CryptDB – A database system that can process SQL queries over encrypted data

CryptDB is cool, but be very careful if you're thinking about actually using it for anything.

Muhammad Naveed, Seny Kamara, and I have a paper appearing at ACM CCS next month where we show that the approach used in CryptDB and other similar systems is vulnerable to a raft of statistical attacks.

http://research.microsoft.com/en-us/um/people/senyk/pubs/edb...

The problem happens when you have a sensitive field in the DB, and you actually need to run queries on that field for your app to function. CryptDB will "peel the onion back" so that your column is protected with a weaker deterministic or order-preserving encryption primitive in order to let the queries succeed. But then you're vulnerable to the statistical attacks that we describe in our paper.

Of course, the CryptDB authors are smart people, so they foresaw some of this. They describe functionality in their tool for marking certain columns as "sensitive", so CryptDB will never peel the onion back too far. But then you can't use that column for most queries.

Like I said, be very careful.

This is a fantastic project from the MIT CSAIL team that builds brings together many cryptographic components into a integrated system. The academic papers on this are available from here https://css.csail.mit.edu/cryptdb/

Conceptually it's a sql proxy that decides how to maximally encrypt data based on the operations that are required. It can then weaken and re-encrypt portions of the data on the fly so that more powerful operations can be performed.

You can chain the encryption to the users login to to prevent cross user leakage.

This model allows you to have a database that is protected from adversary who have access to the database.

The encryption techniques include a partially Homomorphic scheme and other non-homomorphic schemes.

I believe that the team behind it seem to have moved on to a startup using similar technology. The code is not being actively maintained anymore. I was able to build it successfully on Ubuntu 12.04 with Bison 2.x(The code requires a mysql build which chokes on BISON 3).

They claim only a ~15%-25% performance hit although I've not yet be able to replicate that myself yet.

Please note that the attack already shown in http://www.ijiss.org/ijiss/index.php/ijiss/article/view/58/p... in “Section 4 Security & Efficiency of CryptDB” in March 2014. It basically says that “First of all, CryptDB is open to frequency attack where the adversary knows the frequency of the plaintext. Namely, if the RND layer is decrypted to the DET layer in EQ onion, then the frequency attack is possible to apply because of deterministic encryption in the DET layer. In this attack, the adversary that observes the queries can determine the ciphertext simply by looking at the results’ row count. This attack can only be fixed by the RND layer, which has no usable functionality in practice. For example, assume that we have left part of Table 11, and its encrypted form is the right part in Table 11. By using the knowledge of the frequency, one can learn the corresponding plaintexts from the right encrypted part in Table 11. This issue can be solved easily by using random IV based symmetric encryption, however, this will prevent executing all queries.“ You only apply the real medical data to show its practicality.

Related:

http://research.microsoft.com/en-us/projects/cipherbase/

http://www.cs.hku.hk/research/techreps/document/TR-2014-03.p...

I'd like to see this built into PostgreSQL in a way where I can optionally enable it for certain tables only (e.g. accounts, user profiles, sessions).

Do you have any examples of this being used on a live website?