CN106874516A - Efficient cipher text retrieval method based on KCB trees and Bloom filter in a kind of cloud storage - Google Patents

Abstract

It is a kind of efficient retrieval ciphertext data-selected scheme based on keyword complete binary tree and Bloom filter the invention discloses the efficient cipher text retrieval method based on KCB trees and Bloom filter in a kind of cloud storage.Because data query is all related with the height of tree to the time complexity of renewal on complete binary tree, therefore can guarantee that the operating time maintains O (log n), it is suitable with existing optimal case efficiency.On this basis, the present invention is optimized to index structure again using Bloom filter, while keeping search efficiency suitable with existing optimal case, reduces data storage overhead on the server.Program purpose is to combine huge advantage of the keyword complete binary tree in time and Bloom filter in terms of space, propose that one kind efficiently can search for encrypted indexes structure, while keyword efficient retrieval is realized, encrypted indexes storage overhead on the server is reduced.

Description

Efficient cipher text retrieval method based on KCB trees and Bloom filter in a kind of cloud storage

Technical field

The invention belongs to technical field of data security in cloud storage, efficient retrieval ciphertext data are applied to.Specifically, it is Efficient cipher text retrieval method based on KCB trees and Bloom filter in a kind of cloud storage.

Background technology

Cloud storage is gradually grown up based on cloud computing, by by encryption data store on Cloud Server and More convenient, more efficient access is provided the user with management.Therefore, cloud storage is widely used in every field, with ten Divide wide market application foreground, but safety problem therein also result in the concern of user.

For a user, Cloud Server is not completely believable.In order to protect the privacy of data, user would generally be Data transfer to significant data to being encrypted before Cloud Server.So, using plaintext keyword searching techniques just Fail.Therefore how the cryptograph files that are stored on Cloud Server are carried out by key search just becomes a problem.In order to This problem is solved, encryption (Searchable Encryption, SE) technology is can search for and is just arisen at the historic moment.

SE is a new cryptographic primitive, in the same of data confidentiality of the protection storage on insincere third party entity When, moreover it is possible to ciphertext data are efficiently retrieved.SE can be described as following process：Data owner utilizes set of keywords symphysis Into cryptography key word indexing, and all of encryption file and cryptography key word indexing are uploaded into Cloud Server；When user needs It is that keyword generates query token using trapdoor function, and be sent to Cloud Server during search key；Server uses the order Board is retrieved to cryptography key word indexing；Ciphertext set comprising the keyword is returned to user by server, and user utilizes The cryptograph files that secret key decryption is returned.From for security, key search is carried out using the technology, except user can be exposed " search pattern " (whether guessing any two search statement comprising identical keyword), " access module " (i.e. inspection every time The result of rope), outside file cipher text information and some search vouchers, will not expose any about document and the information of keyword. From for access efficiency, using SE mechanism, user need not retrieve, download and decrypt the cryptograph files not comprising keyword, save Substantial amounts of network overhead and computing cost.

Although the appearance of SE mechanism greatly improves user's access efficiency, research shows, different SE index structures Also the recall precision of keyword will be influenceed.2003, Goh proposed the concept of Security Index, using Bloom filter (Bloom Filter, BF) it is one index of each file build.When user needs to retrieve certain keyword, by the keyword Trapdoor be sent to high in the clouds, then matched with the BF of each file using trapdoor by Cloud Server.On time cost, cloud clothes Business device retrieves all cryptograph files one by one, and the keyword query time is O (n) (n is the number of file in file set), efficiency compared with It is low.2006, Curtmola et al. improved the scheme of Goh, first using the method for inverted index.The program is crucial for each Word constructs a keyword chained list, the storage file identifier of each chained list node encryption, the pointer of next node and under One decruption key of node.Under improving herein, the keyword search time has been reduced to O (r), and (r is comprising the keyed file Number), but due to being required for being decrypted present node calculating when accessing every time, therefore computing cost is larger.

