What is Defense in Depth?
Defense in depth in cyber security stacks independent controls so that one failed or bypassed control does not hand an attacker the whole environment. NIST describes variable barriers across multiple layers, built from people, technology and operations. Each layer that stops an attack also writes a log entry, and the SOC investigates from those entries.
Definition
- Defense in Depth
- Defense in depth in cyber security stacks independent controls so that one failed or bypassed control does not hand an attacker the whole environment. NIST describes variable barriers across multiple layers, built from people, technology and operations. Each layer that stops an attack also writes a log entry, and the SOC investigates from those entries.
How Defense in Depth Works
NIST's glossary carries two definitions worth knowing. The one from SP 800-53 Rev. 5, the US federal control catalog, is about strategy: integrate people, technology and operations capabilities to establish variable barriers across multiple layers and missions of the organization. The one NIST quotes from ISA/IEC 62443, the industrial control systems standard, is about engineering: layer heterogeneous security technologies in the common attack vectors so that attacks missed by one technology are caught by another. By that second definition, two controls that fail for the same reason are one layer, however many product names they carry.
Microsoft Defender shows how that happens. Attack surface reduction (ASR) rules and the antivirus scanner look like two endpoint layers, but Microsoft's ASR documentation lists real-time protection in Microsoft Defender Antivirus as a requirement for the rules. An attacker with local admin rights who turns real-time protection off removes both at once. Tamper protection exists to refuse that change, and when it does, Defender writes event 5013 instead.
No standard fixes the number of layers
Microsoft's SC-900 training module lists seven: physical security, identity and access, perimeter, network, compute, application and data. Other models count five by merging some of them. Layered security is the same idea under another name.
For a SOC the useful way to count is by log source: if two layers write to the same log and fail together, count them once. Zero trust sits inside this picture as one way to build the identity and network layers, verifying each request instead of trusting the network it comes from.
One intrusion, layer by layer
What follows is a simulated example generated by SOCSimulator Research: hosts, accounts and times are invented, and the external addresses come from documentation ranges. It follows one attacker through a mid-sized company on 6 October 2026 (UTC) and gives, for each layer, the event that layer writes. Event IDs, log channels and meanings follow Microsoft Learn.
Office hardening, held
09:12:04, workstation WS-ACC-014: user m.reyes opens Invoice_4471.docm and the macro tries to start powershell.exe. The ASR rule Block all Office applications from creating child processes (GUID d4f940ab-401b-4efc-aadc-ad5f3c50688a) is in block mode, so it stops the launch and writes 1121 to Microsoft-Windows-Windows Defender/Operational. In audit mode the same rule would have logged 1122 and let PowerShell run.
Antivirus, held
09:14:37, same host: she opens the second attachment, an ISO with a loader DLL inside. Defender writes 1116 (malware detected, with threat name, path and detection origin) and then 1117 (action taken, here quarantine). Read them as a pair. A 1116 followed by 1118 or 1119 means the action failed and the file may still be on disk.
The user, bypassed
The email also carried a link to a fake sign-in page, and she typed her password into it. Nothing on the company network logs that step. The control here is training, and its failure leaves no event until the stolen password gets used.
Identity, missing
11:02 to 11:04, file server FS-OPS-02, whose RDP port was forwarded for a vendor and never closed. Two 4625 failures from 203.0.113.45 for the accounts admin and scanner, then a 4624 with Logon Type 10 (RemoteInteractive) for m.reyes from the same address. RDP on that server had no MFA, so the identity layer was one password.
Host firewall, held
11:12: from FS-OPS-02 the attacker tries SMB against workstation WS-ACC-022. Its firewall accepts port 445 only from the management subnet, so Windows Filtering Platform drops the connection and WS-ACC-022 writes 5157 (connection blocked, Direction Inbound, local port 445) to its Security log. That needs Audit Filtering Platform Connection enabled for failures; without it the block still happens and nobody sees it.
