TLS Forward Secrecy in Postfix
TLS Forward Secrecy in Postfix
Forward secrecy does not protect against active attacks such as forged DNS replies or forged TLS server certificates. If such attacks are a concern, then the SMTP client will need to authenticate the remote SMTP server in a sufficiently-secure manner. For example, by the fingerprint of a (CA or leaf) public key or certificate. Conventional PKI relies on many trusted parties and is easily subverted by a state-funded adversary.
Postfix supports forward secrecy of TLS network communication since version 2.2. This support was adopted from Lutz Jänicke's "Postfix TLS patch" for earlier Postfix versions. This document will focus on TLS Forward Secrecy in the Postfix SMTP client and server. See TLS_README for a general description of Postfix TLS support.
Topics covered in this document:
Give me some background on forward secrecy in Postfix
Never mind, just show me what it takes to get forward secrecy
The term "Forward Secrecy" (or sometimes "Perfect Forward Secrecy") is used to describe security protocols in which the confidentiality of past traffic is not compromised when long-term keys used by either or both sides are later disclosed.
Forward secrecy is accomplished by negotiating session keys using per-session cryptographically-strong random numbers that are not saved, and signing the exchange with long-term authentication keys. Later disclosure of the long-term keys allows impersonation of the key holder from that point on, but not recovery of prior traffic, since with forward secrecy, the discarded random key agreement inputs are not available to the attacker.
Forward secrecy is only "perfect" when brute-force attacks on the key agreement algorithm are impractical even for the best-funded adversary and the random-number generators used by both parties are sufficiently strong. Otherwise, forward secrecy leaves the attacker with the challenge of cracking the key-agreement protocol, which is likely quite computationally intensive, but may be feasible for sessions of sufficiently high value. Thus forward secrecy places cost constraints on the efficacy of bulk surveillance, recovering all past traffic is generally infeasible, and even recovery of individual sessions may be infeasible given a sufficiently-strong key agreement method.
Early implementations of the SSL protocol do not provide forward secrecy (some provide it only with artificially-weakened "export" cipher suites, but we will ignore those here). The client sends a random "pre-master secret" to the server encrypted with the server's RSA public key. The server decrypts this with its private key, and uses it together with other data exchanged in the clear to generate the session key. An attacker with access to the server's private key can perform the same computation at any later time.
Later revisions to the TLS protocol introduced forward-secrecy cipher suites in which the client and server implement a key exchange protocol based on ephemeral secrets. Sessions encrypted with one of these newer cipher suites are not compromised by future disclosure of long-term authentication keys.
The key-exchange algorithms used for forward secrecy require the TLS server to designate appropriate "parameters" consisting of a mathematical "group" and an element of that group called a "generator". Presently, there are two flavors of "groups" that work with PFS:
FFDHE: Finite-field Diffie-Hellman ephemeral key exchange groups (also EDH or DHE). The server needs to be configured with a suitably-large prime and a corresponding "generator". Standard choices of the prime and generator are specified in RFC7919, and can be used in the TLS 1.3 protocol with the server and client negotiating a mutually supported choice. In earlier versions of TLS (1.0 through 1.2), when FFDHE key exchange is performed, the server chooses the prime and generator unilaterally.
EECDH: This is short for Ephemeral Elliptic Curve Diffie-Hellman (also abbreviated as ECDHE). EECDH offers better security at lower computational cost than FFDHE. Elliptic curves used in cryptography are typically identified by a "name" that stands for a set of well-known parameter values, and it is these "named curves" (or, in certificates, associated ASN.1 object identifiers) that are used in the TLS protocol. When EECDH key exchange is used, a mutually supported named curve is negotiated as part of the TLS handshake.
The Postfix ≥ 2.2 SMTP server supports forward secrecy in its default configuration. If the remote SMTP client prefers cipher suites with forward secrecy, then the traffic between the server and client will resist decryption even if the server's long-term authentication keys are later compromised.
