Interkey Intelligent DPI

Network governance for operators, ISPs and regulators

VPN & tunnelled traffic classification for telecom operators

VPN and tunnelled traffic classification identifies which flows are carried inside tunnels, and which application classes they belong to, without decrypting them. An operator needs this to describe its traffic accurately and to apply policy where its own rules and applicable regulation require it. Interkey Intelligent DPI classifies VPN and other evasive applications from characteristics that stay observable when traffic is encrypted. The result is a traffic class, not an identity, and policies act per application, subscriber group or cell. This page answers how tunnelled classes are used by operators in reports and policies; for what detection covers and who uses it, see VPN detection.

A Saudi product, built by Interkey

For network engineering, policy and regulatory compliance teams at telecom operators and ISPs, and for regulators. Reviewed November 2026.

  • What it does:classifies VPN and other tunnelled or evasive traffic as traffic classes, without decrypting it.
  • Why it is hard:these applications keep changing to avoid detection and disguise their traffic as legitimate services.
  • How it keeps up:features are extracted automatically from live traffic and the AI traffic model is retrained automatically.
  • What it is not:it is not a tool for monitoring individuals. The output is metadata, and any action is set by your policy and applicable regulation.

Why VPN and tunnelled traffic is a classification problem

A tunnel wraps one flow inside another, so what an operator sees at first is the tunnel, not the application inside it. Encryption removes the payload signal, and tunnels often use the same common ports as everyday services. Classification by port number or payload alone therefore either misses tunnelled traffic or attributes it to the wrong class.

Some applications go further. VPN and other evasive applications keep changing to avoid detection and disguise their traffic as legitimate services. For an operator this has two costs: traffic reports undercount what is really on the network, and any policy that refers to these classes acts on the wrong flows.

How Interkey Intelligent DPI classifies tunnelled traffic

Signals used to classify tunnelled and encrypted flows
CharacteristicWhy it helps
Handshake structureTunnels and applications set up encrypted sessions in distinguishable ways
Packet sizing and timingTunnels carrying video, calls or file transfer produce different patterns from the services they imitate
Endpoint behaviourApplications reach characteristic sets of services over time

An AI traffic model built into the engine core combines these features (multi-modal feature extraction) rather than relying on any single one. Automated online probing extracts features from live traffic, the model is retrained automatically, and Live View results feed back to refine models and policies. Traffic is not decrypted at any stage. For the method in depth, see Encrypted traffic classification without decryption.

Governance boundaries

Interkey Intelligent DPI is built and offered for network governance and regulatory compliance, not for surveillance of individuals.

What tunnelled-traffic classification does and does not do
Classification doesClassification does not
Identify which application class a flow belongs to, including tunnelled classesDecrypt traffic or read its content
Report traffic volumes by classDecide what is permitted: that is set by your policy and applicable regulation
Record which policies a flow triggeredSingle out individual users: policies act per application, subscriber group or cell
Produce metadata about applications, traffic volumes and triggered policiesSet retention or access: those follow your data-governance rules and the contract

The platform runs on your infrastructure and can operate entirely in the Kingdom.

What operators use the result for

Uses of VPN and tunnelled traffic classification
UseHow classification helps
Accurate traffic reportingTunnelled traffic appears as its own classes instead of being hidden inside other categories
Capacity planningShows how much load tunnelled classes add, where and when
Policy where rules require itLets operator-defined policies refer to tunnelled classes, with blocking only where your policy and applicable regulation require it
Regulatory reportingGives regulatory teams a description of traffic that includes tunnelled classes
Security visibilityShows tunnelling and traffic that does not behave like the application it presents itself as

Before any policy refers to tunnelled classes, write down which classes are in scope, the basis for each rule, and who approves it. Tunnels also carry legitimate business connectivity, so a rule needs a clear basis and a review route.

Measure accuracy on your tunnelled traffic

Classifier results depend on subscriber mix, application mix, the share of encrypted traffic and how recently the classifier was updated. Evasive applications keep changing, so measure tunnelled classes on your own traffic and re-check after significant updates.