Endpoint protection, switched off
11:21: m.reyes is still in the server's local Administrators group from an old project, and tamper protection is off on servers. Set-MpPreference -DisableRealtimeMonitoring $true produces 5007 (configuration changed, with old and new values) and 5001 (real-time protection disabled). ATT&CK files this under T1685 Disable or Modify Tools, tactic Defense Impairment; the old ID T1562.001 now redirects there. Since ASR needs real-time protection, both endpoint layers on this server are gone.
Logging, tampered with
11:22: auditpol /set /subcategory:"Process Creation" /success:disable writes 4719 (system audit policy changed), which Windows logs regardless of the Audit Policy Change setting. 11:23: wevtutil cl Security empties the log and writes 1102. ATT&CK has these as T1685.001 Disable or Modify Windows Event Log and T1685.005 Clear Windows Event Logs (formerly T1562.002 and T1070.001). Neither command touches the Defender channel, so 5001 and 5007 are still on the server.
SIEM, held
The forwarder shipped the 4624, the 4688 process events for powershell.exe and auditpol.exe, and the 4719 before 11:23. Those 4688s carry full command lines only because the Include command line in process creation events policy was on; by default that field is empty. On the server the Security log now starts with 1102. In the SIEM, the hour before it is intact.
Reading by absence
From 11:30 the server writes a 5156 every few minutes:
Application \device\harddiskvolume2\programdata\sync\rc.exe, Direction Outbound, destination 198.51.100.77 port 443There is no 4688 for rc.exe anywhere. A connection from a process that never started means process auditing is off on that host, or the EDR sensor that should have recorded the launch is not there. In this case the 4719 in the SIEM says which one, and since when.
Four layers held (three preventive, plus the SIEM), one was bypassed (the user), two were never there (MFA on the exposed RDP path, least privilege on the server) and the attacker switched two off (endpoint protection and process auditing on the server). Detection came from the SIEM copy of 4719 and 1102, the one layer the attacker could not reach from the server.
Defense in Depth in SOC Operations
On shift, defense in depth becomes a question you ask of every confirmed alert: which layers should have seen this, and which ones did? Take a SIEM hit on a 5156 from an unknown binary. Endpoint: is there a 4688 or an EDR process event for it? Identity: which logon session owns the process, and did that 4624 come with MFA? Network: did the perimeter firewall log the same flow, allowed or denied? A layer that should have written something and stayed silent is a finding on its own, because either the attacker is past it or it was never deployed on that host. Put the gap in the ticket so it reaches whoever owns that control, or the next intrusion walks through the same hole. Two searches deserve a permanent place in your saved queries: Security 1102 and 4719 outside a change window, and Defender 5001 or 5007 on any host with no change filed against it. Those are the events an attacker leaves when switching a layer off, and the SIEM copy is often the only one left.
Practice Defense in Depth in a Real SOC
SOCSimulator provides hands-on training with realistic SIEM, XDR, and Firewall interfaces. Build real analyst skills investigating defense in depth scenarios with zero consequences, free.
Frequently Asked Questions
- What are the 7 layers of defense in depth?
- The seven-layer list people usually quote is Microsoft's, from its SC-900 security fundamentals training: physical security, identity and access, perimeter, network, compute, application and data, with data as the innermost layer. It is a teaching model, not a standard. NIST's definition talks about multiple layers and does not fix a number.
- What is an example of defense in depth?
- A phishing email whose macro is stopped by an ASR rule (event 1121), whose second payload is quarantined by antivirus (1116, then 1117), and whose follow-up SMB attempt is dropped by a workstation firewall (5157). Any one of those controls could have failed. The attacker had to beat all three, and each block left an event the SOC can query.
- What are the 5 layers of the defense in depth model?
- No standard fixes the number of layers: NIST's definition gives none, and Microsoft's training uses seven. A five-layer list covers the same ground at a coarser grain. For SOC work a practical five is identity, network, endpoint, application and data, because each one writes to a different log source you can search.
- What is the meaning of defense in depth?
- Not relying on any single security control. NIST SP 800-53 Rev. 5 defines it as a strategy that integrates people, technology and operations to establish variable barriers across multiple layers of the organization. In practice: assume each control will fail at some point, and put a different kind of control behind it.
