Skip to content

Est. 2011ยทMicrosoft Partner ID 7033487ยทDelivery under 3 minยทSupport 7 days a week

Your vault is empty.

Free Fix 798

VPN error 798 and 13806: no usable certificate for IKEv2 or EAP logon

11 min read Updated October 5, 2026 Networking, Sharing & Printing

Fix it now

Microsoft publishes 798 as no certificate being found that can be used with the chosen EAP method, and 13806 as IKE failing to find a valid machine certificate. Note what 13806 does not say: it does not say the client. In an IKEv2 deployment the missing machine certificate is as often the RRAS server’s as the laptop’s, so check both.

Run these on the failing machine, in an elevated Command Prompt

certlm.msc
certutil -verify -urlfetch C:\temp\client.cer
  1. In the machine store, open Personal, then Certificates, and check the certificate you expect to be used: in date, showing that a private key is present, and with client authentication among its intended purposes.
  2. On the RRAS server, do the same in its Local Computer Personal store. The server certificate needs a subject or alternative name matching the FQDN clients dial, and its EKUs must include Server Authentication (1.3.6.1.5.5.7.3.1) and IP security IKE intermediate (1.3.6.1.5.5.8.2.2).
  3. Check the chain on both machines. The issuing authority and the root above it must be present under Intermediate Certification Authorities and Trusted Root Certification Authorities, in the machine store rather than the user store.
  4. Export the certificate you believe should be used and run the certutil line against the file. It tells you which check fails, including whether the revocation list could be retrieved.

The certificate authority list in the client’s EAP properties is the Trusted Root Certification Authorities list, and it validates the server’s certificate. It does not decide which certificate the client offers, and clearing entries in it breaks server validation.

If the tunnel now completes you can stop here. If not, the next section covers which of the three certificates involved is missing and how to tell.

Why it happens

A certificate-authenticated tunnel involves up to three certificates. The server holds one proving its identity, with server authentication among its purposes and a name matching the address the client dials. The client holds one proving either the machine or the user, with client authentication among its purposes. Both need the issuing chain to be trusted by the other side. The error tells you which side came up empty, and 13806 is the one people misread.

Microsoft’s text for 13806 is that IKE failed to find a valid machine certificate, and it directs you to install a valid certificate in the appropriate certificate store. It does not scope that to the client. On an IKEv2 or Always On deployment the certificate IKE most often cannot find is the RRAS server’s own: missing, expired, in the wrong store, or lacking the Server Authentication and IP security IKE intermediate purposes. Start by ruling the server in or out, because an article that looks only at the laptop will keep you on the wrong machine all day.

Store location is the next trap. A certificate enrolled into the user’s personal store cannot authenticate the machine, and one in the machine store is not what a user-authenticating method looks for. They look identical in the console until you notice which console you opened. The same applies to a certificate imported without its private key: it appears in the list and is useless for authentication.

Trust and revocation hold the subtler failures. A missing intermediate breaks the connection even though the certificate itself is fine, because the other side cannot build the chain. And a party that checks against a revocation list has to reach it: a list it cannot retrieve, or one that has itself expired, is treated as a failure rather than as a pass.

The RRAS server has no usable machine certificate

You have this one if 13806, and every client fails at the same point regardless of how recently it enrolled.

  1. On the RRAS server, open the Local Computer Personal store and check the certificate is in date and has a private key.
  2. Confirm its subject or subject alternative name matches the FQDN clients dial.
  3. Confirm its EKUs include Server Authentication (1.3.6.1.5.5.7.3.1) and IP security IKE intermediate (1.3.6.1.5.5.8.2.2), and reissue from a template that includes both if they are missing.
  4. Restart the Routing and Remote Access service after replacing the certificate.

There is no machine certificate on the client at all

You have this one if The machine store’s personal container is empty, or holds nothing issued by your own authority.

  1. On the internal network, run gpupdate /force and see whether automatic enrolment issues one.
  2. Check the template’s permissions. The computer or its group needs both enrol and automatic enrol rights, which is the usual reason enrolment silently does nothing.
  3. For machines rarely on the internal network, request the certificate by hand through the certificates console and confirm it lands in the machine store.

A machine that has never been on the internal network cannot enrol automatically. Plan how those devices get their first certificate before they are handed out.

The certificate exists but has the wrong purpose or template

You have this one if A certificate is present and in date, and the connection behaves as though there is none.

  1. Open it and read its intended purposes. Client authentication must be among them for a client certificate.
  2. Where the template is wrong, correct the template and reissue. A certificate cannot be edited after issue.
  3. Remove superseded certificates once the replacement works, so selection cannot pick the old one.

The chain cannot be built or trusted

You have this one if Opening the certificate shows a chain error, or the issuing authority is missing from the other side’s trusted stores.

  1. Import the root into Trusted Root Certification Authorities and any intermediate into Intermediate Certification Authorities, in the machine store.
  2. Publish the chain through Group Policy so every machine receives it rather than fixing them one at a time.
  3. Do not drop an issuing authority certificate straight into the trusted root store as a shortcut. Import the real root and the intermediate in their proper places.

It expired, was revoked, or the revocation list cannot be reached

You have this one if It worked yesterday and fails today, or it fails only for clients whose network cannot reach the publication point.

  1. Check the validity dates on both certificates and renew whichever has lapsed.
  2. Confirm the revocation list is published where the checking party can reach it, and that the list has not itself expired.
  3. Run certutil -verify -urlfetch against an exported copy to see exactly which retrieval fails.

An expired revocation list is treated as a failure, not as an absence. Publishing on schedule matters as much as publishing at all.

