Tag Archives: nsps

Network Security Perimeters – NSPs for Troubleshooting

Posted on by
Network Security Perimeters – NSPs for Troubleshooting

This is part of my series on Network Security Perimeters:

  1. Network Security Perimeters – The Problem They Solve
  2. Network Security Perimeters – NSP Components
  3. Network Security Perimeters – NSPs in Action – Key Vault Example
  4. Network Security Perimeters – NSPs in Action – AI Workload Example
  5. Network Security Perimeters – NSPs for Troubleshooting

Hello folks! The past 3 months have been completely INSANE with customer work, building demos, experimentation and learning. It’s been awesome, but goddamn has it been grueling. I’m back to finally close out my series on Network Security Perimeters. In my past posts I’ve covered NSPs from an conceptional level, the components that make them up, and two separate examples. Today I’m going to cover my favorite use case for NSPs, and this is using them for troubleshooting.

The Setup

At a very high level most enterprises have an architecture similar what you see below. In this architecture network boundaries are setup around endpoints and services. These boundaries are erected to separate hardware and services based on the data stored in that environment, the security controls enforced in that environment, whether that environment has devices connected to the Internet, what type of trust level the humans and non-humans running in those environments have, and other similar variables. For the purposes of this post, we’re gonna keep it simple and stick to LAN (trusted) and DMZ (untrusted). DMZ is where devices connected to the Internet live and LAN is where devices restricted to the private network live.

Very basic network architecture

Environments like these typically restrict access to the Internet through a firewall or appliance/service that is performing a forward web proxy function. This allows enterprises to control what traffic leaves their private network through traditional layer 5-tuple means, deep packet inspection to inspect and control traffic at layer 7, and control access to specific websites based on the user or endpoint identity. Within the LAN there is typically a private DNS service that provides name resolution for internal domains. The DMZ has either separate dedicated DNS infrastructure for DNS caching and limited conditional fowarding or it utilizes some third-party hosted DNS service like a CloudFlare. The key takeaway from the DNS resolution piece is the machines within the LAN and DMZ use different DNS services and typically the DMZ is limited or completely incapable of resolving domains hosted on the internal DNS servers.

You’re likely thinking, “Cool Matt, thanks for the cloud 101. WTF is your point?” The place where this type of setup really bites customers is when consuming Private Endpoints. I can’t tell you how many times over the years I’ve been asked to help a customer struggling with Private Endpoints only to find out the problem is DNS related to this infrastructure. The solution is typically a simple proxy bypass, but getting to that resolution often takes hours of troubleshooting. I’m going to show you how NSPs can make troubleshooting this problem way easier.

The Problem

Before we dive into the NSP piece, let’s look at how the problem described above manifests. Take an organization that uses a high level architecture to integrate with Azure such as pictured below.

Same architecture as above but now with Azure connectivity

Here we have an organization that has connectivity with Azure configure through both an ExpressRoute with S2S VPN as fallback (I hate this fallback method, but that’s a blog for another day). In the ideal world, VM1 hits Service PaaS 1 on that Private Endpoint with the traffic being sent through the VPN or ExpressRoute connection. To do that, the DNS server in the LAN must properly resolve to the private IP address of the Private Endpoint deployed for Service PaaS 1 (check out my series on DNS if you’re unfamiliar with how that works). Let’s assume the enterprise has properly configured DNS so VM1 resolves to the correct IP address.

Now it’s Monday morning and you are the team that manages Azure. You get a call from a user complaining they can’t connect to the Private Endpoint for their Azure Storage account for blob access and gives you the typical “Azure is broken!”. You hop on a call with the user and ask the user to do some nslookups from their endpoint which return the correct private IP address. You even have the user run a curl against the FQDN and still you get back the correct IP address.

This is typically the scenario that at some point gets escalated beyond support and someone from the account team says, “Hey Matt, can you look at this?” I’ll hop on a call, get a lowdown of what the customer is doing, double-check what the customer checked, maybe take a glance at routing and Network Security Groups and then jump to what is almost always the problem in these scenarios. The next question out of my mouth is, “Do you have a proxy?”

