Patent Grant
11341128
Encrypted databases provide data protection (security) in cloud platforms and/or database-as-a-service (DaaS) settings. In encrypted databases, data (plaintext) can be encrypted at the client to provide encrypted data (ciphertext), which can be provided to the database for storage. In some examples, a third-party provides and maintains the database. That is, the database is outsourced to the third-party. For example, a client encrypts data using one or more encryption keys to provide encrypted data, which the client sends to the third-party for storage in the database.
Outsourcing a database offers efficient resource management and low maintenance costs for clients, but exposes outsourced data (client data) to a service provider (the third-party providing the database and its agents). For example, privacy-preserving range queries allow encrypting data, while still enabling tests on ciphertexts, if their corresponding plaintexts fall within a requested range. This provides a data owner the possibility to outsource data collections to a cloud service provider without sacrificing privacy nor losing functionality of filtering the data. However, current schemes for handling range queries either leak additional information (e.g., ordering of the complete data set), or significantly slow the search process.
Implementations of the present disclosure include computer-implemented methods for poly-logarithmic range queries on encrypted data stored in a database on a server. In some implementations, actions include receiving, by a server-side computing device, a range query token from a client-side computing device, determining, by the server-side computing device, one or more of whether a tree list of an encrypted search index is empty and a range of the range query token is intersecting with a range accounted by at least one tree in the tree list, the encrypted search index including the tree list and a point list, receiving, by the server-side computing device, encrypted query results based on one of a search tree, if the tree list is not empty and a range of the range query token is at least a sub-range of a range accounted for in the tree list, and the point list, if the tree list is empty or the range of the range query token is not at least a sub-range of a range accounted for in the tree list, and updating, by the server-side computing device, the encrypted search index based on the range query token and a client-server protocol. Other implementations of this aspect include corresponding systems, apparatus, and computer programs, configured to perform the actions of the methods, encoded on computer storage devices.
These and other implementations can each optionally include one or more of the following features: updating the encrypted search index includes refining a search tree of the tree list at least partially based on transmitting the range query token and a previously received range query token to the client-side computing device, and receiving refined range tokens from the client-side computing device, the encrypted search index being updated based on the refined range tokens; updating the encrypted search index includes extending a search tree of the tree list to provide an extended search tree at least partially based on transmitting a root node of the search tree and a previously received query corresponding to the search tree to the client-side computing device to provide a new range query token corresponding to the extended search tree; updating the encrypted search index includes merging multiple search trees of the tree list to provide a merged search tree at least partially based on transmitting respective root nodes of the multiple search trees to the client-side computing device, and receiving a revised token corresponding to the merged search tree; the multiple trees are merged in response to determining that there is no value gap between the multiple trees; actions further include receiving the encrypted search index and ciphertext from the client-side computing device; and the tree list includes at least one search tree based on a previously received range query token.
The present disclosure also provides a computer-readable storage medium coupled to one or more processors and having instructions stored thereon which, when executed by the one or more processors, cause the one or more processors to perform operations in accordance with implementations of the methods provided herein.
The present disclosure further provides a system for implementing the methods provided herein. The system includes one or more processors, and a computer-readable storage medium coupled to the one or more processors having instructions stored thereon which, when executed by the one or more processors, cause the one or more processors to perform operations in accordance with implementations of the methods provided herein.
It is appreciated that methods in accordance with the present disclosure can include any combination of the aspects and features described herein. That is, methods in accordance with the present disclosure are not limited to the combinations of aspects and features specifically described herein, but also include any combination of the aspects and features provided.
The details of one or more implementations of the present disclosure are set forth in the accompanying drawings and the description below. Other features and advantages of the present disclosure will be apparent from the description and drawings, and from the claims.
Like reference symbols in the various drawings indicate like elements.
Implementations of the present disclosure include computer-implemented methods for poly-logarithmic range queries on encrypted data stored in a database on a server. More particularly, implementations of the present disclosure enable range queries requested from a client (e.g., data owner) to be compared on a server-side (e.g., hosted database). In this manner, the service provider can use information gained from previously queried ranges to speed-up search time for subsequent range queries. In some implementations, values that have fallen within a queried range are stored in an interactively built search index for subsequent range requests. In some implementations, values that have not been queried before do not leak any information to the cloud service provider and stay perfectly secure. As discussed in further detail herein, implementations of the present disclosure reduces information leakage while supporting poly-logarithmic search time on average.
