Skip to content

Azure Network Security Review

A secure Azure network is not simply one with NSGs attached. It is one where traffic flows, trust boundaries, public exposure, private connectivity, routing, and monitoring are intentionally designed and regularly reviewed.

Welcome to the Azure Network Security Review Runbook.

This runbook builds on the practical skills from:

  • Azure Network Security Lab
  • Azure Security Assessment Runbook
  • Microsoft Entra ID Security Review Runbook
  • Microsoft Defender for Cloud Lab
  • Microsoft Sentinel Lab

The objective is to move from:

I know how Azure networking works.

to:

I can systematically review Azure network architecture, identify risky traffic paths and public exposure, produce defensible findings, and recommend practical remediation.

This runbook is directly relevant to roles such as:

  • Azure Security Engineer
  • Cloud Security Engineer
  • Network Security Engineer
  • Cloud Security Consultant
  • Azure Solutions Architect
  • Cloud Security Architect
Item Details
Runbook Azure Network Security Review
Difficulty Intermediate β†’ Advanced
Estimated Time 3–6 hours depending on environment
Primary Focus Network Architecture, Exposure & Segmentation
Target Roles Azure Security Engineer, Cloud Security Engineer, Network Security Engineer
Certification Alignment AZ-104, AZ-500, AZ-305, SC-100
Output Network Findings, Risk Ratings, Exposure Matrix, Remediation Roadmap

You have been asked to perform a security review of an Azure network environment.

The organization operates:

  • multiple Azure VNets
  • frontend workloads
  • application services
  • databases
  • shared services
  • public IP addresses
  • Private Endpoints
  • Azure Firewall
  • VNet peerings
  • hybrid connectivity

Security leadership is concerned that:

  • some systems may be unnecessarily exposed to the internet
  • NSGs may contain overly broad rules
  • application tiers may not be properly segmented
  • unrestricted outbound access may exist
  • route tables may bypass central controls
  • VNet peerings may create excessive trust
  • sensitive PaaS resources may use public network paths
  • network telemetry may be incomplete

Your job is to answer:

Which systems can communicate, through which path, under which controls, and whether that connectivity is actually required?

By the end of this runbook, you should be able to:

  • understand Azure network architecture
  • inventory VNets and subnets
  • identify public exposure
  • review Network Security Groups
  • evaluate rule priorities
  • review outbound access
  • assess route tables and UDRs
  • review Azure Firewall
  • assess hub-and-spoke architecture
  • review VNet peering
  • evaluate Private Endpoints
  • assess private DNS
  • review hybrid connectivity
  • assess administrative access
  • review network telemetry
  • identify segmentation weaknesses
  • create network-security findings
  • prioritize remediation
  • validate network changes

Before beginning, confirm:

  • subscriptions in scope
  • resource groups in scope
  • VNets in scope
  • Regions in scope
  • production vs non-production
  • hybrid connectivity in scope
  • Azure Firewall in scope
  • Private Endpoints in scope
  • whether changes are permitted
  • whether testing is read-only

Do not alter:

  • NSGs
  • routes
  • peerings
  • firewall rules
  • production connectivity

unless explicitly authorized.

Use:

Scope
↓
Understand Architecture
↓
VNet Inventory
↓
Subnet Review
↓
Public Exposure
↓
NSG Review
↓
Routing
↓
Central Firewall
↓
Inter-VNet Connectivity
↓
Private Connectivity
↓
Hybrid Connectivity
↓
Egress
↓
Telemetry
↓
Findings
↓
Remediation
↓
Validation

Before reviewing individual firewall rules, understand the overall architecture.

Identify:

  • how many VNets exist

  • which subscriptions contain them

  • which workloads they support

  • which VNets are production

  • which VNets provide shared services

  • which VNets connect to on-premises infrastructure

Create:

VNet Subscription Region CIDR Environment Purpose

Ask:

Why does this VNet exist?

If there is no clear answer, ownership and governance may already be weak.

For each VNet document:

  • IPv4 address space

  • additional CIDRs

  • overlapping ranges

  • address-space utilization

  • overlapping address ranges

  • unnecessarily large address spaces

  • inconsistent network planning

  • connectivity limitations caused by overlap

Inventory subnets.

