Per-app VPN, and when to use it
A per-app VPN sends only the traffic of selected apps through the tunnel. All other apps use the normal network of the device. This guide explains how mobile platforms do it, how it differs from split tunneling by destination, and which one fits your devices.
The MDM sends the list
An admin selects the apps in the MDM. The MDM sends the list to the device as managed configuration.
Listed apps use the tunnel
The operating system gives the traffic of each listed app to the VPN client.
Other apps go direct
An app that is not on the list uses the normal network of the device. It cannot see the tunnel.
What a per-app VPN is
A per-app VPN is a VPN setup in which the operating system decides, for each app, if its traffic goes through the tunnel. An admin gives the system a list of apps. Traffic from an app on the list enters the tunnel. Traffic from all other apps does not.
A device-wide VPN works differently. It gives the tunnel to the full device, and each app can send traffic into it.
Some people write the term as "per app VPN" or "app-based VPN". The idea is the same.
The split has two effects. Personal apps stay off the company network, which protects the privacy of the user. Apps that are not approved cannot reach company systems, which protects the company.
How iOS and Android do per-app VPN
The operating system does the split, not the VPN app. For that reason, the setup is different on each platform, and it almost always needs an MDM.
iOS and iPadOS
On Apple devices, per-app VPN is a feature of device management. The MDM installs a VPN configuration and connects managed apps to it. The tunnel can start when one of those apps opens a connection. A user cannot set this up by hand in the Settings app.
Android
On Android, a VPN app tells the system which apps are allowed to use the tunnel, or which apps are excluded from it. An MDM can send that list to the VPN app as a managed configuration. A work profile gives a second level of separation, because a VPN in the work profile carries only the apps in that profile.
Desktop systems
Per-app routing is less common on Windows, macOS, and Linux. Most desktop VPN clients split traffic by destination.
Per-app versus per-destination split tunneling
Split tunneling means that only a part of the traffic of a device goes through the tunnel. There are two ways to select that part.
- Per app. The rule names an app. All traffic from that app goes through the tunnel, wherever it goes.
- Per destination. The rule names a destination, such as a subnet, an address, or a DNS name. Traffic to that destination goes through the tunnel, whichever app sends it.
The two methods fail in different ways.
A per-app rule knows nothing about the destination. If the listed app is a browser, the user can reach each system that the tunnel can route to. The VPN gateway or a firewall behind it must still limit the destinations.
A per-destination rule knows nothing about the app. Any app on the device, including a personal app or malware, can send traffic to a permitted destination. An access policy for each user and a device check reduce this risk.
The two methods can work together. The app list decides which apps can use the tunnel, and the destination rules decide where the tunnel goes.
When each method is the better fit
BYOD phones
On a personal phone (BYOD), per-app VPN is a good fit. The company manages a small number of work apps, and personal traffic never touches the company network. The user keeps their privacy, and the company carries less traffic.
Kiosks and point-of-sale terminals
A kiosk or a point-of-sale terminal runs one app. Per-app VPN lets that one app reach the back office and gives no tunnel to any other software on the device. These devices have no user to sign in, so they also need a token that the MDM can install.
Company laptops
On a laptop, people use many tools to reach the same systems, such as a browser, a terminal, and a database client. A list of apps is hard to keep correct. Per-destination split tunneling is the better fit, with a rule for each system.
Regulated traffic
If a rule says that all traffic of a device must go through company controls, use a full tunnel. Neither per-app VPN nor split tunneling meets that requirement.
What Firezone supports
Per destination, on every platform
Firezone always splits traffic by destination. Only traffic for a Resource goes through Firezone. A Resource is a DNS name, an IP address, or a CIDR range, and a Policy says which Group can use it. All other traffic uses the normal network of the device.
This behavior is the same on Windows, macOS, Linux, iOS, Android, and ChromeOS. It needs no MDM and no list of apps.
Per app, on Android
The Firezone IoT use case page describes app selection for Android fleets. You push the Firezone Client to Android devices with Microsoft Intune or a different MDM. You then select which apps use the tunnel, so that a point-of-sale app gets through and no other app does.
The same managed configuration can sign the device in with a service account token. The Android Client then connects with no user at the device. Android keys also apply to the Android Client on ChromeOS.
What the documentation does not describe
The Firezone documentation describes app selection for Android only. It does not describe a per-app setting for iOS, macOS, Windows, or Linux. On those platforms, plan for the per-destination split. Headless operation with a token is also not available on iOS or macOS.
The managed configurations reference lists the settings that an MDM can set on each platform. The MDM provisioning guide shows how to distribute the Clients.
A device-wide VPN, a per-app VPN from an MDM, and Firezone
The three approaches select tunnel traffic in different ways.
| Device-wide VPN | Per-app VPN from an MDM | Firezone | |
|---|---|---|---|
| What selects the tunnel traffic | All traffic, or a list of subnets | A list of apps | Each Resource, and a list of apps on Android |
| Personal traffic stays off the company network | PartialOnly with split tunneling | Yes | YesSplit tunneling is always on |
| Limits which apps can use the tunnel | No | Yes | PartialOn Android, through an MDM |
| Limits which destinations a user can reach | PartialBy subnet | NoThe gateway behind it must do this | YesA Policy for each Resource |
| Works with no MDM | Yes | No | YesFor the per-destination split |
| Same behavior on desktop and mobile | Yes | NoMostly a mobile feature | PartialApp selection is for Android only |
| Unattended devices such as a kiosk | PartialWith a machine certificate | Yes | PartialToken sign-in on Android, Linux, and Windows |
The table describes the usual design of each category. A specific product can be different. Read the documentation of the product before you decide.
Common questions
What is the difference between per-app VPN and split tunneling?
The two are related. Split tunneling usually selects traffic by destination, such as a subnet. A per-app VPN selects traffic by the app that sends it. Some people call per-app VPN a type of split tunneling.
Does per-app VPN need an MDM?
On iOS and iPadOS, yes. Per-app VPN is a device management feature there. On Android, a VPN app can offer the setting itself, but a company usually sends the app list from an MDM so that users cannot change it.
Is a per-app VPN more secure than a device-wide VPN?
It reduces the number of apps that can reach company systems, which helps. It does not limit where a permitted app can go. You still need an access rule for each destination.
Does Firezone support per-app VPN on iOS?
The Firezone documentation does not describe a per-app setting for iOS. The iOS Client splits traffic by destination. Only traffic for Resources goes through Firezone, and all other traffic uses the normal network of the device.
Can I use per-app VPN on a BYOD phone?
Yes, and it is the most common use. The MDM manages only the work apps or the work profile. The tunnel carries only the traffic of those apps, and the company does not see personal traffic.