  1. 1. Build a verified sample.Generate tunnelled and non-tunnelled traffic from known applications on devices you control.

  2. 2. Classify it out-of-band.The platform receives a copy of traffic without touching the forwarding path.

  3. 3. Report per class.Measure detection and false-positive rates for each tunnelled class. A false positive means a policy acts on traffic that was not tunnelled.

Interkey's product material states 99% recognition accuracy, including for VPN and other adversarial encrypted applications. The figure has not been independently benchmarked and is not a specification for any particular network. See Measuring DPI classification accuracy on your traffic.

Test it on your network in 3 stages

  1. 1. Written acceptance criteria.Agree which tunnelled classes are tested, on which traffic, how results are measured and what counts as a pass.

  2. 2. Out-of-band classification against verified traffic.Measure false positives as carefully as detection.

  3. 3. Report-only policy, then the inline decision.Any policy that refers to tunnelled classes shows what it would have done before it acts; inline enforcement follows only after classification has proven itself.

See also: How Interkey runs a DPI proof of concept

Writing a rule that refers to a tunnelled class

A rule that refers to a tunnelled class is a rule about a whole class of traffic, and tunnels also carry legitimate business connectivity. Before the rule is written, settle four things in a short note: which tunnelled classes are in scope, which subscriber groups or cells the rule covers, what the rule does when it triggers, and the basis for it under your policy and applicable regulation. Name an owner for the note. Then run the rule in report-only mode and read the result before it acts. Interkey Intelligent DPI shows what the rule would have done, so the owner can check that it refers to tunnelled flows and not to ordinary traffic that looks similar. Only after that does the move to enforcement come up, and it needs an inline deployment, because policy acts on traffic only there. Keep the note with the rule. It is what you show product, care and compliance teams when they ask why a class is treated the way it is.

If you are looking for what VPN detection covers, who uses it and what it leaves out, read VPN detection.

Questions about VPN traffic classification

Does Interkey Intelligent DPI decrypt VPN traffic?

No. It classifies tunnelled and encrypted flows from observable characteristics such as handshake structure, packet sizing and timing, and endpoint behaviour.

Does classifying VPN traffic mean blocking it?

No. Classification identifies the traffic class. What happens to any class is set by your policy and applicable regulation, and a policy acts on traffic only in an inline deployment.

Does it identify who is using a VPN?

The platform is built for network governance and regulatory compliance, not for surveillance of individuals. Its output is metadata about applications, traffic volumes and triggered policies, and policies act per application, subscriber group or cell. Retention and access follow your data-governance rules and the contract.

How does it keep up with new VPN applications?

Features are extracted automatically from live traffic, the model is retrained automatically, and Live View results feed back into models and policies. Re-check accuracy after significant updates.

Can we test it without affecting traffic?

Yes. Out-of-band classification against a sample you have verified is the first technical step, before any policy acts on traffic.

How is it priced?

On a quote basis, depending on deployment, capacity, throughput, network placement and the capabilities in scope. There is no published price list.

Proof of concept in three phases

  1. Written acceptance criteria.Agree which rules are tested, on which traffic and subscriber groups, how results are measured and what counts as a pass.

  2. Out-of-band classification against verified traffic.Measure false positives as carefully as detection for every class the rules depend on.

  3. Report-only policy, then the inline decision.Each rule shows what it would have done; enforcement follows only after classification has proven itself.

Discuss a proof of concept on your network

Test tunnelled-traffic classification on your network

Tell us which tunnelled or evasive application classes matter to your reporting or policy, whether you want to start out-of-band, and your timeline. Our team in Riyadh replies in Arabic or English.

  1. Our engineers read your request
  2. We agree the scope of a technical session
  3. You see the product on a use case you choose

Tell us your network type (mobile, fixed, ISP), approximate traffic volume and the use case you are evaluating.

By sending this form, you agree to our Privacy Notice.