Transport Layer Security (TLS) and Public Key Infrastructure (PKI)

This page provides a conceptual overview of Transport Layer Security (TLS) and the related notion of Public Key Infrastructure (PKI), and sketches the security-architecture considerations in play when using CockroachDB.

Learn more:

• Manage PKI certificates for a CockroachDB deployment with HashiCorp Vault
• Certificate Authentication for SQL Clients in CockroachDB Advanced Clusters

What is Transport Layer Security (TLS)?

Modern communication systems pose a general problem: How can you safely send data over a network connection to a specific receiving party and guarantee that nobody else can eavesdrop on the message or tamper with the contents? For much of the internet and the systems that support it, the Transport Layer Security (TLS) protocol is part of the solution. TLS can establish securely authenticated and encrypted traffic between a client (the party who initiates the session) and a server (the party receiving the connection request).

Key pairs

How can you send a message that only a specific intended party can receive?

The fundamental mechanism of TLS is a pair of cryptographic keys (usually referred to as a 'key pair' for short):

• The private key is generated using a cryptographically secure random number generator (a random number generator that doesn't produce output that has a pattern to it).

• The public key is derived mathematically from the private key in such a way that:

  • Text encrypted with either can be decrypted with the other.
  • It is prohibitively computationally expensive to compute the private key from the public key.

The useful implication is that if you hold such a key pair, you can distribute the public key (frequently distributed inside of a certificate) to anyone with whom you want to communicate securely. As long as you retain sole possession of the private key, two things are guaranteed:

• Anyone who has the public key can use it to encrypt a message and send it to you, knowing that you alone can decrypt it with the private key.
• Anyone who receives a message or file that was cryptographically signed with your private key (verified with the corresponding public key) knows it must have come from you.

TLS therefore provides two critical security features:

• Encryption: it prevents message from being eavesdropped or tampered with.
• Authentication: it allows one or both parties to prove their identity that cannot be impersonated (or 'spoofed').

Mutual vs one-sided TLS authentication

TLS connections can be either one-sided, meaning that only one party must prove its identity before an encrypted session is established, or two-sided (or mutual), meaning that both must prove their identities.

In mutually authenticated TLS connections, each party must have a key pair issued by a CA that the other party trusts. This is a good system for internal components within a distributed system, and generally works well for long-lived clients.

In one-sided TLS authentication, only the server must have a key pair; clients may be able to access information without any authentication. This works well when:

• The client needs no authentication, for example, for a public read-only website or API.
• When the user must authenticate with another mechanism, such as a username/password combination or Single Sign-On (SSO).

Symmetric and asymmetric encryption in TLS

Encryption using a key pair is "asymmetric", in that different keys are used to encrypt and decrypt. Encryption is called "symmetric" where the same key is used to encrypt and decrypt—as the metal key in a mechanical lock-and-key system is used to both lock and unlock.

Symmetric encryption is more efficient than asymmetric encryption, but it requires all parties to have the single shared key prior to the exchange.

With asymmetric encryption, by contrast, each party can share their public key, allowing anyone to send them messages that only they can decrypt with their private key.

The TLS protocol employs both asymmetric and symmetric encryption:

• Asymmetric encryption using the public/private key pairs is used to securely exchange random data that is used to create shared session keys
• Session keys are then used by both parties to (symmetrically) encrypt and decrypt data for the remainder of the session.

Since a session key is created with information shared using asymmetric encryption, and is used only for the duration of a session, TLS combines the best of both asymmetric encryption (the ability to establish an encrypted communication channel without having to have first securely shared a key) and symmetric encryption (computational efficiency).

What is Public Key Infrastructure (PKI)?

Encryption is powerful and important, but without identity authentication, it's not very useful: if you don't know who is sending you messages in the first place, it's less useful to know that nobody else is tampering with the contents or eavesdropping on the conversation.

If you encrypt a message with a TLS public key, you know that only a holder of the matching private key can decrypt it. Even if the holder of the public certificate identifies themselves as your friend, your bank, your employer, or a government, you have no basis on which to trust that the certificate was ever actually held by the party you want to reach, rather than an imposter.

Authentication is a complex social and technological problem which is addressed today by the internet as a whole, as well as organizations across the world, by a mix of careful practices, social relationships, and digital technologies. This mixed solution is known as Public Key Infrastructure (PKI).

