openpgp-notes/book/source/05-private.md

<!--
SPDX-FileCopyrightText: 2023 The "Notes on OpenPGP" project
SPDX-License-Identifier: CC-BY-SA-4.0
-->

(private_key_chapter)=
# Private key material in OpenPGP

This chapter discusses the handling of private key material, in OpenPGP. Private key material is associated with component keys that are parts of [OpenPGP certificates](certificates_chapter). For a discussion of packet structure internals, see the chapter {ref}`zoom_private`

Historically, terminology around OpenPGP certificates and keys has often been used inconsistently. The pair of terms "OpenPGP public key" and "OpenPGP private/secret keys" were commonly used (while the shorthand "OpenPGP key" can refer to both, depending on context).

## Terms

In this document, we use the term *OpenPGP certificate* to refer "OpenPGP public keys": The combination of component public keys, identity components and bindings.

This chapter is about the counterpart to the public material in certificates: Here, we discuss the handling of *private key material* in OpenPGP.

In this text, we treat the private key material as logically separate from the OpenPGP certificate. Operations that use the private key material are typically handled by a separate subsystem. It is useful to think about OpenPGP certificates on one hand, and the associated private key material, on the other, as two related elements, which are usually handled separately[^pkcs11]:


```{figure} diag/OpenPGPCert_with_privatekeystore.png
:name: fig-openpgp-certificate-with-private-key-store
:alt: Depicts a diagram on white background with an OpenPGP Certificate and a private key store. Gray dotted lines connect the green public key symbols of the OpenPGP Certificate with red dotted private key symbols in the private key store.

An OpenPGP certificate, with the associated private key material handled by a key store subsystem.
```

[^pkcs11]: This kind of distinction between certificates (which combine public key material and identity information) on the one hand, and private key material on the other, is also applied in the data model of [PKCS #11](https://en.wikipedia.org/wiki/PKCS_11) cryptographic systems.

However, there is one exception. Raw cryptographic private key material is sometimes embedded in an OpenPGP framing format that additionally contains the certificate: [*Transferable secret keys (TSK)*](https://www.ietf.org/archive/id/draft-ietf-openpgp-crypto-refresh-11.html#name-transferable-secret-keys).

```{admonition} Terminology
:class: note

Transferable secret keys are sometimes colloquially referred to as "OpenPGP private keys".
```

## Transferable secret keys

Sometimes users handle OpenPGP certificates combined with private key material in the form of *transferable secret keys* (TSK). Transferable secret keys are a serialized format that combines the OpenPGP certificate data with the connected private key material, stored in a single file.

```{figure} diag/TSK.png
:name: fig-transferable-secret-key
:alt: Depicts a box on white background with the title "Transferable secret key". It is identical to the figure depicting an OpenPGP certificate, with the exception, that in each component key box, below the green public key symbol, also the red dotted private key symbol is shown.

OpenPGP certificate with integrated private key material, as a TSK
```

The TSK format can be useful for backups of OpenPGP key material, or to move a key to a different computer[^gpg-tsk].

[^gpg-tsk]: For example, with GnuPG, an OpenPGP key can be exported in (armored) TSK format like this: `gpg --export-secret-key --armor <fingerprint>`

See the chapter {ref}`zoom_private` for insights into the packet structure of a TSK.

(encrypted_secrets)=
## Protecting private key material with a passphrase (using S2K)

In OpenPGP format, private key material can be optionally protected with a [passphrase](https://en.wikipedia.org/wiki/Passphrase). This mechanism applies symmetric encryption to the private key data in component keys.

The symmetric encryption key is derived from a secret that the user knows (the passphrase).

Using a passphrase can be useful when a third party can obtain a copy of the OpenPGP key data, but doesn't know the passphrase. In this scenario, an attacker may have obtained a copy of an OpenPGP key, but is unable to use it, because the private key material is encrypted, and the attacker cannot decrypt it.

OpenPGP defines a mechanism called [string-to-key (S2K)](https://www.ietf.org/archive/id/draft-ietf-openpgp-crypto-refresh-11.html#name-string-to-key-s2k-specifier) that is used to derive (high-entropy) symmetric encryption keys from (lower-entropy) passphrases, using a [key derivation function (KDF)](https://en.wikipedia.org/wiki/Key_derivation_function).

```{figure} diag/passphrase_using_S2K.png
:name: fig-passphrase-using-s2k
:alt: Depicts a diagram on white background with the title "Converting a passphrase into a symmetric key". On the left hand side a box with dotted yellow frame and light yellow background and the text "correct horse battery staple" is shown. It is connected by a dotted yellow line with the word "Passphrase". Right of the passphrase an arrow with green dotted frame, light green background and the text "S2K mechanism (string-to-key)", pointing to the right is shown. On the right hand side the yellow symmetric key symbol is shown.

Transforming a passphrase into a symmetric key
```

The private key material for individual component keys of one certificate can be encrypted with different mechanisms, and/or using different passphrases.