2011, Chai proposed a kind of encrypted indexes structure similar to dictionary tree.In this scenario, keyword query Time is directly proportional to the length of keyword most long.Although this scheme efficiency is higher, and is capable of the deceptive practices of detection service device, Dictionary attack can not be resisted, security is poor.Kamara in 2013 et al. is using keyword RBTree (KRB trees) for file build adds Close index, it is proposed that parallel dynamic symmetry can search for encipherment scheme.The program in the case of p processor, key search Time is O ((r/p) logn)；As p=logn, the parallel key search time is O (r)；As p=r, during key search Between for O (logn), the i.e. size with r it is unrelated.Although the cryptograph index structure based on KRB trees can improve the retrieval of keyword Efficiency, but because the storage organization of all nodes in KRB trees is keyword Hash table, be by key value (key, Value data) are accessed.And each node stores two vectors of m bits, space complexity is at least O (2m² (2n-1)) bits (being directly proportional to mn, wherein m is keyword number, n is file number), this undoubtedly increases Cloud Server Storage overhead.

The content of the invention

It is huge in terms of room and time it is an object of the invention to combine keyword complete binary tree and Bloom filter Big advantage, proposes a kind of efficient SE index structures, while keyword efficient retrieval is realized, reduces encrypted indexes in service Storage overhead on device.

Before the content of the invention is described, we first define the cryptography primitive on several bases：

1. a certainty pseudo-random function：P：{0,1}^k×{0,1}ⁿ→{0,1}^m；

2.y separate hash function：H={ h_i:{0,1}^k×{0,1}ⁿ→{1,x}^m}(1≤i≤y)。

In order to illustrate step of the invention, following four algorithm is defined first：

(1)、Gen(1^k)→K

Key needed for initialization scheme, 1^kIt is system security parameter, as the input of function.In security parameter 1^kUnder, it is raw Into the key K of pseudo-random function₁And the key K of symmetric encipherment algorithm₂.Algorithm output key K=(K₁,K₂)。

(2)、BuildIndex(F,W,K)→I

1), construction keyword complete binary tree (Keyword Complete Binary Tree, KCB)

Input clear text file set F={ f₁,…,f_nAnd set of keywords W={ w₁,w₂,…,w_m, wherein, n is file The number of file in set, m is the number of keyword in set of keywords, the equal changeable of number of n and m.We are by as follows Step constructs KCB trees：

A) All Files in the leaf node respective file set F of KCB trees, therefore the number of plies l of KCB trees can be by file number N decisions, i.e.,The leaf node number of complete binary tree isIfThen fill surplus with empty node Remaining leaf node.

B) node for defining KCB trees is u_ab(1≤a≤l,1≤b≤2^(a-1)).Wherein a is node u_abThe number of plies at place, b It is node u_abIn a layers of sequence number.In node u_abTwo tuples of upper storage (id (u_ab), data_ab)。id(u_ab) it is node identification Symbol.data_abFor a length is the array of m.data_abI-th bit be defined as data_ab[i], wherein i ∈ { 1 ..., m }.

If node u_abIt is leaf node：By id (u_ab) it is defined as the corresponding document f of the node_jDocument identifier ID_j.It is fixed Adopted data_ab[i]=11 (1≤i≤m), file f that and if only if_jIn include keyword w_i；Define data_ab[i]=00, when and only Work as file f_jIn do not include keyword w_i；If the leaf node is filling node, data is set_ab[i]=10.

If node u_abIt is non-leaf nodes：If its left child and right child are present comprising keyword w_iLeaf node, then By internal node array data_abCorrespondence position data_ab[i] is set to 11；If only its left child is present comprising keyword w_i Leaf node, then by internal node array data_abCorrespondence position data_ab[i] is set to 10；If only its right child is present Comprising keyword w_iLeaf node, then by internal node array data_abCorrespondence position data_ab[i] is set to 01；If its left child Sub and right child does not exist comprising keyword w_iLeaf node, then by internal node array data_abCorrespondence position data_ab [i] is set to 00.

