Australian cloud for business software — built and run in Sydney since 2002

NetLeverage
NetLeverage

Wildcard or individual certificates for hosted services: what a shared key puts at risk

Technical article
Hosting

About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.

A wildcard certificate, such as one for *.example.com, covers every name directly under your domain with a single certificate and a single private key. That is its appeal: one purchase, one renewal, no waiting when a new service appears. It also means one secret speaks for all of those services. If the key is copied, anyone holding it can present your certificate for any matching name, and one revocation takes every service with it.

A wildcard is a reasonable choice when the names share one owner, one level of sensitivity and a small, well-controlled set of servers, and when issuance is automated. Individual certificates are the better default when services belong to different teams or suppliers, differ in risk, or run on many machines. This guide is for business owners and IT administrators deciding what to ask for. It sets out the standards behind the trade-off, what a wildcard changes about issuing and revoking, and the questions worth putting to a hosting provider.

Why it matters

Certificate lifetimes are shrinking. The CA/Browser Forum Baseline Requirements, version 2.3.1, set a maximum subscriber certificate validity of 200 days from 15 March 2026, 100 days from 15 March 2027 and 47 days from 15 March 2029 (section 6.3.2). Every renewal of a wildcard has to reach every server that uses it. A short, illustrative calculation of our own: at 47 days, a certificate renews at least eight times a year, so a wildcard on six servers means at least 48 installs a year if copied by hand.

Each copy is also a place the key can leak. If a wildcard key sits on a web server, a mail server, a test machine and a contractor's laptop, an attacker needs to reach only the weakest of them. Our reading: the cost of a wildcard is not paid at issue time but at the moment one of those copies is lost.

What a wildcard actually matches

The rules for matching come from RFC 9525, the IETF standard on checking a server's identity in TLS. A wildcard must be a single * that is the whole of the left-most label, and "A wildcard in a presented identifier can only match one label in a reference identifier." So *.example.com matches shop.example.com, but not a.shop.example.com. Our reading of the same rule is that it does not match example.com itself, because there is no label to stand in for the star; many certificates list the bare name alongside the wildcard for that reason.

Let's Encrypt, a free CA, describes the benefit this way: "A wildcard certificate can secure any number of subdomains of a base domain", and it allows administrators to use "a single certificate and key pair for a domain and all of its subdomains, which can make HTTPS deployment significantly easier."

CAs also check that a wildcard cannot span a whole public registry. The Baseline Requirements section 3.2.2.6 requires a documented procedure to find out whether the name is "registry-controlled" or a "public suffix", and says CAs must refuse such issuance unless the applicant proves control of the entire namespace. For an ordinary business this just means *.co.uk is not available; *.yourname.com.au is.

What the standard says about the risk

RFC 9525 section 7.1 puts the trade-off plainly: "Wildcard certificates automatically vouch for any single-label hostnames within their domain, but not multiple levels of domains." It adds that this "can be convenient for administrators but also poses the risk of vouching for rogue or buggy hosts." The same section notes that application protocols can disallow wildcard certificates entirely as a more foolproof mitigation.

The Baseline Requirements recognise misuse too. Section 4.9.1.1 lists, among reasons for revocation, a CA becoming aware that a wildcard certificate "has been used to authenticate a fraudulently misleading subordinate Fully-Qualified Domain Name". For that item the CA should revoke within 24 hours and must within 5 days.

Three consequences follow, each our reading rather than a published rule:

  • Blast radius. The key is one secret for every name. If it leaks, the right response is to replace the certificate on every service, not just the one that was breached.
  • Revocation is all or nothing. Revoking the wildcard certificate removes trust from all names at once. With individual certificates, a leak affects only the services that shared the key.
  • Trust is only as strong as the weakest holder. A throwaway test server holding the production wildcard key is a production risk.

Cross-protocol confusion: a separate problem from key sharing

The ALPACA research describes an attack that exploits "TLS servers implementing different protocols but using compatible certificates, such as multi-domain or wildcard certificates". An attacker who can redirect a visitor's traffic can steer a secure web connection to, for example, an email or file-transfer server that presents a certificate that also matches the web name, and the browser accepts the session. The authors found 1.4 million web servers generally vulnerable to such cross-protocol attacks, with 114,000 attackable using an exploitable application server.

Two different risks are easy to confuse here, so keep them apart:

  • Key compromise. If one copy of a private key leaks, anyone can impersonate every name on the certificate. Giving each server group its own key (and its own certificate) limits how many servers hold that key, and so how many must be rotated after a leak. It does not narrow what a stolen key can impersonate: if the group's certificate still carries the full wildcard, the thief can still present it for every name the wildcard covers. Only narrower names on the certificate limit that scope.
  • Cross-protocol confusion. The redirected connection reaches a real server that holds a real key. Separate keys alone do not stop it, because the certificate the mail or file-transfer server presents can still match the web name the browser asked for. Our reading of the ALPACA page is that the defence lies in what the client and server check, and in how widely a certificate's names overlap.

The project's page proposes "the use of the Application Layer Protocol Negotiation (ALPN) and Server Name Indication (SNI) extensions in TLS" as the countermeasure. It explains that SNI lets a client name the host it wants and a server may end the connection if the name does not match, while with ALPN a server should end the connection if no supported protocol is requested; it adds that the strict SNI check is rarely implemented, even among web servers. Our recommendation, not a published rule: ask which of your servers enforce strict SNI and ALPN, and avoid a certificate whose names also match a service of a different protocol that does not need them (for example, a wildcard shared by web, mail and file transfer).