Most remote SMTP clients now support forward secrecy (the only choice as of TLS 1.3), but some may prefer cipher suites without forward secrecy. Postfix ≥ 2.8 servers can be configured to override the client's preference by setting "tls_preempt_cipherlist = yes".
Postfix ≥ 3.1 supports 2048-bit-prime FFDHE out of the box, with no additional configuration. You can also generate your own FFDHE parameters, but this is not necessary and no longer recommended. See the quick-start section for details.
Postfix ≥ 3.8 supports the finite-field Diffie-Hellman ephemeral (FFDHE) key exchange group negotiation API of OpenSSL ≥ 3.0. FFDHE groups are explicitly negotiated between client and server starting with TLS 1.3. In earlier TLS versions, the server chooses the group unilaterally. The list of candidate FFDHE groups can be configured via "tls_ffdhe_auto_groups", which can be used to select a prioritized list of supported groups (most preferred first) on both the server and client. The default list is suitable for most users. Either, but not both of "tls_eecdh_auto_curves" and "tls_ffdhe_auto_groups" may be set empty, disabling either EC or FFDHE key exchange in OpenSSL 3.0 with TLS 1.3. That said, interoperability will be poor if the EC curves are all disabled or don't include the most widely used curves.
As of Postfix 3.2 and OpenSSL 1.0.2, a range of supported EECDH curves is enabled in the server and client, and a suitable mutually supported curve is negotiated as part of the TLS handshake. The list of supported curves is configurable via the "tls_eecdh_auto_curves" parameter. With TLS 1.2 the server needs to leave its setting of "smtpd_tls_eecdh_grade" at the default value of "auto" (earlier choices of an explicit single curve grade are deprecated). With TLS 1.3, the "smtpd_tls_eecdh_grade" parameter is not used, and curve selection is unconditionally negotiated.
The Postfix ≥ 2.2 SMTP client supports forward secrecy in its default configuration. All supported OpenSSL releases support both FFDHE and EECDH key exchange. If the remote SMTP server supports cipher suites with forward secrecy (and does not override the SMTP client's cipher preference), then the traffic between the server and client will resist decryption even if the server's long-term authentication keys are later compromised. Forward secrecy is always on in TLS 1.3.
Postfix ≥ 3.2 supports the curve negotiation API of OpenSSL ≥ 1.0.2. The list of candidate curves can be changed via the "tls_eecdh_auto_curves" configuration parameter, which can be used to select a prioritized list of supported curves (most preferred first) on both the Postfix SMTP server and SMTP client. The default list is suitable for most users.
Postfix ≥ 3.8 supports the finite-field Diffie-Hellman ephemeral (FFDHE) key exchange group negotiation API of OpenSSL ≥ 3.0. The list of candidate FFDHE groups can be configured via "tls_ffdhe_auto_groups", which can be used to select a prioritized list of supported groups (most preferred first) on both the server and client. The default list is suitable for most users.
The default Postfix SMTP client cipher lists are correctly ordered to prefer EECDH and FFDHE cipher suites ahead of similar cipher suites that don't implement forward secrecy. Administrators are strongly discouraged from changing the cipher list definitions.
This works "out of the box" with no need for additional configuration.
Postfix ≥ 3.2 supports the curve negotiation API of OpenSSL ≥ 1.0.2. The list of candidate curves can be changed via the "tls_eecdh_auto_curves" configuration parameter, which can be used to select a prioritized list of supported curves (most preferred first) on both the Postfix SMTP server and SMTP client. The default list is suitable for most users.
This works "out of the box" with no need for additional configuration.
Postfix ≥ 3.2 supports the curve negotiation API of OpenSSL ≥ 1.0.2. The list of candidate curves can be changed via the "tls_eecdh_auto_curves" configuration parameter, which can be used to select a prioritized list of supported curves (most preferred first) on both the Postfix SMTP server and SMTP client. The default list is suitable for most users.