In some implementations, the server system 104 includes at least one server device 108 and at least one data store 110. In the example of
In some implementations, the network 106 can include a large computer network, such as a local area network (LAN), a wide area network (WAN), the Internet, a cellular network, a telephone network (e.g., PSTN) or an appropriate combination thereof connecting any number of communication devices, mobile computing devices, fixed computing devices and server systems.
In accordance with implementations of the present disclosure, the server system 104 maintains a database that stores encrypted data (e.g., an encrypted database). In some examples, the data (plaintext) is encrypted at the computing device 102 (client-side), and the encrypted data (ciphertext) is sent to the server system 104 over the network 106 for storage. In some implementations, and as described herein, the server system 104 can be provided by a third-party service provider, which stores and provides access to the encrypted data.
In some implementations, a database is part of a database system that also includes a query execution engine. In some examples, the query execution engine receives a query (e.g., a search token from a client), provides a query plan based on the query, executes the query plan to provide a response (e.g., encrypted data from the database that is responsive to the query), and transmits the response to the client (e.g., from which the query was received). In some examples, the query plan includes one or more operations to be performed on the encrypted data.
Implementations of the present disclosure are described with reference to the example context of outsourcing data to a cloud service provider (e.g., database on the cloud). Cloud computing enables a data owner to outsource data, while enabling the data owner to access the data using one or more computing devices. For example, even computing devices with limited computation power (e.g., a smartphone, relative to a desktop computer) can be used to access relatively large data collections. This is possible by delegating computationally expensive operations (e.g., searching) to the cloud service provider. In some examples, a relatively small subset of data matching a search query is transmitted to and processed by the client-side computing device.
To preserve data privacy, the outsourced data is encrypted. For example, the data (plaintext) is encrypted at the client-side to provide encrypted data (ciphertext), which is transmitted to and stored at the server-side. In some examples, standard encryption schemes are not suitable for the data outsourcing scenario, because they do not enable processing of encrypted data. As a result, the complete encrypted data would need to be transferred to the client-side computing device, and decrypted and processed locally.
Advanced encryption schemes enable a server-side computing device (of the cloud service provider) to perform search operations, such as exact pattern matching and/or range queries on ciphertexts, for example. In some examples, the data owner can encrypt the data and augment the encrypted data with additional information (e.g., keywords, timestamps). The data owner transfers the ciphertexts to the cloud service provider. Using a secret key (e.g., an encryption key known only to the data owner), the data owner can create a search token (e.g., for exact pattern matching of a keyword, a range the timestamp should fall within), which is transmitted to the cloud service provider. Using this search token, the cloud service provider can filter the encrypted data for all ciphertexts that match with the search token, and can return matching ciphertexts to the data owner.
To be practical, searchable encryption schemes should be efficient and secure. In some examples, search operations should be possible in sub-linear search time, which can be achieved by an appropriately encrypted search index. In order to protect privacy of outsourced data, the encrypted data and the encrypted search index should leak as little information as possible. For example, ciphertexts that have never been searched for should not unveil any information.
In view of this, implementations of the present disclosure provide an encrypted search scheme for implementing privacy-preserving range queries on encrypted data. In accordance with the present disclosure, and as described in further detail herein, implementations enable the cloud service provider to compare range tokens that have already been queried in previous search requests. This enables the cloud service provider to decrease its average search time for range queries. While initial search time for a range query is linear in the number of indexed ciphertexts (e.g., files), subsequent queries can be accelerated. For example: in a first, initial search, the cloud service provider learns the result set of a range query; given a range query in a second search request that is a subrange of the already queried range in the first step, it is sufficient to scan the previously learned result set. In this manner, the search space of the second query is downscaled, which results in a quicker search operation. Furthermore, using this approach for every new range query, the cloud service provider can construct and refine an encrypted search index in an interactive protocol between the client and the server, as described herein. Consequently, the encrypted search scheme of the present disclosure achieves decreased search time on average. In addition, ciphertexts that have never fallen within any queried range are not contained in any access pattern. Consequently, using a suitable encryption scheme, the un-queried ciphertexts do not leak any information.
In some implementations, properties of the encrypted search scheme for privacy-preserving range queries can be summarized with respect to security, efficiency, and modularity. Security is formalized using a simulation-based approach in a non-adaptive model. In some examples, leakage functions are provided that give an upper bound for information that is leaked by the encrypted search scheme of the present disclosure. With respect to efficiency, the encrypted search scheme has amortized poly-logarithmic runtime. This is achieved by interactively building a search index. With respect to modularity, the encrypted search scheme is built on a black box interface for functional encryption.