At its core, PKI is a hierarchy of cryptographically-backed trust relationships between the following categories of interested party:

• Subscribers wish to use the PKI to prove their identity to the world and offer secure services to others.
• Relying parties wish to rely on the PKI to authenticate and connect with subscribers.
• Certificate Authorities (CA) are responsible for verifying the identity of subscribers (which can include subordinate certificate authorities).

Certificates, signing, trust, and authority

The core mechanism of PKI is the PKI certificate, also commonly known as a "security certificate", "digital certificate" or "TLS certificate" (because it is used in TLS), or abbreviated "cert". In TLS the x509 certificate format is used.

A PKI certificate is a file containing the following:

• A public key to be used for encryption/decryption and to validate cryptographic signatures.
• Some metadata about:
  • the party that allegedly holds the corresponding private key and who therefore is the only one capable of decrypting messages encrypted with A, most importantly at least one name, such as a domain name in a domain name registry system (DNS). The certificate essentially functions as a badge or nametag, allowing the holder to claim to be the named party.
  • the party signing the certificate, and the certificate authority (if any) that signed its public certificate
  • the cryptographic signature created with the certificate authorities' private key
• A list of actions the holder of the certificate's private key can be trusted to perform, if the certificate is trusted.

On its own, such a digital certificate is of no more value than a paper certificate. Perhaps less value, as it can be neither scribbled upon nor burned. However, if signed by a trusted certificate authority, a digital certificate can then be used as an indicator that the entity with the corresponding private key is also to be trusted.

Signing here means generating a hash of data (such as a certificate) that can be verified with the public key, using a function that is difficult to spoof without the corresponding private key. So how does that help? This is the point where the boundaries between computing systems and social systems become very murky. By signing a certificate (and anyone with a private key can do this) a party is acting as a "certificate authority"; they are in effect asserting the validity of the identity claim being made by the certificate holder.

The premise of PKI is that if I present you with a certificate says that I work for the Cockroach Labs documentation team, and you trust the certificate that signed it (for example, a Cockroach Labs certificate authority), because you trust the signing certificate you trust that I work for the documentation team (or at least, I did when the certificate was signed).

Generally, given that you can put your trust in the organization that backs the Certificate Authority who has signed a server's public certificate, you can extend your trust that the holder of the private key corresponding to that certificate is who the certificate says they are. For example, this is how your web browser or mobile app knows that it's actually talking to your bank, rather than an imposter.

Furthermore, one Certificate Authority may grant certify that another party is authorized to issue certificates on its behalf, acting as a subordinate CA.

Public and private PKIs

On the public internet, certificate authority providers such as Identrust, Digicert, and Let's Encrypt provide the role of root CAs (or "trust anchors") to the entire system. What makes them "trust-worthy"? In practice, just the fact that they are trusted by the parties that distribute hardware and software (such as operating system distributions and browsers) packages that come pre-loaded with trust stores.

Trust store

A trust store is a collection of public certificates for trusted CAs—CAs whose signed certificates will be accepted for purposes of identity verification. When you use a hardware device or software package that comes loaded with a trust store, you are trusting the judgment of the company selected the CAs to add to the package's trust store.

It is up to each hardware or software vendor to decide which root CAs to include in trust stores they distribute, and ultimately the end user decides which vendors to trust. CAs must comply with formalized industry standard baseline requirements to maintain good standing with vendors.

Internally, organizations must maintain their own trust architectures, deciding what parties (individual persons), should have access to what network, computing, data, financial, and physical resources, and using certificates or other means to authenticate identity and establish encryption. Certificates and TLS are powerful tools, provided the trusted certificate authorities are properly protected (and the infrastructure that they exist in are carefully protected according to the principle of least privilege), certificates are only issued for systems that they should be, and the private keys involved are protected against exfiltration from the machines using them.

Revoking trust

Over time, employment and business relationships change, computing systems are deployed and destroyed, and even the most carefully-hidden passwords and private keys may be accidentally leaked or intentionally stolen. For all of these reasons, in order to maintain its trustworthiness, a certificate authority (CA) must be able to revoke guarantees it has issued in the form of signed certificates when those guarantees no longer hold.