File index T is substantially a KCB tree for unencryption, and complete binary search tree can guarantee that and grasp in worst case Make the time for O (logn).Although the advantage of KCB tree efficient lookups can realize the quick-searching of keyword, if by complete two The 2n-1 node of fork tree T is all stored on Cloud Server, and will bring huge storage overhead to server.Therefore it is of the invention Index structure is optimized again using Bloom filter, accelerates the recall precision of keyword, the storage for reducing server is opened Pin.

2), by file index tree T generation encrypted indexes I

File index tree T is generated into encrypted indexes I using Bloom filter.Input key K, file set F={ f₁,…, f_n, file index T.Following steps are performed afterwards：

A) to all node u in file index tree T_ab(1≤a≤l,1≤b≤2^(a-1)) and all of keyword w_i(1≤ I≤m), calculateAnd it is mapped to Bloom filter using y separate hash function On：

B) in order to prevent loss of data on Cloud Server, file index tree T is retained in home server.Deposited on server Storage encrypted indexes I.

Because file index T uses complete binary tree structure, each node identifier in the structure can lead to Cross and be calculated.From u₁₁Node rises, and server is able to know that next node identifier to be accessed.Therefore KCB need not be uploaded Structure, only need to upload to high in the clouds by encryption file and encrypted indexes.

When that can bear some wrong reports, Bloom filter possesses very big space advantage.There is 1% rate of false alarm for one Bloom filter for, storage all elements only need to (9.6m (2n-1)) bits (wherein m be keyword number, n is file Number).

(3)、SrchToken(K,w_i)→τ

Search token is generated using key word of the inquiry.Input key K and keyword w_i, output search token

(4)、Search(C,I,τ)→C_(x,y)

Server receives search token τ, and cryptograph files are retrieved by encrypted indexes I, and return meets the close of condition Literary file set C_(x,y)。

When user needs to retrieve certain keyword, the search token of the keyword is generated by trapdoor function. From section u₁₁Point starts, and token and node identifier, search are judged into symbol is mapped on Bloom filter.According to the value meter for judging symbol The node identifier that next calculation will be retrieved, then carries out recursion searches, until finding the encryption file for meeting search condition Set.Contain keyword w when judging that symbol shows not existing_iFile when, server to user return do not find.

What goal of the invention of the invention was realized in：

Efficient cipher text retrieval method based on KCB trees and Bloom filter in cloud storage of the present invention is a kind of using crucial The characteristics of word complete binary tree query time is soon and Bloom filter space efficiency is high, a kind of efficient SE index structures of proposition. The search time of keyword can be maintained O (logn) by keyword complete binary tree, with existing optimal scheme efficiency phase When.On this basis, the present invention is optimized to file index tree T again using Bloom filter, by owning KCB trees Node is mapped on Bloom filter, it is to avoid all store on Cloud Server the 2n-1 nodal information of complete binary tree T. While realizing keyword efficient retrieval, encrypted indexes storage overhead on the server is reduced.

Brief description of the drawings

Fig. 1 is to can search for encrypting schematic diagram；

Fig. 2 is file index tree T of the present invention using keyword red and black tree construction construction；

Fig. 3 is the schematic diagram that file index T is mapped to the present invention Bloom filter；

Fig. 4 is that the present invention is the efficient cipher text retrieval method based on KCB trees and Bloom filter in a kind of cloud storage of the invention Technical scheme figure, is also Figure of abstract of the invention；

Specific embodiment

Embodiment of the present invention is described in detail below in conjunction with the accompanying drawings, so that those skilled in the art can It is better understood from the present invention.

The present invention, for file set constructs file index tree T, recycles Hash using the structure of keyword complete binary tree T is mapped to construction encrypted indexes I on Bloom filter by function, and concrete scheme is as follows：

(1), first, using key schedule Gen, the key K needed for initializing the present invention₁And K₂。

(2), data owner is encrypted to clear text file, obtains cryptograph files set C={ c₁,…,c_n}。

(3) it is, clear text file collection F={ f followed by algorithm BuildIndex (F, W, K)₁,…,f_nConstruction file index T, such as Fig. 2.All node u of the file index T that will be obtained_abAnd all keyword w_iIt is mapped on Bloom filter.As schemed 3, obtain encrypted indexes I.