Create:

Subnet VNet CIDR Purpose Public Workloads?

Classify subnets such as:

  • frontend

  • application

  • database

  • management

  • shared services

  • firewall

  • Private Endpoint

Ask:

Does subnet design reflect workload trust boundaries?

A three-tier application may conceptually use:

Frontend
↓
Application
↓
Data

rather than placing everything in one subnet.

Document the intended communication.

Example:

Internet
↓
Frontend
↓
Application
↓
Database

Then identify prohibited paths.

Internet ─Xβ†’ Application
Internet ─Xβ†’ Database
Frontend ─Xβ†’ Database Admin Interface

A flat architecture where application tiers can communicate broadly may increase lateral-movement risk.

Identify all public IP resources.

Create:

Public IP Associated Resource Purpose Required? Risk

Review:

  • VMs

  • load balancers

  • Application Gateways

  • firewalls

  • other public resources

Ask:

Does this resource genuinely require direct public network reachability?

Create an exposure matrix.

Resource Public? Service/Port Business Need Risk

Pay special attention to:

  • SSH

  • RDP

  • database ports

  • application admin interfaces

  • custom management ports

TCP 22
Source: Internet
TCP 3389
Source: Internet

These should receive immediate review.

For important NSGs document:

  • associated subnet/NIC

  • inbound rules

  • outbound rules

  • priorities

  • source

  • destination

  • protocol

  • port

Create:

NSG Direction Source Destination Port Action Purpose

Remember:

Lower numerical priority is evaluated first.

When rules appear contradictory, inspect:

  • rule priority

  • source

  • destination

  • protocol

Do not evaluate a rule independently.

Evaluate the effective traffic decision.

Look for:

  • Any

  • Internet

  • very large CIDR ranges

  • broad VNet access

Then ask:

Is this source really required?

Public HTTPS on an internet-facing application may be expected.

Public database access usually requires much stronger justification.

Specifically assess:

  • TCP 22

  • TCP 3389

  • database administration

  • management consoles

Finding ID:
AZNET-001
Title:
Internet-Exposed Administrative Access
Severity:
High
Observation:
A production virtual machine allows RDP from an unrestricted internet source.
Risk:
The workload is directly exposed to internet-based authentication and exploitation attempts.
Recommendation:
Remove unrestricted public administrative access and use an approved private or controlled management path.

Do not focus only on ingress.

Inspect whether workloads can send:

Any
β†’ Internet
β†’ Any Port

Ask:

Does the workload require unrestricted outbound internet connectivity?

Which destinations are actually needed?

Compromised workloads may use outbound access for:

  • command and control

  • malicious downloads

  • data exfiltration

Source Workload Destination Port Purpose Required?

Where applicable, inspect effective security rules on network interfaces.

This helps identify combined effects of:

  • subnet NSG

  • NIC NSG

Do not assume a visible rule on one NSG equals the final traffic decision.

Inventory User Defined Routes.

Create:

Route Table Subnet Destination Next Hop Purpose

Look for:

  • default routes

  • virtual appliances

  • firewall routes

  • gateway routes

Understand whether sensitive traffic is forced through:

  • Azure Firewall

  • network virtual appliance

  • central inspection

Example:

Application Subnet
↓
UDR
↓
Azure Firewall
↓
Destination

A route change can unintentionally bypass:

  • inspection

  • centralized egress

  • expected trust boundaries

For suspicious or unclear network paths, determine the actual next hop.

Ask:

Is the traffic going where the architecture says it should?

If Azure Firewall exists, review:

  • firewall policy

  • network rules

  • application rules

  • DNAT rules

  • logging

  • egress controls

  • threat intelligence settings where applicable

Ask:

Which traffic flows must pass through the firewall?

Then verify that routing enforces that architecture.

Look for:

  • broad source ranges

  • broad destination ranges

  • Any ports

  • rules with unclear business purpose

  • stale temporary rules

Determine whether internet-bound traffic from private workloads is centralized.