So why does this matter? This question is important whenever we consider HTTP/HTTPS traffic because it is almost always sent through through an appliance or service that is performing the architectural function of a forward web proxy before it egresses to the Internet as I covered earlier. This could be a service hosted within the organization’s boundry or it could be a third-party service like a ZScaler. Where it’s hosted isn’t super important (but can play a role in DNS), the key thing to understand is the enterprise using one and is the application being used to access the Azure PaaS service using it.

When HTTP/HTTPS traffic is configured to be proxied, the connection from the endpoint is made to the proxy service. The proxy service examines the endpoint’s request, executes any controls configured, and initiates a connection to the outbound service. This last piece is what we care about because to make a connection, the proxy service needs to do its own DNS query which means it will use the DNS server configured in the proxy. This is typically the problem because as I covered earlier, this DNS service doesn’t typically have resolution to internal domains, which would include resolution to the privatelink domains used by Azure PaaS services.

DNS resolution when using proxy

When you did curl (without specifying proxy settings) or nslookup the machine was hitting the internal DNS service which does have the ability to resolve privatelink domains giving you the false sense that everything looks good from a DNS perspective. The resolution to this problem is to work with the proxy team to put in the appropriate proxy bypass so the endpoint will connect directly to the Azure PaaS (thus using its own DNS service).

All of this sounds simple, right? The reality is getting to the point of identifying the issue was the proxy tends to take hours, if not days, and tons of people from a wide array of teams across the enterprise. This means a lot of money spent diagnosing and resolving the issue.

What if there was an easier way? In comes NSPs.

NSPs to the rescue!

If you were a good Azure citizen, you would have wrapped this service in an NSP (assuming that service has been onboarded by Microsoft to NSPs) and turned on logging as I covered in my prior posts. If you had done that, you could have leveraged the NSP logs (the NSPAccessLogs table) to identify incoming traffic being blocked by the NSP. Below is an example of what the log entry looks like.

NSP Log Sample

In the above log entry I get detail as to the operation the user attempted to perform (in this instance listing the Keys in a Key Vault), the effect of the NSP (traffic is denied), and the category of traffic (Public). If we go back to the troubleshooting steps from earlier, I may have been able to identify this problem WAY earlier and with far less people involved even if the Azure resource platform logs obfuscated the full IP address or didn’t list it at all. I can’t stress the value this presents, especially having a standardized log format. The amount of hours my customers could have saved by enabling and using these logs makes me sad.

Even more uses!

Beyond troubleshooting the proxy problem, there are any scenarios where this comes in super handy. Another such example is PaaS to PaaS traffic. Often times the documentation around when a PaaS talks to another PaaS is not clear. It may not be obvious that one PaaS is trying to communicate with another over the Microsoft public backbone. This is another area NSPs can help because this traffic can also be logged and used for troubleshooting

Troubleshooting PaaS to PaaS

We’re not done yet! Some Azure compute services can integrate into a customer virtual network using a combination of Private Endpoints for inbound traffic and regional VNet integration for outbound access (traffic initiated by the PaaS and destined for customer endpoints or endpoints in the public IP space). A good example is Azure App Services or the new API Management v2 SKUs. Often times, I’ll work with customers who think they enabled this correctly but only actually enabled Private Endpoints and missed configuring regional Vnet integration causing outbound traffic to leave the Microsoft public backbone and hit the PaaS over public IPs or misconfigured DNS for the virtual network the PaaS service has been integrated with. NSP logs can help here as well.

NSPs helping to diagnose regional VNet integration issues

Summing it up

Here are the key takeways for you for this post:

  1. Enable NSPs wherever they are supported. If you’re not comfortable enforcing them, at least turn them on for the logging.
  2. Don’t forget NSPs capture both the inbound AND outbound traffic. You’d be amazed how many Azure PaaS services (service based) can make outbound network calls that you probably aren’t tracking or controlling.
  3. Like platform logs, NSP logs are not simply a security tool. Don’t lock them away from operations behind a SIEM. Make them available to both security and operations so everyone can benefit.

NSPs are more than just a tool to block and get visibility into incoming and outbound traffic for security purposes, but also an important tool in your toolbox to help with day-to-day operational headaches. If you’re not using NSPs for supported services today, you should be. There is absolutely zero reason not to do it, and your late night troubleshooting sessions will only consume 1 Mountain Dew vs 10!

