keyword search on encrypted data

Keyword search on encrypted data

Keyword search problem Linux utility: grep Information retrieval

Basic operation Advanced operations – relevance

analysis and ranking Search engines highly complicated problem

New settings Search data in the cloud Filter encrypted emails Privacy preserving log retrieval

Basic techniques

Symmetric encryptionPublic key encryption

Simple keyword matchingA little bit relevance evaluation

Secure keyword search with symmetric encryption Paper: Song 2000

•Seed is random, different for each Wi•Key idea: Li and Ri are self-verifiable •Advantage of XOR

How to set K?

Setting of ki Ki = Fk’(Wi), k’ is secret User publishes W and k = Fk’(W) Server checks CiW

whether <Li, Fk(Li)> == CiW It reveals nothing if Ci is not the ciphertext

for W. And Li is random for different Wi – server

cannot find any information from Li.

Hidden search In previous schemes, W is revealed

Weakness: each search will have to release k for W Easy to collect information

Solution: encrypt Wi with an private key, then xor with <Li, Fk(Li)>

Still weaknesses Wi encryption should be deterministic Access pattern is leaked Linear scan over the whole doc collection

Typical method for speedy keyword based search Using the “inverted index”

Word -> doc1:pos, doc2:pos,…

Or simply word -> doc1, doc2, …

However, inverted index reveals the word frequency

Recent developments Reza 2006

“Searchable symmetric encryption: improved definitions and efficient constructions”

Completely solved this problem, with a solution indistinguishability under chosen ciphertext attack (IND-CCA) Allow inverted index Hide word frequency

setup D – the set of documents {D1,…,Dn} max - the maximum number of

distinct words in a document Li – the list of document IDs that

contain the keyword w_i , plus some dummy entries to reach max

A – array contains all elements in Li (max * |D|)

T – table that contains the <wi, address of Li’s first node>)

Symmetric encryption function, encrypt words and document ids id(Dj) for wi entry is encoded as enc(wi||

j) to make indistinguishable

Pseudo-random function f Two pseudo-random permutation

functions : for mapping word to table entry : for mapping index to next node of Li

to the index of array A

Building the index table T

The key used to encryptthe node Ni,1

1.

2.

to random values of the same size of the existing entries

Generating Li

with Ki,0, We can decryptall nodes in the list

For the remaining max – |D(wi)| dummy nodes, store the doc id thatAlready appears in the first |D(wi)| entries. This can be done with the help of a look-up table I

Search Generate the trapdoor

Search

Property Each keyword search returns the

same number of encrypted document ids – the attacker cannot distinguish word frequency

Search public-key encrypted data Users who encrypt the data (with

public key) can be different from the owner of the private key

Cyclic group For example, if G =

{ g0, g1, g2, g3, g4, g5 } mod p is a group, then g6 = g0, and G is cyclic. p is the order g is the generator

Bilinear-map construction Two groups G1 G2 of prime order p A bilinear map : G1 X G1 -> G2 Properties: