
System hardening: the default settings to lock down first
System hardening means closing everything a system leaves open by default that you do not use. Vendors ship devices and software to work straight away, not to be secure. This article sets out the order: what to lock down first, which list to base it on and how to stop it all drifting open again within a year.
Hardening does not fix bugs in software; updates do that. It removes what is switched on by default and not needed: passwords nobody changed, legacy protocols, admin rights and open ports.
You do not have to write the list yourself. CIS Benchmarks and Microsoft’s security baselines are worked-out starting points; begin at the basic level and record where you deviate.
Being hardened is a state, not a project. Without policy that enforces and measures the settings, an environment slides back towards the defaults within months.
What system hardening is, and what it is not
System hardening means shrinking a system’s attack surface: you switch off what is not needed, close what is open by default and enforce what is optional by default. It applies to laptops and servers, but just as much to switches, printers, the NAS in the comms cupboard and your Microsoft 365 tenant.
It is not the same as patching. An update fixes a flaw in the vendor’s code. Hardening is about configuration: a fully patched server with an admin account that has no password policy and an open RDP port is up to date and still exposed. You need both, and they are different kinds of work: patching is a rhythm, hardening is a decision you make properly once and then guard.
Defaults fall short for a simple reason. A vendor does not know how you will use the device, so it ships it to work without complaint in as many situations as possible. That means legacy protocols on for compatibility, generous rights so installations do not fail, and a management interface reachable from any network.
The default settings to lock down first
Not every setting carries the same weight. In a small or mid-sized organisation these six deliver the most for the least effort, and an afternoon’s inventory will find them.
- Default passwords on devices. Switches, access points, printers, cameras, the NAS and a server’s management card. They often still match the manual, and the manual is online.
- Local admin rights for users. Anyone who is admin on their own laptop also installs whatever a phishing email brings with it. Remove them and provide a route for the software people genuinely need.
- The same local admin password on every laptop. One leaked password then opens every device. Windows LAPS, now built into Windows, gives each device its own password that rotates.
- Legacy protocols. SMBv1, NTLMv1, LLMNR and NetBIOS name resolution, and TLS 1.0 and 1.1. They are usually still on because nobody switched them off, not because anything needs them.
- Management from the internet. RDP open to the outside, a firewall admin page reachable externally, a NAS behind a port forward. Administration belongs behind a VPN or a management platform.
- Macros and executables from email and downloads. Microsoft now blocks macros in files from the internet by default, but policy can reverse that and older Office versions do not do it. Check what actually applies in your environment.
Endpoint, server, network and cloud: one list each
Hardening becomes manageable when you take it layer by layer. Each layer has a handful of settings that matter most.
- Endpoint: disk encryption on, the operating system firewall on and not switchable by the user, a screen lock after a short idle period, and the attack surface reduction rules in Microsoft Defender that block common attack techniques.
- Server: only the roles and services the server actually performs, no browsing or email on the server itself, and administration through separate admin accounts that are never used for day-to-day work.
- Network: management interfaces reachable only from a separate management network, Telnet and unencrypted HTTP management off, SNMP at version 3 or not at all, and unused switch ports disabled.
- Microsoft 365 and cloud: legacy authentication off, the audit log on, and a deliberate setting for external sharing rather than the default.
Start with the layer that has the most devices and where users bring in the most risk. In almost every organisation that is the endpoint.
Pick a baseline instead of inventing one
You do not need to work out which of Windows’ thousands of settings matter. Worked-out lists exist for that, and they are better reasoned than anything an organisation invents on its own.
- CIS Benchmarks. Free configuration guides from the Center for Internet Security, per operating system, application and network platform. They have two levels: Level 1 is the baseline that rarely breaks anything, Level 2 is stricter and meant for environments where that is worth the effort.
- Microsoft Security Baselines. Microsoft’s own recommended settings for Windows, Edge and Microsoft 365, also available as ready-made security baselines in Intune.
- What regulators and customers ask. In the UK, secure configuration is one of the five technical controls in Cyber Essentials, and the NCSC publishes guidance on it. Under NIS2, basic cyber hygiene is one of the listed measures. A baseline is the most concrete evidence that you have it in place.
