Mutual TLS for DICOM connections
- Mutual TLS requires the client to present a certificate, authenticating the sender during the TLS handshake itself.
- A client certificate is a stronger trust anchor for DICOM senders than IP-based trust, since devices can change IP or sit behind NAT.
- Each sender can hold its own certificate, so removing one sender does not affect others on a multi-hospital service.
- mTLS secures the connection and sender identity but does not validate, de-identify, or repair DICOM content itself.
- A common pattern converts authenticated inbound DICOM C-STORE traffic into DICOMweb STOW-RS over standard HTTPS.
Transport Layer Security (TLS) protects a network connection. Mutual TLS (mTLS) adds one more check: the client also proves its identity with a certificate. For DICOM, that means the sender can be authenticated before any imaging data moves.
That matters when a hospital scanner or Picture Archiving and Communication System (PACS) connects to your product over the internet. You need encryption, but you also need to know which sender is on the other side. mTLS gives you both in the same handshake.
How is mutual TLS different from normal HTTPS?
In normal HTTPS, the server presents a certificate. The client checks that certificate against a trusted certificate authority (CA). If the check passes, the encrypted connection opens.
The client’s identity is different. In a browser flow, the server often learns who the client is later, with a password, cookie, or token. That can work for web apps, but it is a poor fit for many DICOM senders.
mTLS moves client authentication into the TLS handshake:
- The server presents its certificate.
- The client presents its own certificate.
- The server verifies that client certificate before the connection opens.
After that, both sides have a cryptographic identity for the peer. The application does not need to accept unauthenticated DICOM traffic first and then sort it out later.
Why does mTLS fit DICOM senders?
A DICOM sender is often a real device or a long-running system on another organization’s network. It may be a CT scanner, an MRI console, a modality gateway, or a PACS. It may sit behind network address translation. Its public IP address may change.
That makes IP-based trust weak. A client certificate is a better anchor because the sender must hold the matching private key. If it cannot prove that, the TLS handshake fails before the DICOM association starts.
This is also a natural fit for multi-hospital products. Each sender can have its own certificate, so your service can treat devices as separate identities. If 1 sender must be removed, you can remove that sender without changing the others.
The hard part is not the idea. It is the operating work around it. You must keep certificates, trust chains, DICOM Application Entity (AE) titles, and destination routing aligned across many senders. Once you make that order explicit, the setup becomes a checklist instead of a mystery.
What happens during setup?
A typical setup has 4 moving parts.
First, the receiving side needs a TLS endpoint that the hospital sender trusts. Second, each sender needs its own client certificate and private key. Third, the sender must be configured to use TLS when it opens the DICOM association. Fourth, the receiving side must know what to do with traffic from that sender.
Most modern DICOM software already supports this pattern. Open tools such as DCMTK’s storescu can test a TLS C-STORE connection before a hospital connects a production sender. That gives your team a safe lab path: prove the handshake, then prove the DICOM association, then send a small synthetic study.
The failure you will actually hit is usually mundane. One side trusts the wrong CA. The certificate is installed without its private key. The Called AE title does not match the configured value. These are fixable configuration errors, not unknown protocol problems.
How does mTLS help with routing?
When you receive DICOM from many hospitals, authentication and routing are connected. You need to answer 2 questions early: who is this sender, and where should this sender’s studies go?
mTLS answers the first question before DICOM data arrives. The client certificate identifies the sender during the TLS handshake. Your application can then map that authenticated sender to the correct customer, facility, or inbound workflow.
This is safer than a shared password on many devices. Shared secrets tend to spread. They also make removal hard, because changing 1 secret may affect many senders. A per-sender certificate gives you a smaller blast radius.
The limitation is real too. Certificates do not fix bad study content. They do not de-identify tags, validate patient names, or repair broken DICOM objects. mTLS proves the sender and protects the connection; your product still owns the imaging workflow after delivery.
How does this fit with STOW-RS delivery?
mTLS secures the hospital-facing DICOM connection. It is separate from what happens after your product receives the instance. Many modern products convert that inbound DICOM C-STORE flow into a DICOMweb Store Over the Web RESTful Service (STOW-RS) request.
That split is useful. Hospitals can keep sending with standard DICOM C-STORE, while your cloud service receives a normal HTTPS POST. The C-STORE and STOW-RS comparison explains the protocol boundary in more detail.
Running that boundary yourself is possible, but it becomes long-lived infrastructure. You need the listener online, the sender identities mapped correctly, and failures translated back into DICOM status codes for years. A managed DICOM endpoint such as Dicomly handles this hospital-facing edge, so your team works on the product itself. The trade-off is less low-level control and a monthly cost.
For a broader view, read how hospitals transmit DICOM over the internet. It shows how outbound hospital connections, TLS, and HTTPS delivery fit together.
What should you build first?
Start with 1 sender on paper. Write down its AE title, the host and port it will call, the client certificate it will present, and the HTTPS destination that should receive the study. Then test that path with synthetic DICOM data before you involve a hospital.
Your next useful step is to compare the authentication options. Read client certificates versus API keys for healthcare integrations, then decide which identity model you want every sender to use.
Dicomly builds infrastructure for healthcare software teams that receive DICOM studies from hospitals. Why we built Dicomly →