In Postfix < 3.8, or OpenSSL prior to 3.0, FFDHE for TLS 1.2 or below works "out of the box", no additional configuration is necessary. The most one can do is (not advisable) disable all "kDHE" ciphers, which would then disable FFDHE key exchange in TLS 1.2 and below.
With OpenSSL 1.1.1, FFDHE is not supported for TLS 1.3, which uses only EECDH key exchange. Support for FFDHE with TLS 1.3 was added in OpenSSL 3.0. With OpenSSL 3.0 and Postfix 3.8 the list of supported TLS 1.3 FFDHE groups becomes configurable via the "tls_ffdhe_auto_groups" parameter, which can be set empty to disable FFDHE in TLS 1.3, or conversely expanded to support more groups. The default should work well for most users.
In Postfix < 3.8, or OpenSSL prior to 3.0, FFDHE for TLS 1.2 or below works "out of the box", no additional configuration is necessary. One can of course (not advisable) disable all "kDHE" ciphers, which would then disable FFDHE key exchange in TLS 1.2 and below.
The built-in default Postfix FFDHE group is a 2048-bit group as of Postfix 3.1. You can optionally generate non-default Postfix SMTP server FFDHE parameters for possibly improved security against pre-computation attacks, but this is not necessary or recommended. Just leave "smtpd_tls_dh1024_param_file" at its default empty value.
The set of FFDHE groups enabled for use with TLS 1.3 becomes configurable with Postfix ≥ 3.8 and OpenSSL ≥ 3.0. The default setting of "tls_ffdhe_auto_groups" enables the RFC7919 2048 and 3072-bit groups. If you need more security, you should probably be using EECDH.
Postfix can be configured to report information about the negotiated cipher, the corresponding key lengths, and the remote peer certificate or public-key verification status.
With "smtp_tls_loglevel = 1" and "smtpd_tls_loglevel = 1", the Postfix SMTP client and server will log TLS connection information to the maillog file. The general logfile format is shown below. With TLS 1.3 there may be additional properties logged after the cipher name and bits.
postfix/smtp[process-id]: Untrusted TLS connection established to host.example.com[192.168.0.2]:25: TLSv1 with cipher cipher-name (actual-key-size/raw-key-size bits) postfix/smtpd[process-id]: Anonymous TLS connection established from host.example.com[192.168.0.2]: TLSv1 with cipher cipher-name (actual-key-size/raw-key-size bits)
With "smtpd_tls_received_header = yes", the Postfix SMTP server will record TLS connection information in the Received: header in the form of comments (text inside parentheses). The general format depends on the smtpd_tls_ask_ccert setting. With TLS 1.3 there may be additional properties logged after the cipher name and bits.
Received: from host.example.com (host.example.com [192.168.0.2])
(using TLSv1 with cipher cipher-name
(actual-key-size/raw-key-size bits))
(Client CN "host.example.com", Issuer "John Doe" (not verified))
Received: from host.example.com (host.example.com [192.168.0.2])
(using TLSv1 with cipher cipher-name
(actual-key-size/raw-key-size bits))
(No client certificate requested)
TLS 1.3 examples. Some of the new attributes may not appear when not applicable or not available in older versions of the OpenSSL library.
Received: from localhost (localhost [127.0.0.1])
(using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits)
key-exchange X25519 server-signature RSA-PSS (2048 bits) server-digest SHA256)
(No client certificate requested)
Received: from localhost (localhost [127.0.0.1])
(using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits)
key-exchange X25519 server-signature RSA-PSS (2048 bits) server-digest SHA256
client-signature ECDSA (P-256) client-digest SHA256)
(Client CN "example.org", Issuer "example.org" (not verified))
The "key-exchange" attribute records the type of "Diffie-Hellman" group used for key agreement. Possible values include "DHE", "ECDHE", "X25519" and "X448". With "DHE", the bit size of the prime will be reported in parentheses after the algorithm name, with "ECDHE", the curve name.
The "server-signature" attribute shows the public key signature algorithm used by the server. With "RSA-PSS", the bit size of the modulus will be reported in parentheses. With "ECDSA", the curve name. If, for example, the server has both an RSA and an ECDSA private key and certificate, it will be possible to track which one was used for a given connection.