Sources
- NIST CSRC Glossary: defense-in-depth (SP 800-53 Rev. 5; ISA/IEC 62443 via NISTIR 8183)
- Microsoft Learn: Describe defense in depth (SC-900 training module)
- Microsoft Learn: Microsoft Defender Antivirus event IDs and error codes (1116, 1117, 1118, 1119, 1121, 5001, 5007, 5013)
- Microsoft Learn: ASR rules overview (requirements and audit-mode events)
- Microsoft Learn: ASR rules reference
- Microsoft Learn: Tamper protection overview
- Microsoft Learn: 4688(S) A new process has been created
- Microsoft Learn: 5156(S) The Windows Filtering Platform has permitted a connection
- Microsoft Learn: 5157(F) The Windows Filtering Platform has blocked a connection
- Microsoft Learn: 4624(S) An account was successfully logged on
- Microsoft Learn: 4625(F) An account failed to log on
- Microsoft Learn: 4719(S) System audit policy was changed
- Microsoft Learn: 1102(S) The audit log was cleared
- MITRE ATT&CK: T1685 Disable or Modify Tools
- MITRE ATT&CK: T1685.001 Disable or Modify Windows Event Log
- MITRE ATT&CK: T1685.005 Clear Windows Event Logs
Related Terms
Zero Trust is a security architecture philosophy based on "never trust, always verify," requiring co...
The principle of least privilege states that users, processes, and systems should hold only the mini...
EDR (endpoint detection and response) is security software that puts an agent on each laptop, server...
A firewall is a network security control that inspects traffic crossing a boundary and permits or de...
A Next-Generation Firewall (NGFW) combines traditional stateful packet inspection with deep packet i...
MFA (multi-factor authentication) is a sign-in check that asks for two or more different kinds of pr...
Security Information and Event Management (SIEM) is a platform that aggregates, normalizes, and corr...
Log management is the process of generating, collecting, storing, protecting and finally deleting th...
In cyber security, the attack surface is every point where an attacker could get into your environme...
Lateral movement is the attack phase where adversaries expand access from an initial foothold to add...
More Concepts Terms
Related SOC Training Resources
SOC Analyst (Tier 1) Career Guide: Salary & Skills
Tier 1 SOC Analysts are the front line. You monitor alert queues, triage incoming detections, classify them as true or f…
Read more Career PathDetection Engineer Career Guide: Salary & Skills
Detection Engineers build the rules, analytics, and automated workflows that determine what the SOC can see. You transla…
Read more Career PathSecurity Engineer Career Guide: Salary & Skills
Security Engineers build and maintain the infrastructure that SOC analysts depend on. You deploy SIEMs, configure firewa…
Read more ComparisonSOCSimulator vs LetsDefend: Comparison
SOCSimulator wins on operational realism. You get multi-tool shift simulation with SLA pressure, noise injection, and al…
Read more ComparisonSOCSimulator vs CyberDefenders: Comparison
SOCSimulator trains the operational workflow: alert triage, correlation, and response under pressure. CyberDefenders tra…
Read more ToolSIEM Training Console: SOCSimulator
The SIEM console in SOCSimulator replicates the workflow of enterprise platforms like Splunk Enterprise Security, Micros…
Read more ToolXDR Training Console: SOCSimulator
The XDR console in SOCSimulator replicates the investigation workflow of platforms like CrowdStrike Falcon, Microsoft De…
Read more ToolFirewall Training Console: SOCSimulator
The Firewall console in SOCSimulator replicates the log analysis experience of enterprise platforms like Palo Alto Netwo…
Read more TechniqueMITRE ATT&CK® Techniques: Detection Training Library
Browse all MITRE ATT&CK® techniques with detection strategies and example alerts.
Read more Career PathCybersecurity Career Paths: 2026 Guide
Explore SOC analyst career paths with salary data, required skills, and certification roadmaps.
Read more PlaybookSOC Investigation Playbooks: Step-by-Step Guides
Practitioner investigation playbooks with decision trees and real SIEM queries.
Read more FeatureShift Mode: Real-Time SOC Simulation
Practice alert triage under realistic time pressure with SLA timers and noise injection.
Read more