support of the business.
What is the 3-2-1-1-0 backup rule?
The 3-2-1-1-0 rule means keeping three copies of data on two different media, with one copy off-site and one copy immutable or offline, and confirming zero errors when restores are tested. The immutable or offline copy is the one ransomware cannot encrypt or delete, and the restore test is what shows the backups actually work.
How often should a security hardening checklist be reviewed?
It should be reviewed on a recurring schedule, with the date of the last pass recorded along with anything deferred and the reason. Environments drift as people leave, new cloud resources are created without tags, and baselines are loosened, so a checklist that was completed once and never revisited can suggest coverage that no longer exists.
Which domains does a complete IT hardening checklist cover?
The Hardened Stack's checklist covers identity and access, endpoints, network, cloud on Azure and AWS, Microsoft 365 and Entra ID, email, data and backup, logging and monitoring, vulnerability and change management, and people and process. It is deliberately longer than any single organisation needs, so teams can choose the items that fit their size and risk.