Concept:

Spoke Workloads
↓
UDR
↓
Azure Firewall
↓
Internet

Centralized egress can improve:

  • visibility

  • policy enforcement

  • threat detection

  • governance

But it can also introduce:

  • cost

  • complexity

  • availability dependencies

Where applicable, map:

Hub VNet
β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”Όβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
↓ ↓ ↓
Prod Dev Shared
Spoke Spoke Services

The hub may contain:

  • Azure Firewall

  • DNS

  • VPN/ExpressRoute

  • shared services

Assess:

  • route propagation

  • firewall enforcement

  • shared-service access

  • administrative access

  • segmentation between spokes

Ask:

Can every spoke communicate with every other spoke?

They should not necessarily be able to.

Inventory peerings.

Create:

Source VNet Destination VNet Subscription Purpose Required?

Ask:

What new trust path does this peering create?

Assess relevant settings such as:

  • virtual-network access

  • forwarded traffic

  • gateway transit

  • remote gateways

depending on architecture.

A development VNet should not automatically have unrestricted access to production simply because both connect to a shared topology.

Create:

Source Destination Allowed? Required? Control
Dev Prod DB No No NSG/Firewall
Frontend App Yes Yes NSG
App DB Yes Yes NSG

This is valuable professional documentation.

Inventory Private Endpoints.

Create:

Endpoint Service VNet/Subnet Resource Public Access Disabled?

Private Endpoints can reduce public network exposure by providing private IP-based connectivity.

But they do not replace:

  • authentication

  • authorization

  • RBAC

  • data access controls

Look for services such as:

  • Storage

  • Key Vault

  • SQL/database services

  • other sensitive PaaS resources

Ask:

Does this service need a public network path?

Finding ID:
AZNET-002
Title:
Sensitive PaaS Resource Uses Unnecessary Public Network Access
Severity:
Medium / High
Observation:
An internal application accesses a sensitive Azure service through a public endpoint although private connectivity is available.
Risk:
The design expands network exposure and increases attack surface.
Recommendation:
Evaluate Private Endpoint-based connectivity and restrict unnecessary public access after validating application dependencies.

For Private Endpoints, verify that name resolution maps services to private addresses as intended.

A Private Endpoint can exist while an application still resolves the public endpoint if DNS is misconfigured.

Document:

  • private DNS zones

  • VNet links

  • ownership

  • forwarders/resolvers where applicable

Private Endpoint Exists
+
DNS Resolves Public Address
=
Traffic May Still Use Public Path

Identify:

  • Site-to-Site VPN

  • Point-to-Site VPN where applicable

  • ExpressRoute

  • gateway infrastructure

Create:

Connection Source Destination Purpose Criticality

Assess:

  • business purpose

  • routing

  • tunnel redundancy

  • monitored status

  • permitted networks

Ask:

Which on-premises networks can reach Azure production workloads?

Understand:

  • connectivity purpose

  • routing

  • redundancy

  • security requirements

Dedicated connectivity does not automatically replace all other security requirements.

Still consider:

  • segmentation

  • authentication

  • encryption requirements

  • monitoring

Map:

On-Premises
↓
VPN / ExpressRoute
↓
Hub
↓
Azure Workloads

Then document which network segments should communicate.

Determine how administrators access workloads.

Possible methods:

  • public SSH/RDP

  • Azure Bastion

  • VPN/private management subnet

  • just-in-time controls

  • managed administration services

Administrative access should be:

  • authenticated

  • least privileged

  • limited

  • monitored

  • preferably not broadly public

If Azure Bastion is used, assess:

  • target resources

  • deployment architecture

  • public exposure reduction

  • ownership

  • logging where appropriate

Do not assume Bastion automatically solves all administrative security issues.

Identity and VM access controls still matter.

For internet-facing load balancers, assess:

  • frontend IP

  • listeners

  • backend pools

  • NSGs

  • health probes

  • intended workload

Where Application Gateway exists, review:

  • frontend configuration

  • listeners

  • backend pools

  • TLS

  • Web Application Firewall if used

  • logging

For public web applications, architecture may include:

Internet
↓
Application Gateway / WAF
↓
Application Tier
↓
Private Data Tier

For important databases assess:

  • public endpoint

  • Private Endpoint

  • firewall/network restrictions

  • application source networks

  • administrative access

Internet
↓
Database Port

should receive immediate attention.