The main solutions to this problem are:

• Issuing only short-lived certificates, which are effectively revoked by discontinuing their rotation.
• Certificate Revocation Lists (CRLs) allow relying parties to check if a subscriber's certificate has been revoked.
• The Online Certificate Revocation Protocol (OCSP) allows relying parties to check the status of certificates in real time.

Each strategy has pros and cons in terms of security and operational overhead. CockroachDB does not support CRLs.

See Revoking Certificates in CockroachDB.

PKI in CockroachDB

In a CockroachDB cluster, each node must be able to initiate HTTP requests to any of the others. In a normal operating mode, these requests are TLS encrypted with mutual authentication, requiring that each node present its own certificate, signed by the same CA.

In addition, SQL clients and DB Console clients must be able to reach the nodes. The client must authenticate the server with a certificate signed by a trusted CA, and the server must authenticate the client, either with its own certificate, or with another method, such as username/password.

Therefore, the nodes must each have a private key/public certificate pair where the public certificates have been signed by a common CA (called the node CA here).

If the client is to use mutual authentication the client must have a private key/public certificate pair, where the public certificate is signed by a CA trusted by the nodes, i.e. the CA's public certificate must be in the nodes' trust stores.

CockroachDB self-hosted customers must provision PKI certificates for both internode and client-cluster communication.

Refer to Manage PKI certificates for a CockroachDB deployment with HashiCorp Vault for procedural information on administering and using client certificate authentication.

PKI in CockroachDB Cloud

Certificate authentication for SQL clients is available for CockroachDB Advanced clusters.

Refer to Certificate Authentication for SQL Clients in CockroachDB Advanced for procedural information on administering and using client certificate authentication.

CockroachDB's TLS support and operating modes

TLS encryption and server authentication are supported in all communication between CockroachDB nodes, and from clients to nodes.

Currently, mutual TLS authentication, in which the client as well as the server uses a private key/public certificate pair to authenticate itself, is not supported in CockroachDB Cloud. Clients must use username/password combinations.

CockroachDB self-hosted does supports TLS authentication for clients.

Default mode

By default, CockroachDB clusters require TLS. Client connection requests must be made with sslmode=on. The CockroachDB CLI cockroach sql command, by default, is made with sslmode=on.

Clients always require a public certificate for the CA that issued the server certificate.

If the connection is mutually TLS-authenticated (i.e., if the client authentication method is a certificate rather than a username/password combination), then a private key/public certificate pair for the client is also required.

--accept-sql-without-tls mode

CockroachDB clusters can be started with this option in order to allow clients to opt out of TLS server authentication.

Client connections can be made with sslmode=on, in which case TLS authentication proceeds as in default mode. Client connections can also be made sslmode=off (or by using cockroach sql --insecure on the CLI), in which case TLS authentication is skipped.

Regardless of whether TLS authentication is performed or skipped, TLS encryption is still used for all traffic.

--insecure mode

CockroachDB can be operated entirely without TLS. If a CockroachDB cluster is started with the --insecure flag, all TLS encryption and authentication is skipped.

This can be useful to quickly deploy a short-lived local cluster for testing and hands on experimentation, but is unsuitable for long-lived deployments.

Note that client connections must also be made insecurely, or the connection request will fail. Do this by using cockroach sql --insecure on the CLI, or by setting sslmode=off in the database connection string.

The CockroachDB certificate Trust Store

A node's trust store is the set of CA public certificates contained in the directory specified by the --certs-dir argument when the node is started using cockroach start. For each CA public certificate in the trust store, the node will accept all valid certificates signed by the CA or any CA subordinate to it.

CockroachDB ignores operating system certificate trust stores.

TLS between CockroachDB nodes

Connections between CockroachDB nodes are always mutually TLS-authenticated. Each node must have at least one private-key/public certificate pair, which it can use both as server and as client when initiating internode traffic.

CockroachDB Cloud

Customers using CockroachDB Cloud need not worry about managing TLS keys for CockroachDB nodes, as cluster management is delegated to the Cockroach Labs team.

CockroachDB self-hosted

By default, a CockroachDB node makes double use of a single key pair, using it to authenticate both as client when initiating a connection to another node, and as server when receiving requests.

