190 likes | 353 Views
Efficient Search on Encrypted Data. Outline. SWP Linear Scan SWP encrypted index Goh Bloom Filter Hybird scheme Discussion. SWP Linear Scan. K i = f k’ (L i ) X i = E k’’ (W i ). SWP encrypted index. motivation :當文件量大時, sequential scan 不是一個有效率的方法。 ex :在大型資料庫上,用
E N D
Outline • SWP Linear Scan • SWP encrypted index • Goh Bloom Filter • Hybird scheme • Discussion
SWP Linear Scan • Ki = fk’(Li) • Xi = Ek’’(Wi)
SWP encrypted index • motivation:當文件量大時,sequential scan不是一個有效率的方法。 • ex:在大型資料庫上,用 pre-computed index是加速搜尋最常用的技巧。 • ex:index keyword pointer document
Simple way • Simple way: • Operation: • B search E(W) ,return A encrypted list. • A may decrypt the encrypted entries and send B another request to retrieve the documents • advantage:the request could be embedded in other retrievals • disadvantage:A has to spend round-trip time
Another way • motivation:save time • Another way:Kw = Fk’’’(E(W)) pointer Ekw(Pi) • Search:send < E(W), Kw > to server
Goh Bloom Filter • Document D:S={s1,s2,…,sn} • n : the number of words chosen by client • S : m bit array
Methodology The motivations: • Have a method that allows the owner of the data to find the required data from a remote and untrusted storage • Supports any types of data • Allows Alice to choose just the required keywords describing the data • Preserve the keywords where the keywords can be retrieved if needed • Time complexity of 0(1) to search for a keyword • Easy integration with any existing indexing scheme • Good performance time in terms of encryption, decryption and search
Scheme • Setup/Encryption Phase • Search Phase • Single Document Searching Mode • Multiple Documents Search Mode • Decryption Phase
Hash Table Loc(Wi) = H( Ek’(W)+id ) HT 2 HT 1 HT n Loc(W1) c1 Loc(W1) c1 Loc(W1) c1 c2 c2 Loc(W2) c2 Loc(W2) Loc(W2) c3 c3 Loc(W3) c3 Loc(W3) Loc(W3) … … … cn cn cn Loc(Wn) Loc(Wn) Loc(Wn)
Search Phase:Single Document Searching Mode • 傳給server id // Loc(Wi) // Ek’(W) // ki
Search Phase :Multiple Documents Search Mode • motivation: • Document number is unknown • Find a certain word in multiple documents • This allows server to do the hash function H on behalf of client
SWP Linear Scan disadvantage • Long processing time • The search time increases as with the number of word not suitable for files with many words
SWP encrypted index • The fastest processing time • disadvantage: • A single master index manages all the documents • Maintain a pool of keywords
Goh Bloom Filter • High security • disadvantage: • Large index size • Long processing time • Not suitable for active file server where changes document occur frequently