By default, PostgreSQL will not perform any verification of the server certificate. This means that it is possible to spoof the server identity (for example by modifying a DNS record or by taking over the server IP address) without the client knowing. In order to prevent spoofing, the client must be able to verify the server's identity via a chain of trust. A chain of trust is established by placing a root (self-signed) certificate authority (CA) certificate on one computer and a leaf certificate signed by the root certificate on another computer. It is also possible to use an “intermediate” certificate which is signed by the root certificate and signs leaf certificates.

To allow the client to verify the identity of the server, place a root certificate on the client and a leaf certificate signed by the root certificate on the server. To allow the server to verify the identity of the client, place a root certificate on the server and a leaf certificate signed by the root certificate on the client. One or more intermediate certificates (usually stored with the leaf certificate) can also be used to link the leaf certificate to the root certificate.

Once a chain of trust has been established, there are two ways for the client to validate the leaf certificate sent by the server. If the parameter sslmode is set to verify-ca, libpq will verify that the server is trustworthy by checking the certificate chain up to the root certificate stored on the client. If sslmode is set to verify-full, libpq will also verify that the server host name matches the name stored in the server certificate. The SSL connection will fail if the server certificate cannot be verified. verify-full is recommended in most security-sensitive environments.

In verify-full mode, the host name is matched against the certificate's Subject Alternative Name attribute(s), or against the Common Name attribute if no Subject Alternative Name of type dNSName is present. If the certificate's name attribute starts with an asterisk (*), the asterisk will be treated as a wildcard, which will match all characters except a dot (.). This means the certificate will not match subdomains. If the connection is made using an IP address instead of a host name, the IP address will be matched (without doing any DNS lookups).

To allow server certificate verification, one or more root certificates must be placed in the file ~/.postgresql/root.crt in the user's home directory. (On Microsoft Windows the file is named %APPDATA%\postgresql\root.crt.) Intermediate certificates should also be added to the file if they are needed to link the certificate chain sent by the server to the root certificates stored on the client.

Certificate Revocation List (CRL) entries are also checked if the file ~/.postgresql/root.crl exists (%APPDATA%\postgresql\root.crl on Microsoft Windows).

The location of the root certificate file and the CRL can be changed by setting the connection parameters sslrootcert and sslcrl or the environment variables PGSSLROOTCERT and PGSSLCRL. sslcrldir or the environment variable PGSSLCRLDIR can also be used to specify a directory containing CRL files.

If the server attempts to verify the identity of the client by requesting the client's leaf certificate, libpq will send the certificate(s) stored in file ~/.postgresql/postgresql.crt in the user's home directory. The certificates must chain to the root certificate trusted by the server. A matching private key file ~/.postgresql/postgresql.key must also be present. On Microsoft Windows these files are named %APPDATA%\postgresql\postgresql.crt and %APPDATA%\postgresql\postgresql.key. The location of the certificate and key files can be overridden by the connection parameters sslcert and sslkey, or by the environment variables PGSSLCERT and PGSSLKEY.

On Unix systems, the permissions on the private key file must disallow any access to world or group; achieve this by a command such as chmod 0600 ~/.postgresql/postgresql.key. Alternatively, the file can be owned by root and have group read access (that is, 0640 permissions). That setup is intended for installations where certificate and key files are managed by the operating system. The user of libpq should then be made a member of the group that has access to those certificate and key files. (On Microsoft Windows, there is no file permissions check, since the %APPDATA%\postgresql directory is presumed secure.)

The first certificate in postgresql.crt must be the client's certificate because it must match the client's private key. “Intermediate” certificates can be optionally appended to the file — doing so avoids requiring storage of intermediate certificates on the server (ssl_ca_file).

The certificate and key may be in PEM or ASN.1 DER format.

The key may be stored in cleartext or encrypted with a passphrase using any algorithm supported by OpenSSL, like AES-128. If the key is stored encrypted, then the passphrase may be provided in the sslpassword connection option. If an encrypted key is supplied and the sslpassword option is absent or blank, a password will be prompted for interactively by OpenSSL with a Enter PEM pass phrase: prompt if a TTY is available. Applications can override the client certificate prompt and the handling of the sslpassword parameter by supplying their own key password callback; see PQsetSSLKeyPassHook_OpenSSL.

For instructions on creating certificates, see Section 19.9.5.

The different values for the sslmode parameter provide different levels of protection. SSL can provide protection against three types of attacks:

For a connection to be known SSL-secured, SSL usage must be configured on both the client and the server before the connection is made. If it is only configured on the server, the client may end up sending sensitive information (e.g., passwords) before it knows that the server requires high security. In libpq, secure connections can be ensured by setting the sslmode parameter to verify-full or verify-ca, and providing the system with a root certificate to verify against. This is analogous to using an https URL for encrypted web browsing.

Once the server has been authenticated, the client can pass sensitive data. This means that up until this point, the client does not need to know if certificates will be used for authentication, making it safe to specify that only in the server configuration.

All SSL options carry overhead in the form of encryption and key-exchange, so there is a trade-off that has to be made between performance and security. Table 34.1 illustrates the risks the different sslmode values protect against, and what statement they make about security and overhead.

The difference between verify-ca and verify-full depends on the policy of the root CA. If a public CA is used, verify-ca allows connections to a server that somebody else may have registered with the CA. In this case, verify-full should always be used. If a local CA is used, or even a self-signed certificate, using verify-ca often provides enough protection.

The default value for sslmode is prefer. As is shown in the table, this makes no sense from a security point of view, and it only promises performance overhead if possible. It is only provided as the default for backward compatibility, and is not recommended in secure deployments.

Table 34.2 summarizes the files that are relevant to the SSL setup on the client.

If your application initializes libssl and/or libcrypto libraries and libpq is built with SSL support, you should call PQinitOpenSSL to tell libpq that the libssl and/or libcrypto libraries have been initialized by your application, so that libpq will not also initialize those libraries.