(4), when user needs to retrieve keyword, using algorithm SrchToken generation search tokensAnd it is sent to Cloud Server.

(5), server receives search token τ, and cryptography key word indexing I is retrieved using algorithm Search, finds Corresponding cryptograph files.

If judging symbol π₁=00, π₂=01, π₃=10, π₄=11.From u₁₁Start, the value of the judgement symbol according to hit is calculated Identifier id (the u of the node next to be retrieved_ab), recursive lookup is then carried out, until finding the ciphertext for meeting querying condition File set.

By identifier id (u_ab)、And π_i(1≤i≤4) do following computing：

1), current retrieval is internal node：

If a) π₁Hit, represents not existing and contains keyword w_iFile.Cloud Server is returned to user and not found.

If b) π₂Hit, represents that the right child of present node is present comprising keyword w_iFile, calculating next to retrieve Node be u_(a+1)(2b)。

If c) π₃Hit, represents that the left child of present node is present comprising keyword w_iFile, calculating next to retrieve Node be u_(a+1)(2b-1)。

If d) π₄Hit, represents that the left child and right child of present node are present comprising keyword w_iFile, calculating connects The node to be retrieved that gets off is u_(a+1)(2b-1)And u_(a+1)(2b)。

2), current retrieval is leaf node：

If a) π₁Hit, represents file f_jWithout keyword w_i。

If b) π₄Hit, represents file f_jContain keyword w_i。

If c) π₃Hit, represents that the leaf node is filling node.

According to above step, cryptograph files set C corresponding with search token is found from Cloud Server_(x,y)。

(6), decrypting ciphertext file set C_(x,y), obtain corresponding clear text file collection F_(x,y)。

Claims (1)

1. the efficient cipher text retrieval method of KCB trees and Bloom filter is based in a kind of cloud storage, it is characterised in that including following Several steps：

(1), the key needed for initialization scheme

Using key schedule Gen (1^k) → K, the key K needed for initializing the present invention₁And K₂, wherein 1^kIt is that system is joined safely Number, in security parameter 1^kUnder, generate the key K of pseudo-random function₁And the key K of symmetric encipherment algorithm₂；

(2) it is, clear text file collection construction file index tree T using algorithm BuildIndex (F, W, K)

File index tree T is substantially a KCB tree for unencryption (Keyword Complete Binary Tree), and input is in plain text File set F={ f₁,…,f_nAnd set of keywords W={ w₁,w₂,…,w_m, wherein, n is the number of file in file set, M is the number of keyword in set of keywords, the equal changeable of number of n and m, and we construct KCB trees as follows：

A), the All Files in the leaf node respective file set F of KCB trees, therefore the number of plies l of KCB trees can be determined by file number n It is fixed, i.e.,The leaf node number of complete binary tree isIfThen remaining leaf is filled with empty node Node；

B) node for, defining KCB trees is u_ab(1≤a≤l,1≤b≤2^(a-1)), wherein a is node u_abThe number of plies at place, b is section Point u_abIn a layers of sequence number, in node u_abTwo tuples of upper storage (id (u_ab), data_ab), id (u_ab) it is node identifier, data_abFor a length is the array of m, data_abI-th bit be defined as data_ab[i], wherein i ∈ { 1 ..., m }；

If node u_abIt is leaf node：By id (u_ab) it is defined as the corresponding document f of the node_jDocument identifier ID_j, definition data_ab[i]=11 (1≤i≤m), file f that and if only if_jIn include keyword w_i, define data_ab[i]=00, and if only if File f_jIn do not include keyword w_iIf the leaf node is filling node, sets data_ab[i]=10；

If node u_abIt is non-leaf nodes：If its left child and right child are present comprising keyword w_iLeaf node, then should Internal node array data_abCorrespondence position data_ab[i] is set to 11, if only its left child is present comprising keyword w_iLeaf Node, then by internal node array data_abCorrespondence position data_ab[i] is set to 10, if only its right child exist comprising Keyword w_iLeaf node, then by internal node array data_abCorrespondence position data_ab[i] is set to 01, if its left child and Right child does not exist comprising keyword w_iLeaf node, then by internal node array data_abCorrespondence position data_abIt is set to 00；