Assess important storage accounts for:

  • public network access

  • network rules

  • Private Endpoints

  • trusted-service exceptions

Assess:

  • public network access

  • firewall settings

  • Private Endpoint

  • application dependencies

40 β€” Review Kubernetes / AKS Networking Where Applicable

Section titled β€œ40 β€” Review Kubernetes / AKS Networking Where Applicable”

If AKS exists, assess high-level areas such as:

  • API-server exposure

  • network architecture

  • ingress

  • private-cluster design

  • workload segmentation

Do not treat AKS as only another VM network.

It introduces additional layers of networking.

Determine whether sufficient network evidence exists.

Possible sources may include:

  • Azure Firewall logs

  • NSG-related telemetry where supported

  • Application Gateway logs

  • load-balancer diagnostics

  • Azure Monitor

  • Microsoft Sentinel

Ask:

If suspicious network traffic occurred yesterday, could the security team reconstruct the path today?

For important network/security resources, assess whether diagnostics are sent to appropriate destinations.

Review:

  • Log Analytics

  • storage

  • Event Hub where applicable

Assess whether Microsoft Sentinel receives relevant network telemetry.

Create:

Network Source Sentinel Connected? Detection Use Case
Azure Firewall
Identity
Azure Activity
Application Gateway

Correlate available:

  • Defender for Cloud findings

  • Sentinel alerts

  • firewall telemetry

  • Azure Activity

Suspicious Sign-In
↓
NSG Rule Changed
↓
Public Port Opened
↓
Suspicious Traffic

This is a network + identity incident, not only a firewall event.

Look for changes involving:

  • NSGs

  • route tables

  • public IPs

  • VNet peerings

  • firewalls

  • network interfaces

Ask:

Who changed the network control, and what happened afterward?

Determine whether sensitive network changes require:

  • change approval

  • infrastructure as code

  • peer review

  • monitoring

  • rollback

Direct manual changes to critical production networking with no audit/change process increase operational and security risk.

If networking is deployed through:

  • Terraform

  • Bicep

  • ARM templates

assess whether configuration changes can be:

  • reviewed

  • version controlled

  • validated

  • rolled back

Assess whether:

Production
Development
Test
Sandbox
Shared Services

have appropriate network boundaries.

Development systems should not automatically reach production databases.

Identify high-value systems such as:

  • identity services

  • payment systems

  • regulated databases

  • security infrastructure

Ask whether stronger network isolation is appropriate.

Create:

Internet
β”‚
β”œβ”€β”€ Application Gateway :443
β”œβ”€β”€ VM :3389 ← Review
β”œβ”€β”€ Database ← High Risk
└── Public PaaS Endpoint

Which of these paths are intentionally public?

Example:

Frontend
↓
Application
↓
Database
Dev ─Xβ†’ Production

This makes lateral-movement opportunities easier to explain.

Create:

ID Finding Severity Affected Resource Priority
AZNET-001
AZNET-002

Use:

Finding ID:
Title:
Severity:
Affected Resource:
Network Path:
Observation:
Evidence:
Risk:
Business Impact:
Recommendation:
Validation:
Finding ID:
AZNET-003
Title:
Production Virtual Machine Exposes RDP to the Internet
Severity:
High
Observation:
TCP 3389 is permitted from an unrestricted internet source.
Risk:
The workload is directly reachable by internet-based attackers attempting credential compromise or exploitation.
Recommendation:
Remove unrestricted public administrative access and move administration to an approved private or controlled access path.
Validation:
Confirm RDP is no longer reachable from unapproved internet sources.
Finding ID:
AZNET-004
Title:
Application Tiers Lack Effective Segmentation
Severity:
High
Observation:
Frontend, application, and database workloads can communicate broadly without documented least-privilege restrictions.
Risk:
Compromise of one application component may enable unnecessary lateral movement.
Recommendation:
Define approved traffic flows and enforce segmentation using NSGs and/or centralized network-security controls.
Finding ID:
AZNET-005
Title:
Sensitive Workloads Have Unrestricted Internet Egress
Severity:
Medium / High
Observation:
Sensitive application workloads can initiate outbound connections to unrestricted internet destinations.
Risk:
Compromised resources may communicate with malicious infrastructure or exfiltrate data.
Recommendation:
Identify required outbound dependencies and implement appropriate egress controls and monitoring.
Finding ID:
AZNET-006
Title:
Development VNet Has Unnecessary Connectivity to Production
Severity:
High
Observation:
VNet peering and network controls permit development workloads to access sensitive production networks beyond documented requirements.
Risk:
Compromise of lower-trust development workloads may provide a path toward production systems.
Recommendation:
Restrict inter-VNet communication to explicitly required flows.