To provide further context for implementations of the present disclosure, governing definitions are provided. In some examples, denotes the set of natural numbers, and [i,j] is denoted with i≤j and i,j∈
as the set of integers starting at i and including j (i.e., the set {i, . . . , j}). An output z of a (possible probabilistic) algorithm
is written as z←
. In some examples, λ denotes a security parameter. A function ƒ:
→
is called negligible (in x), if, for every positive polynomial p(⋅) there exists a x0, such that, for all x>x0, it holds that ƒ(x)<1/p(x). Given a matrix M, M[i] denotes the i-th line, and M[i][j] denotes the j-th element in the i-th line. A message m is provided as a tuple (ƒ,
), where ƒ is an arbitrary file (e.g., a document, an image), and a value point
, which is used for indexing. In some examples, each message m has a unique identifier ID(m), and each file ƒ can be accessed given the associated identifier ID(m). For range Q=[q(s), q(e)], IDQ can be defined as the set of file identifiers indexed under values v with
∈Q (i.e., IDQ={ID(m): m=(ƒ,
) with q(s)≤
≤q(e)}).
In some implementations, the scheme for secure and efficient range queries is composed of the following (partly probabilistic) polynomial-time algorithms: SRQ-Setup, SRQ-Enc, SRQ-IndexFile, SRQ-Token, SRQ-Search. In an initial step, the data owner creates public parameters and a master key for a desired value domain by running SRQ-Setup. In some examples, the public parameters are known by all parties. In a next step, a message collection (plaintext) is encrypted and indexed under given value points by running SRQ-Enc. In some examples, each value point lies in the value domain used in the initial setup. The result of these steps includes an encrypted index and a ciphertext collection, which are transferred to the server-side (e.g., cloud service provider) using SRQ-IndexFile. The data owner (holding the master key mk) is able to create range tokens (for respective range queries) by calling SRQ-Token. A range token can be transmitted to the server-side, such that the respective range query can be executed over encrypted data. In some examples, given a range token, the server can run SRQ-Search to filter all (encrypted and indexed) messages associated with value points falling within the requested range.
Definition 1: The scheme SRQ for secure range queries includes the following (probabilistic) polynomial time algorithms:
In some implementations, an order-preserving encryption scheme can be used. For example:
OPE=(OPE-Setup,OPE-Enc,OPE-Dec), where
OPE-Enc(x)≤OPE-Enc(y)x≤y.
This example OPE scheme can be used for building an encrypted search index. For example, in each message m=(ƒ, ), the value point v is encrypted using OPE-Enc, and is associated with the message identifier ID (m). All index entries of the form ci=(OPE-Enc(
i), ID(mi)) are sorted according to their first element. For search queries for range [q(s), q(e)] a range token is implemented as a tuple τQ=(OPE-Enc(q(s)), OPE-Enc(q(e))). Given τQ to the server storing the search index, the server-side is able to obtain the set {(OPE-Enc(
i), ID(mi)): OPE-Enc(q(s))≤OPE-Enc(
i)≤OPE-Enc(q(e))} in logarithmic time by running binary search. However, even indexed, but not queried points can be compared with all other indexed (queried and not queried) points.
In another example approach to hide the information about the order of all points not queried before, range predicate encryption (RPE) can be used. An example RPE scheme includes the following example algorithms:
In view of the example RPE scheme, a scheme for range queries with linear runtime (in the number of indexed elements) can be provided. That is, all attributes are encrypted using RPE-Enc and added to the point list . For every range query of range Q a token τQ is created by the data owner holding the master key using RPE-Token, which is transmitted to the server. The server can provide IDQ by adding all entries c∈
with RPE-Dec(tkQ, c)=1. In some examples, all entries are checked. Consequently, runtime is linear in the number of all indexed files (ciphertexts).
In some implementations, the search index is structured (e.g., all indexed messages are arranged) in a manner to increase average search speed. Furthermore, the index structure is designed to leak as little information as possible. For all ciphertexts, security can be defined using the following example security game.
Definition 2: RPE is a scheme for range predicate encryption. The following example security game between attacker and a challenger includes:
In some examples, RPE has selective secure plaintext privacy, if, for all probabilistic polynomial-time attackers running this security game, it holds that
where ∈ is negligible in λ.