KCB trees can guarantee that the operating time is O (logn) in worst case, although the advantage of KCB tree efficient lookups can be real The quick-searching of existing keyword, if but all store on Cloud Server the 2n-1 node of complete binary tree T, again will be to service Device brings huge storage overhead, therefore the present invention to be optimized to index structure again using Bloom filter, accelerates crucial The recall precision of word, reduces the storage overhead of server；

(3), by file index tree T generation encrypted indexes I

File index tree T is generated into encrypted indexes I using Bloom filter, key K, file set F={ f is input into₁,…,f_n, File index T, performs following steps afterwards：

A), to all node u in file index tree T_ab(1≤a≤l,1≤b≤2^(a-1)) and all of keyword w_i(1≤i≤ M), calculateAnd be mapped on Bloom filter using y separate hash function：

$h_p\left(\mathrm{id}\left(u_{\mathrm{ab}}\right)\right|\left|P_{K_1}\left(w_i\right)\right|\left|{\mathrm{data}}_{\mathrm{ab}}\right[i\left]\right)(1\≤p\≤y,1\≤a\≤l,1\≤b\≤2^{(a-1)},1\≤i\≤m)$

B), in order to prevent loss of data on Cloud Server, file index tree T is retained in home server, is stored on server Encrypted indexes I；

Because file index T uses complete binary tree structure, each node identifier in the structure can be by meter Obtain, from u₁₁Node rises, and server is able to know that next node identifier to be accessed, therefore need not upload KCB structures, Only encryption file and encrypted indexes need to be uploaded to high in the clouds；

When that can bear some wrong reports, Bloom filter possesses very big space advantage, the cloth for having 1% rate of false alarm for one For grand filter, storage all elements only need to (9.6m (2n-1)) bits, and (wherein m is keyword number, and n is file Number)；

(4), using algorithm SrchToken generation search tokens

When user needs to retrieve certain keyword, using algorithm SrchToken (K, w_i) → τ, is input into key K and pass Key word wi, output search token

(5), server receives search token τ, and encryption keyword is retrieved using algorithm Search

Server receives search token τ, and cryptograph files are retrieved by encrypted indexes I, and return meets the ciphertext text of condition Part set C_(x,y), perform step as follows：

If judging symbol π₁=00, π₂=01, π₃=10, π₄=11, from u₁₁Start, the value of the judgement symbol according to hit is calculated and connect down Carry out the identifier id (u of the node to be retrieved_ab), recursive lookup is then carried out, until finding the cryptograph files for meeting querying condition Set；

By identifier id (u_ab)、And π_i(1≤i≤4) do following computing：

$h_p\left(id\right(u_{ab}\left)\right|\left|P_{K_I}\right(w_i\left)\right|\left|{\mathrm{\π}}_i\right),(1\≤p\≤y,1\≤i\≤4,1\≤a\≤l,1\≤b\≤2^{(a-1)})$

Current retrieval is internal node：

If a), π₁Hit, represents not existing and contains keyword w_iFile, Cloud Server to user return do not find；

If b), π₂Hit, represents that the right child of present node is present comprising keyword w_iFile, next calculating to be retrieved Node is u_(a+1)(2b)；

If c), π₃Hit, represents that the left child of present node is present comprising keyword w_iFile, next calculating to be retrieved Node is u_(a+1)(2b-1)；

If d), π₄Hit, represents that the left child and right child of present node are present comprising keyword w_iFile, calculating connect down Carry out the node to be retrieved for u_(a+1)(2b-1)And u_(a+1)(2b)；

Current retrieval is leaf node：

If a), π₁Hit, represents file f_jWithout keyword w_i；

If b), π₄Hit, represents file f_jContain keyword w_i；

If c), π₃Hit, represents that the leaf node is filling node；

According to above step, cryptograph files set C corresponding with search token is found from Cloud Server_(x,y)。