However, this is not customarily done. Usually, when creating a certificate, the user's software will use the same encryption mechanism and passphrase for all component keys (to limit complexity).

However, when adding new subkeys to a certificate at a later date, the user might choose to use a different passphrase. Or the user's software may choose a different encryption mechanism, e.g. based on updated best practices.

### S2K mechanisms for symmetric key generation

Over time, OpenPGP has specified a series of [S2K mechanisms](https://www.ietf.org/archive/id/draft-ietf-openpgp-crypto-refresh-11.html#name-string-to-key-s2k-types-reg), following the current state of the art. Of the specified S2K mechanisms, two remain relevant today:

- [Iterated and Salted S2K](https://www.ietf.org/archive/id/draft-ietf-openpgp-crypto-refresh-11.html#name-iterated-and-salted-s2k), which OpenPGP version 4 implementations can handle
- [Argon2](https://www.ietf.org/archive/id/draft-ietf-openpgp-crypto-refresh-11.html#name-argon2), which was newly added in OpenPGP version 6, and additionally protects the passphrase against brute-force attacks because it is memory-hard (which reduces the efficiency of attacks with specialised hardware)

### Mechanisms for secret key encryption with S2K

Different mechanisms are specified [for the encryption of the secret key data](https://www.ietf.org/archive/id/draft-ietf-openpgp-crypto-refresh-12.html#name-secret-key-encryption).


## Private key operations

The core of private key operations doesn't require access to the whole certificate.

Designs of private key subsystems in the OpenPGP space differ:

- Some designs perform the primitive cryptographic operations in a separate backend, only using the cryptographic key material itself. This type of design matches well with general purpose hardware cryptographic devices (such as TPMs).
- In other designs, the OpenPGP private key subsystem requires the additional metadata that is stored in the component key (the key creation time, for all keys, and in the case of keys that use ECDH algorithms: the KDF parameters).

Either way: at most, a private key subsystem should need access to individual component keys and the associated private key material.

```{note}

Keeping a copy of the full certificate information in the private key subsystem can cause "split brain" problems. E.g.: expiration times could then differ between different subsystems, when they keep different copies of the certificate information.

(Thunderbird users currently suffer from such issues, I think)
```

### OpenPGP card for private keys

[OpenPGP card](https://en.wikipedia.org/wiki/OpenPGP_card) devices are a type of hardware security device. They are one popular way to handle OpenPGP private key material.

OpenPGP card devices implement an open specification: [Functional Specification of the OpenPGP application on ISO Smart Card Operating Systems, Version 3.4.1](https://gnupg.org/ftp/specs/OpenPGP-smart-card-application-3.4.1.pdf). Multiple vendors produce devices that implement this specification, and there are a number of Free Software implementations (some of which can even be run on open hardware designs).

Effectively, the OpenPGP card specification outlines one model for an OpenPGP private key subsystem:

OpenPGP card devices do not store the full OpenPGP certificate. Instead, they have three "key slots": one each for signing, decryption and authentication. Each key slots stores the component key data and the associated cryptographic private key material[^missing-ecdh]. In addition, for each key slot, the OpenPGP fingerprint for the contained component key is explicitly stored on the card (note that storing the fingerprint on the card is different from how OpenPGP key data is usually handled: normally, fingerprints are not explicitly stored, but calculated on the fly from the component key data).

[^missing-ecdh]: In the case of ECDH keys, the KDF parameters (hash function ID and a symmetric encryption algorithm ID) are not stored on the OpenPGP card. This is considered a flaw in the OpenPGP card specification. These missing parameters can be handled in two ways, by OpenPGP software running on the host computer: Either by consulting a copy of the component key (e.g. by inspecting a copy of the certificate), or by deducing the missing KDF parameters from the OpenPGP fingerprint that is stored on the card.

## Advanced topics

### TSKs: Best practices S2K + S2K migration?

### The KOpenPGP attack

See [https://www.kopenpgp.com/](https://www.kopenpgp.com/)