That’s it for me. Off to snowblow!

Network Security Perimeters – NSPs in Action – AI Workload Example

Posted on by
Network Security Perimeters – NSPs in Action – AI Workload Example

This is part of my series on Network Security Perimeters:

  1. Network Security Perimeters – The Problem They Solve
  2. Network Security Perimeters – NSP Components
  3. Network Security Perimeters – NSPs in Action – Key Vault Example
  4. Network Security Perimeters – NSPs in Action – AI Workload Example
  5. Network Security Perimeters – NSPs for Troubleshooting

Hello again! Today I’ll be covering another NSP (Network Security Perimeters) use case, this time focused on AI (gotta drive traffic, am I right?). This will be the fourth entry in my NSP series. If you haven’t read at least the first and second post, you’ll want to do that before jumping into this one because, unlike my essays back in college, I won’t be padding the page count by repeating myself. Let’s get to it!

Use Case Background

Over the past year I’ve worked with peers helping a number of customers get a quick and simple RAG (retrieval augmented generation) workload into PoC (proof-of-concept). The goal of these PoCs were often to validate that the LLMs (large language models) could provide some level of business value when supplementing them with corporate data through a RAG-based pattern. Common use cases included things like building a chatbot for support staff which was supplemented with support’s KB (knowledge base) or chatbot for a company’s GRC (governance risk and compliance) team which was supplemented with corporate security policies and controls. You get the gist of it.

In the Azure realm this pattern is often accomplished using three core services. These services include the Azure OpenAI Service (now more typically AI Foundry), AI Search, and Azure Storage. In this pattern AI Search acts as the as the search index and optional vector database, Azure Storage stores the data in blob storage before it’s chunked and placed inside AI Search, and Azure OpenAI or AI Foundry hosts the LLM. Usage of this pattern requires the data be chunked (think chopped up into smaller parts before it’s stored as a record in a database while still maintaining the important context of the data). There are many options for chunking which are far beyond the scope of this post (and can be better explained by much smarter people), but in Azure there are three services (that I’m aware of anyway) that can help with chunking vs doing it manually. These include:

  1. Azure AI Document Intelligence’s layout model and chunking features
  2. Azure OpenAI / AI Foundry’s chat with your data
  3. Azure AI Search’s skillsets and built-in vectorization

Of these three options, the most simple (and point and click) options are options 2 and 3. Since many of these customers had limited Azure experience and very limited time, these options tended to serve for initial PoCs that then graduated to more complex chunking strategies such as the use of option 1.

The customer base that was asking for these PoCs fell into one or more of the these categories:

  1. Limited staff, resources, and time
  2. Limited Azure knowledge
  3. Limited Azure presence (no hybrid connectivity, no DNS infrastructure setup for support of Private Endpoints

All of these customers had minimum set of security requirements that included basic network security controls.

RAG prior to NSPs

While there are a few different ways to plumb these services together, these PoCs would typically have the services establish network flows as pictured below. There are variations to this pattern where the consumer may be going through some basic ChatBot app, but in many cases consumers would interact direct with the Azure OpenAI / AI Foundry Chat Playground (again, quick and dirty).

Network flows with minimalist RAG pattern

As you can see above, there is a lot of talk between the PaaS. Let’s tackle that before we get into human access. PaaS communication almost exclusively happens through the Microsoft public backbone (some services have special features as I’ll talk about in a minute). This means control of that inbound traffic is going to be done through the PaaS service firewall and trusted Azure service exception for Azure OpenAI / AI Foundry, AI Search, and Azure Storage (optionally using resource exception for storage). If you’re using the AI Search Standard or above SKU you get access to the Shared Private Access feature which allows you to inject a managed Private Endpoint (this is a Private Endpoint that gets provisioned into a Microsoft-managed virtual network allowing connectivity to a resource in your subscription) into a Microsoft-managed virtual network where AI Search compute runs giving it the ability to reach the resource using a Private Endpoint. While cool, this is more cost and complexity.