The rule of thumb: choose one baseline per platform, start at the basic level and write down where you deviate. Mixing two lists produces settings that contradict each other.
The order: inventory, test, roll out
Hardening rarely goes wrong on the setting itself and often goes wrong on the order. Switching off a legacy protocol on every device at once is the quickest way to find out which scanner, which finance package or which workshop machine depended on it.
- Inventory first. Which devices exist, what runs on them and what talks to what. Without that list you do not know what you are touching.
- Measure before you enforce. Many settings have an audit mode: the system logs what it would block without blocking it. A few weeks of logs tell you more than a meeting.
- Roll out in rings. IT first, then a representative pilot group, then everyone else. The rings you use for updates work here too.
- Record what you did. Which baseline, which version, which deviations and why. At the next audit, cyber insurance renewal or change of supplier, that is the document everyone asks for.
Exceptions are the real work
No environment meets a baseline without deviations. There is always an old application that needs SMBv1, a production machine that must not restart or a supplier that needs remote access. That is fine. What is not fine is an exception nobody can explain.
Give every exception four things: a reason, an owner, an end date and a compensating control. A machine that needs a legacy protocol goes into its own network segment rather than the protocol being switched back on everywhere. An exception without an end date has become the standard within a year.
The question to ask of every exception: does this apply to this one system, or are we opening it for everyone because that is easier? The second is how most gaps appear.
Staying hardened: preventing drift
An environment that meets the baseline today will not do so in six months on its own. Someone opens something temporarily to fix a problem, a new laptop arrives with factory settings, a major Windows update adds new settings. That is configuration drift.
You prevent drift by enforcing settings rather than setting them once. Through Intune or Group Policy the baseline is applied again and again, and a new device receives it at first sign-in. That only works well if endpoints are standardised; how to approach that is covered in standardising the Windows workplace.
Measure as well: a report showing per device whether it meets the baseline, and a fixed check after every major update and every new version of the baseline. For network equipment, which Intune does not manage, a periodic comparison of the configuration against a stored reference is the minimum.
Where our work stops
What we do. On the systems we manage, we choose a baseline with you, roll it out in phases, keep the register of exceptions and enforce the settings, with a report showing where each device stands. That covers endpoints, servers, network equipment and Microsoft 365.
What we do not do. We are not an auditor and we certify no one: a report showing that your environment meets a CIS Benchmark is a measurement, not a certificate, and not a judgement on whether you comply with NIS2, Cyber Essentials or any other standard. That judgement comes from a qualified assessor. Nor do we harden third-party applications from the inside: the configuration of your ERP, point-of-sale software or sector package belongs to its vendor, and we agree with that vendor what is possible on the system underneath.
Questions we get about this
The five that come up most often.
What is system hardening?
System hardening is reducing a system’s attack surface by switching off what is not needed and closing what is open by default. Think of default passwords, local admin rights, legacy protocols and administration reachable from the internet. It applies to endpoints, servers, network equipment and cloud environments.
What is the difference between hardening and patching?
A patch fixes a flaw in the vendor’s software. Hardening is about configuration: which features are on, who has which rights and what is reachable. A fully patched system with poor settings is still exposed, so you need both.
What are CIS Benchmarks?
CIS Benchmarks are free configuration guides from the Center for Internet Security for operating systems, applications and network platforms. They have a basic level (Level 1) that rarely breaks anything and a stricter level (Level 2). They are a sound starting point if you do not want to decide yourself which settings matter.
Can hardening break applications?
Yes, which is why you roll it out in phases. Many settings can run in audit mode first, so you see what would be blocked. A pilot group follows, then everyone else. Anything that genuinely needs an old setting gets a recorded exception with an end date and, where needed, its own network segment.
Can you certify that our systems are hardened?
No. We are not an auditor and we certify no one. We can measure and report whether systems meet a chosen baseline and which exceptions apply. A formal judgement against a standard or legal requirement comes from a qualified assessor.
Further reading
What system hardening connects to.
Want to know what is still open by default?
We check your endpoints, servers and network equipment against a baseline and show which settings to close first, and what might break when you do.
Practical IT knowledge in your inbox
New guides on management, security and the workplace, written by the people doing the work. No sales talk, and you can unsubscribe in one click.