The new "server-digest" attribute records the digest algorithm used by the server to prepare handshake messages for signing. The Ed25519 and Ed448 signature algorithms do not make use of such a digest, so no "server-digest" will be shown for these signature algorithms.
When a client certificate is requested with "smtpd_tls_ask_ccert" and the client uses a TLS client-certificate, the "client-signature" and "client-digest" attributes will record the corresponding properties of the client's TLS handshake signature.
The next sections will explain what cipher-name, key-size, and peer verification status information to expect.
There are dozens of ciphers that support forward secrecy. What follows is the beginning of a list of 51 ciphers available with OpenSSL 1.0.1e. The list is sorted in the default Postfix preference order. It excludes null ciphers that only authenticate and don't encrypt, together with export and low-grade ciphers whose encryption is too weak to offer meaningful secrecy. The first column shows the cipher name, and the second shows the key exchange method.
$ openssl ciphers -v \
'aNULL:-aNULL:kEECDH:kEDH:+RC4:!eNULL:!EXPORT:!LOW:@STRENGTH' |
awk '{printf "%-32s %s\n", $1, $3}'
AECDH-AES256-SHA Kx=ECDH
ECDHE-RSA-AES256-GCM-SHA384 Kx=ECDH
ECDHE-ECDSA-AES256-GCM-SHA384 Kx=ECDH
ECDHE-RSA-AES256-SHA384 Kx=ECDH
ECDHE-ECDSA-AES256-SHA384 Kx=ECDH
ECDHE-RSA-AES256-SHA Kx=ECDH
ECDHE-ECDSA-AES256-SHA Kx=ECDH
ADH-AES256-GCM-SHA384 Kx=DH
ADH-AES256-SHA256 Kx=DH
ADH-AES256-SHA Kx=DH
ADH-CAMELLIA256-SHA Kx=DH
DHE-DSS-AES256-GCM-SHA384 Kx=DH
DHE-RSA-AES256-GCM-SHA384 Kx=DH
DHE-RSA-AES256-SHA256 Kx=DH
...
To date, all ciphers that support forward secrecy have one of five values for the first component of their OpenSSL name: "AECDH", "ECDHE", "ADH", "EDH" or "DHE". Ciphers that don't implement forward secrecy have names that don't start with one of these prefixes. This pattern is likely to persist until some new key-exchange mechanism is invented that also supports forward secrecy.
The actual key length and raw algorithm key length are generally the same with non-export ciphers, but they may differ for the legacy export ciphers where the actual key is artificially shortened.
Starting with TLS 1.3 the cipher name no longer contains enough information to determine which forward-secrecy scheme was employed, but TLS 1.3 always uses forward-secrecy. On the client side, up-to-date Postfix releases log additional information for TLS 1.3 connections, reporting the signature and key exchange algorithms. Two examples below (the long single line messages are folded across multiple lines for readability):
postfix/smtp[process-id]: Untrusted TLS connection established to 127.0.0.1[127.0.0.1]:25: TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) key-exchange X25519 server-signature RSA-PSS (2048 bits) server-digest SHA256 client-signature ECDSA (P-256) client-digest SHA256 postfix/smtp[process-id]: Untrusted TLS connection established to 127.0.0.1[127.0.0.1]:25: TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) key-exchange ECDHE (P-256) server-signature ECDSA (P-256) server-digest SHA256
In the above connections, the "key-exchange" value records the "Diffie-Hellman" algorithm used for key agreement. The "server-signature" value records the public key algorithm used by the server to sign the key exchange. The "server-digest" value records any hash algorithm used to prepare the data for signing. With "ED25519" and "ED448", no separate hash algorithm is used.