Use:

Reachability
+
Exposure
+
Trust
+
Asset Criticality
+
Exploitability
+
Business Impact
=
Risk

Examples:

  • internet-reachable sensitive database

  • exposed privileged administration surface with major impact

Examples:

  • unrestricted SSH/RDP

  • major production segmentation weakness

  • development-to-production trust

Examples:

  • unrestricted egress

  • unnecessary public PaaS access

  • incomplete telemetry

Examples:

  • stale NSG

  • unused public IP

  • unclear documentation

Recommended order:

Critical Internet Exposure
↓
Administrative Exposure
↓
Sensitive Lateral Access
↓
Routing / Firewall Bypass
↓
Unrestricted Egress
↓
Telemetry Gaps
↓
Architecture Cleanup

Use:

Immediate
0–7 Days
Short Term
7–30 Days
Medium Term
30–90 Days
Strategic
90+ Days

Example:

Finding Timeframe
Public database Immediate
Public RDP Immediate
Dev-to-prod access Short Term
Unrestricted egress Medium Term
Hub redesign Strategic

Network remediation must be tested carefully.

Validate:

Required Traffic
β†’ Works
Unauthorized Traffic
β†’ Fails

Security improvements that cause unexpected outages are not successful changes.

Validate both:

  • security

  • application functionality

For important flows document:

Source Destination Port Expected Result
Frontend Application 8080 Allow
Internet Application 8080 Deny
Application Database DB Port Allow

After Private Endpoint changes confirm:

  • service hostname

  • resolved IP

  • expected private path

  • application connectivity

Confirm:

  • expected traffic reaches firewall

  • required applications still work

  • bypass routes do not remain

Create:

Exposure Type Count
Public VMs
Public Admin Ports
Public PaaS Endpoints
Public Databases
Broad Egress Paths

Document:

  • trusted VNets

  • restricted VNets

  • hybrid networks

  • production access paths

  • administrative networks

Example:

The Azure network security review identified excessive public exposure, weak segmentation between application tiers, and broad outbound connectivity.
The highest-priority issues include internet-facing administrative access and unnecessary connectivity between lower-trust and production networks.
Immediate remediation should focus on removing unnecessary public management paths and restricting high-risk lateral connectivity.
Longer-term improvements should strengthen private service connectivity, centralized egress controls, network telemetry, and enterprise network governance.

Recommended structure:

1. Executive Summary
2. Scope
3. Network Architecture
4. VNet & Subnet Inventory
5. Public Exposure
6. NSG Review
7. Routing Review
8. Azure Firewall
9. VNet Peering
10. Private Connectivity
11. Hybrid Connectivity
12. Administrative Access
13. Network Telemetry
14. Findings
15. Risk Summary
16. Remediation Roadmap
17. Positive Observations

Include good controls such as:

  • no public administrative ports

  • strong subnet segmentation

  • centralized Azure Firewall

  • Private Endpoints for sensitive services

  • documented traffic flows

  • Sentinel network monitoring

  • controlled development-to-production trust

🎀 Azure Network Security Review Interview Questions

Section titled β€œπŸŽ€ Azure Network Security Review Interview Questions”

Practise without notes.

1. How would you perform an Azure network-security review?

Section titled β€œ1. How would you perform an Azure network-security review?”

14. How would you determine whether traffic bypasses Azure Firewall?

Section titled β€œ14. How would you determine whether traffic bypasses Azure Firewall?”

17. What risks exist with overly broad firewall rules?

Section titled β€œ17. What risks exist with overly broad firewall rules?”

26. How would you investigate an unexpected NSG change?

Section titled β€œ26. How would you investigate an unexpected NSG change?”