Full reference

Which certificate each code is complaining about

Code Published meaning Which machine to look at
798 A certificate could not be found that can be used with this Extensible Authentication Protocol The client, in the store the method uses
13806 IKE failed to find valid machine certificate Either end; check the RRAS server first on IKEv2
13801 IKE authentication credentials are unacceptable The certificate was found and rejected – purpose, chain or revocation
13805 Negotiation timed out Not a certificate fault by itself; look at reachability first
781 The connection requires a certificate, and no valid certificate was found The client
766 A certificate could not be found. L2TP over IPsec requires a machine certificate The client’s machine store

The pair to keep apart is 13806 and 13801. 13806 is nothing found. 13801 is something found and refused, which means you are past enrolment and into purposes, chains and revocation. Treating the second as the first sends people to reissue certificates that were never the problem.

Choosing which certificate the client offers

There is no control on the client’s connection properties that filters which client certificate is presented. The certificate authority list inside the EAP properties dialog is the Trusted Root Certification Authorities list, and Microsoft documents it as the list used to validate the server’s certificate. Clearing entries there does not make client selection deterministic; it breaks server validation.

  • To constrain what the client offers, turn off simple certificate selection and choose explicitly, or set certificate issuer and EKU filtering in the EAP configuration when the VPN profile is built.
  • In a deployed environment that filtering belongs in the profile, so every machine gets the same rule rather than each being fixed by hand.
  • Removing expired and superseded certificates from the store is a legitimate tidy-up and reduces what there is to choose between, but it is not a substitute for filtering.
  • Export anything you intend to remove, with its private key, before removing it.

Do not delete certificates to tidy up before you know what uses them. A machine certificate is often shared by wired and wireless authentication, Remote Desktop and management agents, and removing it to fix a VPN can take those down at the same time.

Checking a certificate without guessing

Question How
Is it in the machine store or the user store certlm.msc for the machine, certmgr.msc for the user
Does it have a private key The certificate’s General tab states it, and the icon carries a key
What are its intended purposes The Details tab, Enhanced Key Usage
Can the chain be built The Certification Path tab, on the machine that has to validate it
Does every check pass, including revocation certutil -verify -urlfetch <exported file>

When the certificate is beyond reproach

  • Confirm the name the client dials matches the server certificate’s subject or one of its alternative names, exactly.
  • Confirm the client and the server agree on the authentication method, since a method mismatch produces a certificate-shaped failure.
  • Check the clock on both machines. A certificate that is not yet valid behaves exactly like one that has expired.
  • Check whether an inspection device on the path is substituting its own certificate, which breaks validation without touching either store.
  • Test from a machine that has never connected before, which proves enrolment and trust work for a new device rather than only for one you repaired by hand.

Every code this article covers

Code What it points at Source
798 A certificate could not be found that can be used with this Extensible Authentication Protocol Microsoft Learn
13806 ERROR_IPSEC_IKE_NO_CERT: IKE failed to find a valid machine certificate. Microsoft does not scope this to the client, and on IKEv2 it is often the RRAS server’s own certificate Microsoft Learn
13801 ERROR_IPSEC_IKE_AUTH_FAIL: IKE authentication credentials are unacceptable. Something was found and refused Microsoft Learn
13805 ERROR_IPSEC_IKE_TIMED_OUT: negotiation timed out Microsoft Learn
781 The connection requires a certificate, and no valid certificate was found Microsoft Learn
766 A certificate could not be found. Connections that use L2TP over IPsec require a machine certificate Microsoft Learn

Confirm the fix worked

  1. Connect and confirm the tunnel completes without a certificate prompt or warning.
  2. On the client, confirm the certificate used is the one you intended rather than an older one still in the store.
  3. On the RRAS server, confirm its machine certificate is in date, matches the dialled name and carries both required EKUs.
  4. Run certutil -verify -urlfetch against the certificate and confirm every check passes, including revocation.
  5. Test from a machine that has never connected before, to prove enrolment and trust work for a new device.

Questions people ask about this

Do I need to buy certificates for this?

Usually not. Client and machine certificates come from your own internal authority at no cost. A public certificate is worth buying for the server side, where clients that are not domain members would otherwise have to be told to trust your internal root.

I get 13806 and the client certificate is definitely fine. What now?

Look at the server. Microsoft’s text says IKE failed to find a valid machine certificate without saying which machine, and on IKEv2 the usual answer is the RRAS server’s own certificate – missing, expired, in the wrong store, or without Server Authentication and IP security IKE intermediate in its EKUs.

How do I stop the client offering the wrong certificate?

Not in the connection’s properties. The certificate authority list there validates the server’s certificate. Turn off simple certificate selection and choose explicitly, or set issuer and EKU filtering in the EAP configuration when the profile is built.

Can I turn certificate checking off to get people connected?

Some of it can be relaxed, and doing so removes the assurance the certificate exists to provide. If you need a temporary measure, prefer a short-lived exception you have recorded over a permanent setting nobody remembers making.

Why does it work for domain machines and not for personal ones?

Domain machines enrol automatically and already trust the internal authority. A personal device has neither, so it needs the root imported and a certificate issued some other way, which is why many organisations use a public certificate on the server for those users.

Related error codes

Was this article helpful?

Your feedback helps us improve our documentation.

Related articles

Free Fix Error 0x8009030C: logon denied when NTLM is blocked or restricted Free Fix VPN error 691: username or password rejected by the remote access server License Error Wi-Fi adapter Code 43 or Code 45: the device stopped or is not present Free Fix 0x80070005 access denied on a network share you are sure you can open
โ† Back to Knowledge Base