System and architecture for supporting analytics on encrypted databases
A method is provided for generating an encrypted database. The method includes: receiving a plaintext database having plaintext data entries in one or more columns; augmenting the received plaintext database to generate an augmented plaintext database, the augmenting including the addition of one or more columns to the received plaintext database, each added column corresponding to an attribute which is to be made available for conditional queries; and encrypting the augmented plaintext database to generate the encrypted database including encrypted data entries. The encrypted database supports at least one form of conditional query for those attributes corresponding to the added columns, the at least one form of conditional query being computed on the encrypted data entries without the decryption thereof to produce an encrypted result.
1. A method of generating an encrypted database, said method comprising:
(a) receiving a plaintext database having plaintext data entries in one or more columns;
(b) augmenting the received plaintext database to generate an augmented plaintext database, said augmenting including an addition of one or more columns to the received plaintext database, each said added column corresponding to an attribute which is to be made available for conditional queries; and
(c) encrypting the augmented plaintext database to generate the encrypted database including encrypted data entries;
wherein the encrypted database supports at least one form of conditional query for those attributes corresponding to said added columns, the at least one form of conditional query being computed on the encrypted data entries without the decryption thereof to produce an encrypted result.
2. The method of claim 1 , wherein the encrypted data is encrypted with a semantically secure encryption.
3. The method of claim 2 , wherein the encrypted data is encrypted using a homomorphic cryptosystem.
4. The method of claim 3 , wherein the homomorphic cryptosystem is an additively homomorphic cryptosystem.
5. The method of claim 3 , wherein the homomorphic cryptosystem supports 2-DNF (Disjunctive Normal Form) operations.
6. The method of claim 1 , wherein the at least one form of conditional query is one of a WHERE query or a GROUPBY query.
7. The method of claim 1 , further comprising:
(d) establishing a public-private key pair, wherein a public key of said pair is used to encrypt the encrypted database and a private key of said pair is provided to decrypt an encrypted result of a query.
8. The method of claim 7 , wherein said private key is not made available to a database management system in which the encrypted database is maintained or a server which computes queries conducted on the encrypted database.
9. A system for administering an encrypted database, said system comprising:
an Extract, Transfer and Load (ETL) server, implemented via hardware, wherein said ETL server is operative to (i) receive as input a plaintext database having unencrypted data entries therein in one or more columns, (ii) augment the received plaintext database to generate an augmented plaintext database, wherein said augmented plaintext database includes an addition of one or more columns to the input plaintext database, each said added column corresponding to an attribute which is to be made available for conditional queries, and (iii) encrypt said augmented plaintext database to generate the encrypted database including encrypted data entries;
a database (DB) server, implemented via hardware, wherein the DB server is operative to (i) receive and maintain the encrypted database from said ETL server, and (ii) return encrypted data in response to queries submitted to the DB server; and
a computation server, implemented via hardware, wherein the computation server is operative to (i) submit queries to the DB server, and (ii) conduct computations on encrypted data returned from the DB server;
wherein said computations are conducted on the encrypted data from the encrypted database without decryption of the encrypted data, and results obtained from said computations are encrypted, the encrypted database being configured so as to support obtaining correct encrypted results in response to at least one form of conditional query without revealing samples of the unencrypted data underlying the encrypted data.
10. The system of claim 9 , wherein the encrypted data is encrypted with a semantically secure encryption.
11. The system of claim 10 , wherein the encrypted data is encrypted using a homomorphic cryptosystem.
12. The system of claim 11 , wherein the homomorphic cryptosystem is an additively homomorphic cryptosystem.
13. The system of claim 11 , wherein the homomorphic cryptosystem supports 2-DNF (Disjunctive Normal Form) operations.
14. The system of claim 9 , wherein the at least one form of conditional query is one of a WHERE query or a GROUPBY query.
15. The system of claim 9 , further comprising:
a key authority which administers a public-private key pair, wherein a public key of said pair is used to encrypt the encrypted database and a private key of said pair is usable to decrypt the encrypted result of a query.
16. The system of claim 15 , wherein said private key is not made available to the DB server or the computation server.