This is a pre-production deployment of Warehouse. Changes made here affect the production instance of PyPI (
Help us improve Python packaging - Donate today!

Secure Content Store

Project Description
seccs --- the SECure Content Store

.. image::

.. image::

.. image::
:alt: Documentation Status

What it is

`seccs` is a Python library that realizes a secure and efficient hash-table-like
data structure for contents on top of any existing key-value store as provided
by, e.g., cloud storage providers.

It has been developed as part of the work [LS16]_ at CISPA, Saarland University.



$ pip install seccs

If you want to use AES-SIV encryption (you probably want!), you also need to install PyCrypto 2.7a1 which is not yet available in PyPI::

$ pip install

If you are using Python >=3.5, pip might fail to build the package `fastchunking`
on which `seccs` depends. This is due to the latest `pybindgen` release on PyPI
being incompatible to Python 3.5. To resolve this issue, install the latest
`pybindgen` version from github before installing `seccs`:


$ pip uninstall pybindgen
$ pip install git+

Usage and Overview

`seccs` is a Python implementation of `sec-cs`, a secure and efficient
hash-table-like data structure for contents. It stores its data on top of any
existing database providing a key-value store interface. Thus, it is likewise
usable with in-memory :code:`dict` objects, persistent databases like
:code:`ZODB`, and many cloud storage providers.

Its details are described in [LS16]_. In short, it is suitable for usage on
`untrusted` cloud storage and has the following desirable properties:

* Confidentiality:
Stored contents are securely encrypted using a symmetric key.
* Authenticity:
`sec-cs` guarantees authenticity of all stored contents,
irrespective of gurantees of the underlying database.
* Storage Efficiency:
Data deduplication strategies are applied to all stored contents. When
storing new contents, overlapping parts of existing contents are
automatically reused as to avoid redundancy.
`sec-cs` is optimized for efficiency in presence of `many` similar
contents: Storage costs of an n-bytes content that differs only slightly
from an existing content are in O(log n).

Typical Use Case

In the most-typical configuration, `sec-cs` chunks its contents hierarchically
using ML-CDC (see [LS16]_), usually relying on Rabin Karp hashes, and stores the
resulting nodes in a `database` after applying AES-SIV-256 for encryption and
authentication. From a user perspective, we have to initialize a suitable
database object and a 32-bytes key first.

Database and key setup:
>>> database = dict()
>>> import os
>>> key = os.urandom(32)

Note that we might want to store the database and the key at some persistent
location in practice.

Next, we need to create a `crypto wrapper` which is in charge of all the
cryptographic operations. Depending on our security goals (e.g., whether
encryption is required), we could choose any suitable wrapper from
:code:`seccs.crypto_wrapper`. Afterwards, we can instantiate the data structure.

Choice of `crypto wrapper` and instantiation of data structure:
>>> import seccs
>>> crypto_wrapper = seccs.crypto_wrapper.AES_SIV_256(key) # install PyCrypto>=2.7a1 to use AES-SIV
>>> seccs = seccs.SecCSLite(256, database, crypto_wrapper) # 256 is the chunk size

.. note::

Internally, `sec-cs` splits contents into chunks, creates a tree of chunks
for each of them and inserts each node separately into the `database`. The
first parameter specifies the desired `average` size of nodes inserted into
the database. As deduplication is performed at the chunk level, large chunk
sizes decrease deduplication performance, but they also create less storage
overhead when storing non-deduplicable contents as fewer nodes have to be

Performance is discussed in detail in [LS16]_. If high redundancy is
expected, 256 bytes is typically a good compromise; otherwise, larger chunk
sizes might be more suitable.

We can now insert contents...
>>> content = b"This is a test content."
>>> digest = seccs.put_content(content)
>>> repr(digest)

...retrieve them...
>>> seccs.get_content(digest)
This is a test content.

...and delete them as soon as they are not needed anymore:
>>> seccs.delete_content(digest)

Storage Efficiency

`seccs` avoids redundancy in the `database` wherever possible, as gets clear
in the following example.

Consider this function for measuring the `database`'s current storage costs in bytes:
>>> import sys
>>> def dbsize(db):
>>> return sum([sys.getsizeof(k) + sys.getsizeof(v) for (k, v) in db.items()])

Initially, the database is empty:
>>> dbsize(database)

Insertion of a 1 MiB content clearly causes some storage costs:
>>> content1 = os.urandom(1024*1024)
>>> digest1 = seccs.put_content(content1)
>>> dbsize(database)

But inserting the same content for a second time does not incur additional costs:
>>> content2 = content1
>>> digest2 = seccs.put_content(content2)
>>> digest1 == digest2 # identical contents yield identical digests
>>> dbsize(database)

Clearly, the database grows if different contents are inserted. However, these
costs are low if inserted contents are similar to existing ones.

Only about 2.3 KiB are required to store another 1 MiB content with one byte changed:
>>> content3 = b''.join([content1[:512*1024], b'x', content1[512*1024+1:]])
>>> digest3 = seccs.put_content(content3)
>>> dbsize(database)

Costs are similar even if the identical parts are shifted...
>>> content4 = b''.join([content1[:512*1024], b'xyz', content1[512*1024+1:]])
>>> digest4 = seccs.put_content(content4)
>>> dbsize(database)

...and deduplication is also performed if a content consists of parts of different existing contents:
>>> content5 = b''.join([content1, content3, content4])
>>> digest5 = seccs.put_content(content5)
>>> dbsize(database)

In the last example, the growth was about 3 KiB.

Furthermore, storage space is reclaimed completely when contents are removed:
>>> seccs.delete_content(digest5)
>>> seccs.delete_content(digest4)
>>> seccs.delete_content(digest3)
>>> seccs.delete_content(digest2)
>>> dbsize(database)
>>> seccs.delete_content(digest1)
>>> dbsize(database)

.. note::

Every :code:`seccs.delete_content` call undos eactly one
:code:`seccs.put_content` call. Thus, even if the same content has been
inserted twice, yielding only a single digest, it has to be deleted twice as
well to get actually removed.


`seccs` uses tox for testing, so simply run:


$ tox

.. [LS16] Dominik Leibenger and Christoph Sorge (2016). sec-cs: Getting the
Most out of Untrusted Cloud Storage.
`arXiv:1606.03368 <>`_
Release History

Release History

This version
History Node


History Node


History Node


History Node


Download Files

Download Files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

File Name & Checksum SHA256 Checksum Help Version File Type Upload Date
seccs-0.0.4.tar.gz (20.4 kB) Copy SHA256 Checksum SHA256 Source Apr 5, 2017

Supported By

WebFaction WebFaction Technical Writing Elastic Elastic Search Pingdom Pingdom Monitoring Dyn Dyn DNS Sentry Sentry Error Logging CloudAMQP CloudAMQP RabbitMQ Heroku Heroku PaaS Kabu Creative Kabu Creative UX & Design Fastly Fastly CDN DigiCert DigiCert EV Certificate Rackspace Rackspace Cloud Servers DreamHost DreamHost Log Hosting