Definition 3: RPE is a scheme for range predicate encryption. The following example security game between attacker and a challenger includes:
In some examples, RPE has selective secure predicate privacy, if, for all probabilistic polynomial-time attackers running this security game, it holds that
where ∈ is negligible in λ.
Given such an RPE scheme, a scheme for range queries with linear runtime (in the number of indexed elements) can be provided. That is, all attributes are encrypted using RPE-Enc and added to the point list . For every range query of range Q a token τQ is created by the data owner holding the master key using RPE-Token. Given this token, the server can create IDQ by adding all entries c∈
with RPE-Dec(tkQ, c)=1. Because all entries are checked, runtime is linear in the number of all indexed files.
In some examples, to increase average search speed, all indexed messages are arranged in a suitable index structure. Furthermore, the index structure is designed to leak as little information as possible. Given a SRQ scheme for secure and efficient range queries, the encryption operation is extended in way, such that it is possible to encrypt a file ƒ indexed under a value point . As a result, the security game for plaintext privacy is slightly modified to provide the following:
Definition 4: SRQ scheme is a scheme for secure and efficient range queries. The following example security game between attacker and a challenger includes:
In some examples, SRQ has selective secure plaintext privacy, if, for all probabilistic polynomial-time attackers running this security game, it holds that
where ε is negligible in λ.
Regarding predicate privacy, implementations of the present disclosure use properties given directly from RPE predicate privacy. Using these definitions and resulting properties, a security definition for the SRQ scheme of the present disclosure can be provided. In some examples, security properties are defined using a leakage based definition.
Definition 5: Given a scheme for secure range queries SRQ=(SRQ-Setup, SRQ-Enc, SRQ-IndexFile, SRQ-Token, SRQ-Search) and security parameter λ∈, the following probabilistic experiments with an adversary
, a simulator
and leakage functions
1,
2 are provided:
In some examples, the SRQ scheme is (1 (M),
2 (Q))-secure against non-adaptive chosen-range attacks, if, for all probabilistic polynomial-time algorithms
there exists a probabilistic polynomial-time simulator
so that advantage of
defined as:
|Pr[Rea(1λ)=1]−Pr[Idea
(1λ))=1]|
is negligible in λ.
In accordance with implementations of the present disclosure, the server uses information gained from previous range queries to reduce the number of files for executing future queries. In this manner, search time is decreased. For further discussion, it can be assumed that the server (cloud service provider) has learned a range token τQ and its result set IDQ after performing a scan over the complete set of indexed messages. Given a new range token τR with R⊂Q scanning result set IDQ (that is a subset of the complete set of indexed values) suffices to obtain IDR. Furthermore, the server can compute the result set for the relative complement of R in Q (i.e., IDQ\R). This information can also be used for responding more quickly to queries in the future.
In some implementations, in order to increase average search speed, all indexed messages are arranged in the search index in a suitable manner, and the search index is updatable efficiently. Furthermore, structure of the search index is designed to leak as little information as possible. In accordance with implementations of the present disclosure, the search index is organized into multiple components:
Within the tree list , encrypted (1-dimensional) R-trees are stored. In some examples, an R-tree Γ completely covers a coherent range. Each inner node includes up to t entries.
Further, for any two entries (p1, R1) and (p2, R2) of the same node, it holds that R1∩R2=Ø (i.e., the ranges in one node do not overlap). For every entry (p, R), it holds that the subtree rooted at the node pointed to by p covers range R (i.e., Γ[p]=R). Furthermore, all leafs consist of up to t entries, and each entry has the form (obj, R), where R is a range and obj is a pointer to a list of all files indexed under values v∈R (i.e., obj points to IDR). Given a range Q=[q(s), q(e)], the server holding a tree F covering a superset of Q (i.e. Q⊂Γ) can calculate IDQ by using Algorithm 1 in logarithmic time (in the number of nodes):
∈ Q
In some examples, the only two operations needed for such a range query are the following: first, checking if start point q(s) and end point q(e) of Q falls within range R (i.e., if R and Q intersect), and second, checking if range Q is a subrange of range R. In some examples, this functionality is provided by a modified RPE scheme. Consequently, a range query can also be answered over search trees that include ranges encrypted by the modified RPE scheme. In some examples, a range token for range Q created by RPE-Token must be augmented with encrypted limiting points (that is start and end point) encrypted using RPE-Enc.
In some examples, in order to add new values to an existing tree, the correct leaf and entry must be identified. That is, the entry that contains the range the new value(s) falls within. In some examples, this can be done using Algorithm 2:
∈ R.