Examples of Postfix SMTP server logging:
postfix/smtpd[process-id]: Untrusted TLS connection established from localhost[127.0.0.1]:25: TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) key-exchange X25519 server-signature RSA-PSS (2048 bits) server-digest SHA256 client-signature ECDSA (P-256) client-digest SHA256 postfix/smtpd[process-id]: Anonymous TLS connection established from localhost[127.0.0.1]: TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) server-signature RSA-PSS (2048 bits) server-digest SHA256 postfix/smtpd[process-id]: Anonymous TLS connection established from localhost[127.0.0.1]: TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) server-signature ED25519
Note that Postfix ≥ 3.4 server logging may also include a "to sni-name" element to record the use of an alternate server certificate chain for the connection in question. This happens when the client uses the TLS SNI extension, and the server selects a non-default certificate chain based on the client's SNI value:
postfix/smtpd[process-id]: Untrusted TLS connection established from client.example[192.0.2.1] to server.example: TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) key-exchange X25519 server-signature RSA-PSS (2048 bits) server-digest SHA256 client-signature ECDSA (P-256) client-digest SHA256
The verification levels below are subject to man-in-the-middle attacks to different degrees. If such attacks are a concern, then the SMTP client will need to authenticate the remote SMTP server in a sufficiently-secure manner. For example, by the fingerprint of a (CA or leaf) public key or certificate. Remember that conventional PKI relies on many trusted parties and is easily subverted by a state-funded adversary.
Postfix SMTP client: With opportunistic TLS (the "may" security level) the Postfix SMTP client does not verify any information in the peer certificate. In this case it enables and prefers anonymous cipher suites in which the remote SMTP server does not present a certificate (these ciphers offer forward secrecy of necessity). When the remote SMTP server also supports anonymous TLS, and agrees to such a cipher suite, the verification status will be logged as "Anonymous".
Postfix SMTP server: This is by far most common, as client certificates are optional, and the Postfix SMTP server does not request client certificates by default (see smtpd_tls_ask_ccert). Even when client certificates are requested, the remote SMTP client might not send a certificate. Unlike the Postfix SMTP client, the Postfix SMTP server "anonymous" verification status does not imply that the cipher suite is anonymous, which corresponds to the server not sending a certificate.
Postfix SMTP client: The remote SMTP server presented a certificate, but the Postfix SMTP client was unable to check the issuing CA signature. With opportunistic TLS this is common with remote SMTP servers that don't support anonymous cipher suites.
Postfix SMTP server: The remote SMTP client presented a certificate, but the Postfix SMTP server was unable to check the issuing CA signature. This can happen when the server is configured to request client certificates (see smtpd_tls_ask_ccert).
Postfix SMTP client: The remote SMTP server's certificate was signed by a CA that the Postfix SMTP client trusts, but either the client was not configured to verify the destination server name against the certificate, or the server certificate did not contain any matching names. This is common with opportunistic TLS (smtp_tls_security_level is "may" or else "dane" with no usable TLSA DNS records) when the Postfix SMTP client's trusted CAs can verify the authenticity of the remote SMTP server's certificate, but the client is not configured or unable to verify the server name.
Postfix SMTP server: The remote SMTP client certificate was signed by a CA that the Postfix SMTP server trusts. The Postfix SMTP server never verifies the remote SMTP client name against the names in the client certificate. Since the client chooses to connect to the server, the Postfix SMTP server has no expectation of a particular client hostname.
Postfix SMTP client: The remote SMTP server's certificate was signed by a CA that the Postfix SMTP client trusts, and the certificate name matches the destination or server name(s). The Postfix SMTP client was configured to require a verified name, otherwise the verification status would have been just "Trusted".
Postfix SMTP client: The "Verified" status may also mean that the Postfix SMTP client successfully matched the expected fingerprint against the remote SMTP server public key or certificate. The expected fingerprint may come from smtp_tls_policy_maps or from TLSA (secure) DNS records. The Postfix SMTP client ignores the CA signature.
Postfix SMTP server: The status is never "Verified", because the Postfix SMTP server never verifies the remote SMTP client name against the names in the client certificate, and because the Postfix SMTP server does not expect a specific fingerprint in the client public key or certificate.