The node must also have a trust store containing the public certificate of at least one trusted CA—the one that issued the public certificates of any nodes it must connect with.

TLS in CockroachDB SQL client connections

CockroachDB provides a number of SQL clients, including a CLI, and a number of drivers and object-relational mapping (ORM) tools. Regardless of which client you are using, how you are able to authenticate to a CockroachDB cluster depends on that cluster's authentication configuration, specifically whether that configuration requires the user to authenticate with username/password combination, certificate or another method.

In turn, which authentication methods are available depends on the sort of environment in which a CockroachDB cluster is deployed, as described in the following subsections.

CockroachDB Cloud

CockroachDB Cloud does not support certificate-authenticated client requests. TLS is used to authenticate the server and encrypt all traffic, but the user must authenticate to the database with a username/password combination.

Because the server must still be TLS-authenticated, the client must know to trust the certificate authority that signed the public certificate identifying the server. The path to the CA's public certificate is passed as the sslrootcert parameter in a database connection string, or by being placed in the directory specified by the certs-dir argument in a connection made with the cockroach sql CLI command.

CockroachDB self-hosted

CockroachDB self-hosted clusters support TLS authentication for clients, i.e. mutual TLS authentication. Other supported authentication methods are username/password combination, and GSSAPI/Kerberos (Enterprise only).

Non-TLS client authentication

When using a non-TLS client authentication method, such as username/password or GSSAPI/Kerberos (Enterprise only), the server must still be TLS-authenticated. Therefore, the client must know to trust the certificate authority that signed the public certificate identifying the server. Therefore, the root CA certificate, called ca.crt, must be provided to client authentication attempts. This can be passed as the sslrootcert parameter in a database connection string, or by being placed in the directory specified by the certs-dir argument in a connection made with the cockroach sql CLI command.

TLS client authentication

For a SQL client to authenticate using TLS, that client must be provisioned with its own key-pair:

• client.<username>.crt is the client's public certificate.
• client.<username>.key is the client's private key.

<username> here corresponds to the username of the SQL user as which the client will issue SQL statements if authentication is successful.

These key files are used with the CockroachDB CLI by placing them in a directory and --certs-dir

Revoking Certificates in CockroachDB

CockroachDB Cloud

Customers of CockroachDB Cloud delegate responsibility for maintaining cluster internal PKI to the Cockroach Labs team, and need not worry about securing authentication and encryption between cluster nodes.

Clusters in CockroachDB Cloud do not support certificate based client authentication, but rely instead on username/password combination.

CockroachDB self-hosted

CockroachDB does not support certificate revocation lists (CRLs). The remaining options are the Online Certificate Status Protocol (OCSP), and reliance on a rapid cycle of generating and propagating short-lived certificates.

OCSP

Securely operating an OCSP responder is a significant task, and it would not be recommended to undertake this solely for the purposes of securing a CockroachDB cluster.

CockroachDB can be configured to check an OCSP responder.

Short-lived certificates

For many self-hosted customers, we recommend a strategy of relying on a short "lifetime", i.e., validity duration, for credentials (private key/public certificate pairs) issued to CRDB nodes and SQL clients. This strategy is relatively easy to implement in an automated pipeline using cloud native secrets management tools such as GCP secrets manager or HashiCorp Vault to securely propagate certificates, and synergizes with the use of cloud native CA tools, such as GCP's Certificate Authority Service (CAS). Using these tools, an operator can create an automation pipeline to generate and propagate certificates.

By relying on certificates with a short validity duration, we can greatly reduce the threat posed by such a certificate being leaked, since the certificate's value to an attacker is limited by its validity duration. To maintain connection to a network secured by short-lived credentials, a would-be-attacker must maintain access to the secret manager used to propagate the credentials. This consolidates the risk surface of the certificate itself into the shadow of access to the secrets manager.

The trade-off with short-lived certificates (or requiring any low-latency revocation system), is that it can become a single point of failure for service availability. If network connections depend on hourly propagation of fresh credentials, then a leaked credential only offers an attacker a one hour window of exploitation, but taking the credential automation offline for more than an hour can take the entire system offline. Fine tuning the validity duration to meet your threat model and available resources is an important component of designing a minimally secure private PKI without using another revocation mechanism.

See: Managing security certificates with HashiCorp Vault