Keep this in proportion. The project's own FAQ calls the vulnerability "very situational and can be challenging to exploit", and answers the question of whether an administrator should drop everything and fix it with "Probably not."

How issuing changes

An individual certificate can usually be validated by placing a file on the web server (the HTTP-01 method). A wildcard cannot. Let's Encrypt's FAQ states that "Wildcard issuance must use the DNS-01 challenge", and its challenge guide says HTTP-01 "cannot be used to issue wildcard certificates".

DNS-01 proves control of the domain by placing a record in its DNS zone. For automatic renewal, software needs write access to DNS. The Let's Encrypt guide warns that "putting your full DNS API credentials on your web server significantly increases the impact if that web server is hacked." It recommends narrowly scoped credentials, or doing DNS validation from a separate server and copying the certificate to the web server. It also describes delegating the validation record with a CNAME to a separate zone.

So a wildcard moves risk in two places. The certificate key is now shared, and the credential that can edit your DNS may now sit on a machine you would not otherwise trust with it. Our recommendation: when you ask for a wildcard, ask where the DNS credential lives and what it is allowed to change.

One wildcard does not require one key

A wildcard name and a shared key are different things. Our reading is that a certificate for the same wildcard name can be requested more than once, each time with a new key pair held by one server group. That keeps the convenience of one name while limiting each key to a smaller set of servers, which contains a key leak but does not by itself address cross-protocol confusion (see above). Check whether your CA and tooling allow this and what limits apply; the sources we read do not specify. It also costs more installs and more tracking.

Visibility is the other side. Let's Encrypt notes in its FAQ that issued certificates are submitted to public Certificate Transparency logs, where automated crawlers can find the names. Our reading: a wildcard hides which individual hostnames you use inside the zone, which can be a legitimate reason to choose one, but is not a substitute for protecting the key.

Decision: when each fits

Our recommendation, not a published rule:

  • A single wildcard is reasonable for names under one owner and one security level, on few servers, with automated renewal and a short list of people who can export the key.
  • Separate wildcards by zone (for example one for test and one for production, on different subdomains) when you want the convenience without letting a test machine vouch for production.
  • Individual certificates suit services with different owners, third-party hosts, payment or login systems, or anything needing separate revocation. The cost is more certificates to track, which automation handles better than people do.
  • Separate keys by server group even where a wildcard name is kept, to reduce how many servers hold a key and must be rotated after a leak; to limit which names a stolen key can impersonate, narrow the names on the certificate.
  • Reduce overlapping names across protocols and ask whether servers enforce strict SNI and ALPN, because that, not key separation, is how the ALPACA authors propose to prevent cross-protocol confusion.

Read-only checks

These steps read information; they change nothing. The command below was not tested by the author. It prints the subject, the names a certificate covers and its public key from a certificate file (the public certificate, never the private key). Options are on the OpenSSL x509 manual page.

Linux shell, or any system with OpenSSL:

openssl x509 -in cert.pem -noout -subject -ext subjectAltName -pubkey

If the same public key appears in certificates collected from different servers, our reading is that those certificates share one key. That does not prove each named server holds a copy of the private key: TLS may end on a shared proxy, load balancer or hardware key store in front of them. Next, find out where TLS actually terminates for each name and which systems can use or export the key. Build a list of those systems, who administers each, and where the key is also copied (backups, deployment tools, repositories).

Questions to put to your provider

  1. Is our certificate a wildcard, and which names and servers use it?
  2. How many copies of the private key exist, where are they, and who can export them?
  3. Do different services or different customers share the same key?
  4. Where is the DNS credential used for wildcard renewal stored, and what can it change?
  5. How would you replace the certificate on every service if the key leaked, and how long would that take?
  6. Can mail, file transfer and web services use separate keys?

A worked example (illustrative)

The following is invented. A firm buys a wildcard certificate for its domain and installs it on its customer portal, its mail gateway, a file-transfer server and a development server run by an outside contractor. Each renewal is copied by hand. When the contractor's laptop is lost, the firm cannot tell whether the key was on it.

The firm replaces the certificate on each of those four systems in one afternoon. It then splits the setup: the portal and mail gateway get individual certificates with separate keys, the development server gets a wildcard for a test subdomain only, and renewal is automated from a separate validation server with a narrowly scoped DNS credential.

Risks and limits

  • Wildcards are widely used and accepted by browsers; nothing here says they are forbidden or unsafe by themselves.
  • Which model suits you depends on your servers, owners and tooling. The standards describe risk but do not rank options for a particular business.
  • ALPACA is a research finding with stated preconditions; this guide does not assess whether any system is vulnerable.
  • Some software treats wildcards differently from the RFC 9525 rules. We did not test any.
  • The article is not lab tested and the command was not run.

Next step

Count the servers, tools and backups that hold the private key for any wildcard certificate you use, and write down who can export it. If you are reviewing a hosted service, the Managed Services page is a starting point for that conversation; ask for the scope in writing.

Sources

All accessed 15 November 2026.

Learning notes

  • A wildcard certificate covers many names with one key. Whoever holds the key can speak for all of them.
  • A wildcard matches one label only: shop.example.com, not a.shop.example.com.
  • Wildcards need the DNS-based validation method, so renewal automation often needs DNS write access.
  • A shared name does not force a shared key; splitting keys by server group limits the damage of a key leak. It does not by itself stop cross-protocol confusion, which is addressed by strict SNI and ALPN checks and by fewer overlapping names.
  • If a wildcard key leaks, every service using it needs a new certificate.