Outbound access controls are limited in this pattern. There are some data exfiltration controls that can be used for Azure OpenAI / AI Foundry which are inherited from the Cognitive Services framework which I describe in detail in this post. AI Search and Azure Storage don’t provide any native outbound network controls that I’m aware of. This lack of outbound network controls was a sore point for customers in these patterns.

For inbound network flows from human actors (or potentially non-human if there is an app between the consumer and the Azure OpenAI / AI Foundry service) you were limited to the service firewall’s IP whitelist feature. Typically, you would whitelist the IP addresses of forward web proxy in use by the company or another IP address where company traffic would egress to the Internet.

RAG design network controls prior to NSPs

Did this work? Yeah it did, but oh boy, it was never simple to approved by organizational security teams. While IP whitelisting is pretty straightforward to explain to a new-to-Azure customer, the same can’t be said for the trusted services exception, shared private access, and resource exceptions. The lack of outbound network controls for AI Search and Storage went over like a lead balloon every single time. Lastly, the lack of consistent log schema and sometimes subpar network-based logging (I’m looking at you AI Search) and complete lack of outbound network traffic logs made the conversations even more difficult.

Could NSPs make this easier? Most definitely!

RAG with NSPs

NSPs remove every single one of the pain points described above. With an NSP you get:

  1. One tool for controlling both inbound and outbound network controls (kinda)
  2. Standardized log schema for network flows
  3. Logging of outbound network calls

We go from the mess above to the much more simple design pictured below.

The design using NSPs

In this new design we create a Network Security Perimeter with a single profile. In this profile there is an access rule which allows customer egress IP addresses for human users or non-human (in case users interact with an app which interacts with LLM). Each resource is associated to that profile within the NSP which allows non-human traffic between PaaS services since it’s all within the same NSP. No additional rules are required which prevents the PaaS services from accepting or initiating any network flows outside of what the access rules and communication with each other within the NSP.

In this design you control your inbound IP access with a single access rule and you get a standard manner to manage outbound access. No more worries about whether the product group baked in an outbound network control, every service in the NSP gets one. Logging? Hell yeah we got your logging for both inbound and outbound in a standard schema.

Once it’s setup you get you can monitor both inbound and outbound network calls using the NSPAccessLogs. It’s a great way to understand under the hood how these patterns work because the NSP logs surface the source resource, destination resource, and the operation being performed as seen below.

NSP logs surfacing operations

One thing to note, at least in East US 2 where I did my testing, outbound calls that are actually allowed since all resources are within the NSP falsley record as hitting the DenyAll rule. Looking back at my notes, this has been an issue since back in March 2025 so maybe that’s just the way it records or the issue hasn’t yet been remediated.

The other thing to note is when I initially set this all up I got an error in both AI Foundry’s chunking/loading method and AI Search’s. The error complains that an additional header of xms_az_nwperimid was passed and the consuming app wouldn’t allow it. Oddly enough, a second attempt didn’t hit the same error. If you run into this error, try again and open a support ticket so whatever feature on the backend is throwing that error can be cleaned up.

Summing it up

So yeah… NSPs make PaaS to PaaS flows like this way easier for all customers. It especially makes implementing basic network security controls far more simple for customers new to Azure that may not have a mature platform landing zone sitting around.

Here are your takeaways for today:

  1. NSPs give you standard inbound/outbound network controls for PaaS and standardized log schema.
  2. NSPs are especially beneficial to new customers who need to execute quickly with basic network security controls.
  3. Take note as of the date of this blog both Azure OpenAI Service and AI Foundry support for NSPs in public preview. You will need to enable the preview flag on the subscription before you go mucking with it in a POC environment. Do not use it in production until it’s generally available. Instructions are in the link.
  4. I did basic testing for this post testing ingestion, searching, and submitting prompts that reference the extra data source property. Ensure you do your own more robust testing before you go counting on this working for every one of your scenarios.
  5. If you want to muck around with it yourself, you can use the code in this repo to deploy a similar lab as I’ve built above. Remember to enable the preview flag and wait a good day before attempting to deploy the code.

Well folks, that wraps up this post. In my final post on NSPs, I’ll cover a use case for NSPs to help assist with troubleshooting common connectivity issues.

Thanks!