OPAQUE-Store is a simple protocol that allows anyone to store encrypted blobs of information online, with only a password needed to retrieve the information. As the name implies it uses the OPAQUE protocol to do so. OPAQUE-Store uses the export_key feature of OPAQUE to encrypt the data that is stored on the OPAQUE-Storage server.

You might want to read this blog-post on this topic and on more info: https://www.ctrlc.hu/~stef/blog/posts/How_to_recover_static_secrets_using_OPAQUE.html

OPAQUE-Store goes beyond the original OPAQUE protocol as specified by the IRTF/CFRG and also supports a threshold variant of OPAQUE. In a threshold setup you have a number N of servers that all hold a share of your secret and at least a threshold number T of these need to cooperate to recover the secret. This provides extra robustness and dillution of responsibility (losing a server is not the end of the world!) while at the same time increases security, as an attacker now has to compromise at least T servers to get access to some information.

opaquestore depends on https://github.com/stef/libopaque/ which in turn depends on

• libsodium,
• https://github.com/stef/pysodium available on pypi,
• https://github.com/stef/liboprf, and
• pyoprf (part of https://github.com/stef/liboprf) available on pypi.

When you have a working libopaque, a simple pip install opaquestore should get you started.

Configuration will be looked for in the following order

• /etc/opaque-store/config
• ~/.config/opaque-store/config
• ~/.opaque-storerc
• ./opaque-store.cfg

For an example and documentation on the values in the config files see: opaque-store.cfg for the client configuration, and - in case you want to run your own server(s) - opaque-stored.cfg for the server configuration.

Example configuration with inline comments about each value:

[client]
# you must change this value, it ensures that your record ids are
# unique you must also make sure to not lose this value, if you do,
# you lose access to your records.
id_salt="Please_MUST-be_changed! and backed up to something difficult to guess"
# the number of servers successfully participating in an
# operation. must be less than 129, but lower 1 digit number are
# probable the most robust.
threshold=2
# the time in seconds a distributed keygen (DKG) protocol message is
# considered fresh. anything older than this is considered invalid and
# aborts a DKG. Higher values help with laggy links, lower values can
# be fine if you have high-speed connections to all servers.
ts_epsilon=1200

# the list of servers, must be 1 item, if threshold is 1, or one more
# than threshold.
[servers]
[servers.zero]
# address of server
host="127.0.0.1"
# port where server is running
port=23000
# self-signed public key of the server
# - not needed for proper Lets Encrypt certs
ssl_cert = "/etc/opaquestore/zero/cert.pem"
ltsigkey="/etc/opaquestore/zero/zero.pub"

[servers.eins]
# address of server
host="127.0.0.1"
# port where server is running
port=23001
# public key of the server
ltsigkey="/etc/opaquestore/eins/eins.pub"

[servers.dva]
# address of server
host="127.0.0.1"
# port where server is running
port=23002
# public key of the server
ltsigkey="/etc/opaquestore/dva/dva.pub"

The client config file, contains a [servers] section which lists all servers you want to use in a threshold setup. Each server has an address, port and ltsigkey variable that needs to be set accordingly. In case your server runs with a self-signed certificate there is a ssl_cert variable that can pin it to the correct cert. It is also important to note, that the name of the server - which is given after a dot in the [servers.name] sub-section title is also used to generate record ids specific to that server. Thus once chosen, it should not change, unless you want to lose access to the records on that server. The name doesn't have to be unique by users, but should be unique among all configured servers in this setup, this guarantees that for a record each server has a different record it and thus makes the records unlinkable across servers.

In the config files [client] section the threshold variable specifies the threshold for the setup.

The minimum sane configuration for a threshold setup is threshold=2 with at least 3 servers listed. The maximum of servers is 128, but that is way too many, a reasonable max is around 16 or so.

It is warmly recommended to use pwdsphinx (https://github.com/stef/pwdsphinx) as a front-end to opaquestore, since it handles passwords in a most secure manner. If you want to use a different password manager, you can use the CLI interface documented below.

opaquestore takes the password always on the standard input. If you are creating or updating a record, the record itself is also expected on the standard input. The password and the record - if required - are separated by a newline character.

KeyIds are the identifiers that you use to address your records, they can be any kind of string. Internally this keyId is hashed using the id_salt from the configurations [client] section into a unique identifier. It is very warmly recommended to set this to some random value, and to back this value up. As this salt is necessary to access your records. If you use a commonly used salt (i.e. the default salt) chances are high that there are collisions for record ids, and that people can guess your record ids.

Storing a record needs 3 parameters:

• the password, on standard input, terminated by a newline.
• the record itself until the end of the standard input
• and a keyId with which you can reference and act on this record

$ echo -en 'password\ntoken2store' | opaquestore create <keyId>

Here is a contrived example:

echo -en "mypassword\!sMyV0ice\nmy secretty token data that i need to protect and store using opaque" | opaquestore create myfirstblob

In this example:

• the password is "mypassword!sMyV0ice"
• the record is: "my secretty token data that i need to protect and store using opaque"
• and the keyId is "myfirstblob"

Retrieving a record has to parameters:

• the password on standard input
• the keyId as the 2nd parameter to opaquestore

$ echo -n 'password' | opaquestore get <keyId>

An example fetching the record created in the previous example:

echo -en "mypassword\!sMyV0ice" | opaquestore get myfirstblob

It is possible to update a record in place, it is essentially the same as the creation of a record. It is important to note, that this operation only succeeds, if all servers need to process this request, not only those needed for matching the threshold, you want to update the record on all servers not just some.

$ echo -en 'password\ntoken2update' | opaquestore update <keyId>

If you do not care if some servers will not be updated and you really know what you are doing, you can use the alternative command force-update, in this case the operation will succeed if at least the threshold is matched. Note however if any of the servers that did not participate in the forced update will participate in later operations will corrupt later operations, so you might want to remove those servers from your config.

$ echo -en 'password\ntoken2update' | opaquestore force-update <keyId>

Deleting a record is very straight forward, you need your password and keyId, and ensure that all servers that store this record will all be available. The operation will fail if some servers are not available.

$ echo -n 'password' | opaquestore delete <keyId>

Similarly to the update operation there is also a forced delete operation, which will succeed if at least the threshold is matched. Servers not available during this forced delete will still hold the record, if your setup has a n-out-of-2*n setup could mean that you still have enough shares even after a forced-delete.

$ echo -n 'password' | opaquestore force-delete <keyId>

An attacker might be trying different passwords for your record, after a certain amount of consecutive password failures (by default 3) the server locks down the record. A locked record can only be unlocked with a recovery-token. It is not possible to ask for recovery-tokens when a record is already locked.

$ echo -n 'password' | opaquestore recovery-tokens <keyId>

If a record is locked, and you have a valid recovery-token you can reset the failure counter:

$ echo -n <recovery-token> | opaquestore unlock <keyId>

If you run server, you need to generate some long-term signing keys if you want to use this server in a threshold setup. If you don't provide the path to the keys, the secret-key will be taken from the ltsigkey config value in your opaque-storaged configuration, and the public-key will be the same as the secret-key, but with a .pub extension.

$ opaquestore genltsigkey [secret-key path] [public-key path]