You cannot protect what you cannot see
An asset is anything that can be attacked or that holds something worth protecting: physical devices, virtual machines, cloud resources, SaaS applications, network equipment, identities and data. Most organisations can list their laptops and servers but underestimate everything else.
Unknown assets are dangerous because they receive no care. Nobody patches, monitors or notices the server nobody remembers. Many real compromises begin at a forgotten test environment, an unsupported legacy system or a demo resource that was never removed.
The aim is not a perfect manual register, which decays within weeks because updating it depends on someone remembering. The aim is a living inventory derived automatically from the systems that already know about assets, trusted enough to use, and owned by a named person. An outdated spreadsheet is worse than nothing because people believe it.
Discovery from several directions
Start with authoritative sources before scanning anything. Entra ID and Active Directory know about identities and domain-joined devices, Intune knows about managed endpoints, and the Azure and AWS control planes know about every resource in their accounts. DHCP and DNS see devices appearing on the network. The chapter shows read-only queries for Active Directory computers with last logon dates, an Azure Resource Graph query across subscriptions, and an EC2 listing that must be repeated for each AWS region.
Active discovery, such as a scan of your own address ranges that you are authorised to scan, finds what those sources miss. Hosts that answer a scan but are absent from the directory or device management are the ones to investigate.
Two categories escape both approaches. SaaS applications live outside your network; the identity provider's sign-in records and expense reports reveal them, including services bought without IT. Data also needs mapping: knowing which systems hold personal, financial or credential data tells you which catalogued assets matter most. In the book's example, a cloud query found seven unexplained Azure resources, including a test VM that had been running and billing for nineteen months.
Configuration baselines as code
A configuration baseline is an agreed set of settings for a type of system. Rather than inventing one, adopt CIS Benchmarks, starting with Level 1 settings that harden without much breakage and considering Level 2 for sensitive systems, or Microsoft's security baselines through Intune and group policy.
The baseline should live as policy or code, not as a document someone follows by hand. Group policy or Intune profiles enforce it on Windows; Ansible or similar tools do it on Linux. The chapter's example SSH hardening task is idempotent and validates the configuration before reloading, with a clear warning to confirm key-based admin access before disabling password logins. Every baseline also needs a way to verify it, such as a read-only check that SMBv1 is disabled.
Roll baselines out in stages, from a test machine to a pilot group to the whole fleet, with a documented way to roll back. A baseline that breaks production teaches people to distrust hardening.
Catching configuration drift
Settings do not stay applied. Late-night troubleshooting, vendor installers and old images all pull systems away from the baseline. Drift is caught by regularly comparing actual state against the desired state: configuration management tools in check mode, Azure Policy, or AWS Config. A simple, high-value check is a read-only scan for storage buckets that do not block public access.
Treat drift as a signal to triage. Settings that should never change, such as public access on a store of customer exports, can be corrected automatically. Settings where an engineer may have had a good reason should raise an alert for a human decision. For advanced readers, the chapter describes golden images and immutable infrastructure, where servers are rebuilt from a versioned definition rather than modified in place, and suggests moving towards it gradually.
A modest CMDB and cloud tagging
A configuration management database only needs to answer a few questions for each asset: what it is, who owns it, how critical it is, where it lives and when it was last seen. Ownership should be a named person or team, never IT in general, because an unowned asset turns a quick incident decision into a long search.
In the cloud, enforce tags for owner, environment, cost centre and data classification through infrastructure as code, and run a read-only query for resources missing an owner tag. Reconcile discovery results against the CMDB on a schedule, and make resolving differences someone's job.
KEY POINTS
Key points
- Export devices and servers from Entra ID, Active Directory and Intune, and query every cloud subscription and region.
- Give every unexplained asset an owner and a decision: keep and manage, or decommission by a set date.
- Choose a baseline for each major platform, such as CIS Level 1, and apply at least one as code or policy.
- Turn on Azure Policy or AWS Config, starting with a check for publicly accessible storage.
- Enforce owner, environment and data-class tags on new cloud resources and audit existing ones.
- Schedule a monthly inventory reconciliation with a named owner.
QUESTIONS
Common questions
- How should a small IT team build an asset inventory?
- Start by exporting data from authoritative sources such as Entra ID, Active Directory, Intune and the Azure and AWS control planes, then run authorised scans of your own network ranges to find what those sources miss. Reconcile the two, add SaaS applications found through sign-in and expense records, and assign every unexplained asset an owner and a keep-or-decommission decision.
- Which configuration baseline should I use for hardening?
- The Hardened Stack recommends adopting an existing baseline rather than writing your own, with CIS Benchmarks Level 1 as a sound default and Microsoft's security baselines for Intune and group policy on the Microsoft estate. Apply the baseline as policy or code, roll it out in stages with a rollback plan, and pair it with a way to verify each setting.
- What is configuration drift and how do you detect it?
- Configuration drift is the gradual gap between how systems should be configured and how they actually are, caused by troubleshooting changes, installers and outdated images. Detect it by regularly comparing actual state with the baseline using configuration management in check mode, Azure Policy or AWS Config. Auto-correct only settings that should never change, and alert on the rest.
- What should a simple CMDB record for each asset?
- A small team's CMDB needs only a few fields per asset: what it is, who owns it, how critical it is, where it lives and when it was last seen. Ownership should be a named person or team, and the database must be reconciled against discovery results on a schedule to stay trustworthy.
- Why is cloud resource tagging important for security?
- Tags such as owner, environment, cost centre and data classification turn the cloud control plane into a live inventory you can query by meaning. Enforcing them through infrastructure as code means new resources are created with an owner, and a query for resources without an owner tag quickly reveals orphaned resources nobody is responsible for.