A production VM allows RDP from the entire internet. What would you do?

A strong answer should include:

  1. confirm business requirement

  2. determine current exposure

  3. review previous activity

  4. move administration to a controlled private path

  5. remove unrestricted RDP

  6. validate required access

  7. monitor for recurrence

An application needs access to Azure Storage, but policy prohibits public connectivity.

Discuss:

  • Private Endpoint

  • DNS

  • authorization

  • public network restrictions

  • application validation

Development workloads can reach production databases.

A strong answer should discuss:

  • business requirement

  • current network path

  • peering/routing

  • NSGs/firewall

  • least privilege

  • segmentation

  • monitoring

Traffic should pass through Azure Firewall but appears to bypass it.

Use:

Source
↓
Subnet
↓
Route Table
↓
Effective Route
↓
Firewall
↓
Destination

Review:

  • UDR association

  • route specificity

  • peering

  • gateway behavior

A Private Endpoint is deployed, but the application still connects to the service’s public IP.

Investigate:

  • DNS resolution

  • private DNS zone

  • VNet link

  • application DNS configuration

  • endpoint status

Use:

Source
↓
Destination
↓
DNS
↓
Route
↓
NSG
↓
Firewall
↓
Service
↓
Return Path
↓
Telemetry

Avoid:

β€œI would check the NSG and firewall.”

A stronger answer is:

β€œI would first define the expected source, destination, port, and network path. Then I would validate DNS, effective routes, NSG rules, centralized firewall controls, the target service, return path, and telemetry to determine exactly where connectivity is permitted or blocked.”

That sounds like a Network Security Engineer.

Create sanitized examples of:

  • VNet architecture diagram

  • subnet inventory

  • public-exposure matrix

  • NSG review

  • route analysis

  • VNet peering matrix

  • Private Endpoint review

  • traffic matrix

  • network attack-path diagram

  • findings register

  • remediation roadmap

Instead of:

Azure network security experience.

Use:

Performed a structured Azure network-security review covering VNets, subnet segmentation, NSGs, route tables, VNet peering, public exposure, Private Endpoints, and centralized network controls.

Or:

Identified internet exposure, excessive lateral connectivity, and unrestricted outbound paths in an Azure lab environment and produced prioritized remediation with traffic-flow validation.

You should be able to:

  • inventory VNets and subnets

  • identify public exposure

  • review NSGs

  • understand effective security rules

  • assess outbound access

  • review routing

  • assess Azure Firewall

  • understand hub-and-spoke

  • review VNet peering

  • assess Private Endpoints

  • troubleshoot private DNS

  • review hybrid connectivity

  • assess network telemetry

  • create defensible network findings

  • validate remediation

The objective is not:

I reviewed every NSG rule.

The objective is:

I understand which traffic paths exist, which trust relationships are necessary, which paths create unnecessary attack surface, and how to reduce risk without breaking required business communication.

You now have a repeatable methodology for:

Architecture Discovery
↓
Exposure Analysis
↓
Segmentation Review
↓
NSG Review
↓
Routing Analysis
↓
Central Firewall Review
↓
Private Connectivity
↓
Network Monitoring
↓
Risk Analysis
↓
Remediation
↓
Validation

This directly supports roles such as:

  • Azure Security Engineer

  • Cloud Security Engineer

  • Network Security Engineer

  • Cloud Security Consultant

  • Azure Solutions Architect

  • Cloud Security Architect

You have now completed the broad Azure security assessment, identity review, and network review:

Azure Security Assessment
↓
Microsoft Entra ID Security Review
↓
Azure Network Security Review

The next runbook focuses on what happens when those preventive controls fail and suspicious activity must be investigated.

➑️ Next: Azure Incident Investigation Runbook

In the next runbook, you will systematically work through:

  • incident intake

  • alert validation

  • Microsoft Sentinel incidents

  • Microsoft Entra sign-in investigation

  • Azure Activity Logs

  • Defender for Cloud alerts

  • compromised identities

  • suspicious Azure resource changes

  • compromised VMs

  • network evidence

  • incident scoping

  • containment

  • evidence preservation

  • root-cause analysis

  • recovery

  • post-incident reporting

  • incident-response interview scenarios