∈ R then
∈ R;
∈ R;
In some implementations, using the encrypted index structure, an SRQ scheme can be defined as described in the following. Given an RPE scheme including the algorithms RPE-Setup, RPE-Enc, RPE-Token, RPE-Dec, an (IND-CPA) secure encryption scheme Π1=(GenIND-CPA, EncIND-CPA, DecIND-CPA) and a second encryption scheme Π2=(Gen, Enc, Dec), an SRQ scheme is provided as follows:
Using the above-described algorithms, it is possible to outsource encrypted data, and support range queries. For example, in an initial step the data owner creates a master key by calling SRQ-Setup and defines a possible value domain. The data owner encrypts files by calling SRQ-Enc, where every file is associated with one value point and indexed under this point. The encrypted file is transferred to the server and added to the search index by calling SRQ-IndexFile. In some examples, the files (e.g., encrypted files) can be deleted on the client side. The data owner holding the master key can create search tokens for ranges by calling SRQ-Token. In some examples, the server can compare different range tokens without knowing the master key. More specifically, holding two tokens τQ=(cQ(s), cQ(e), tkQ, cQ) and τR=(cR(s), cR(e), tkR, cR) the server is able to check for the following example properties:
In some implementations, the server receiving a range token τQ=(cQ(s), cQ(e), tkQ, cQ) for range Q=[q(s), q(e)] can search for all files associated with values falling within the range Q by using SRQ-Search. In an initial step, the server checks, for every tree Γi∈, if the tree covers a subrange of Q or intersects with range Q. For example, all entries ((
1, TR
m
)) contained in the root node of any tree Γi are compared with range Q, where Γi covers a subrange, if all ranges R1, . . . , Rm
={Γi|Γi⊂Q} of all trees that cover a subrange of Q is created. In some examples, partial intersections of indexed search trees and the queried range are computed. In some examples, Γ(s) is the tree containing range {circumflex over (R)} covering cQ(s) (i.e., {circumflex over (R)}∈Γ(s) so that RPE-Dec ({circumflex over (R)}, cQ(0))=1). If no such tree was found, set Γ(s)=⊥. The same is done for the encrypted end value cQ(e), which results in a tree Γ(e)∈
and Γ(e)=⊥ otherwise.
There are multiple scenarios that can happen for a given range token τQ. Each scenario results in different behavior of SQR-UpdateIndex described herein. Example scenarios include:
From a high-level perspective, a new range token contains some new information given to the server, such as the result set IDQ for this new range token. This newly gained information is implicit in the search and access pattern, and can be used by the server to update the encrypted search index for future queries. As noted above, given SQR-UpdateIndex(τQ, IDQ, Γ(s), Γ(e), ), four different update situations can occur, where the server has to either refine one tree, create a new tree, extend one tree, or merge trees. In addition, trees that are covered completely by Q (i.e., contained in
) are composed using a combination of tree extension and tree merges.
Because operations can make it necessary to create new range tokens for encrypted trees, and this creation is only possible with the master key, these updates are interactive protocols between server and data owner. In the continued discussion, actions performed at the client-side are denoted as @c in the algorithms below. This could be necessary, because the operation must be performed on plaintext or the creation of new range tokens is necessary.
Because operations add new entries to one or more existing trees, these operations can include a rebalancing step, discussed herein as Algorithm 3, to guarantee every node's size is lower than threshold t afterwards. Again, rebalancing a tree forces the creation of new range tokens, so this is also provided as an interactive protocol.
With regard to refining a tree, the server sends the new range token and previous range tokens that intersect with this new token to the data owner asking for help. The data owner decrypts the range tokens creates (up to) four not intersecting, but more refined ranges and sends back their tokens generated by SRQ-Token. The server can replace the old range tokens with the new, more refined tokens and the indexed file lists are segmented according to these new tokens. Algorithm 4 provides a formal description. Because this replacement increases the entries in a node, the server runs RebalanceTree.
In some implementations, if Γ(s)=Γ(e)=⊥ and is empty, the server creates a new tree. In some examples, the server creates a new, encrypted tree Γ with one entry τQ and indexed item IDQ. This tree Γ is added to tree list
.
In some implementations, a tree is extended, if a new range token intersects partially with a tree (i.e., the range token intersects with the tree), but at least one limiting point of this newly queried range does not. In some examples, extending a tree is started by the server sending the newly learned range token and the root node to the data owner. The data owner decrypts all ranges to reconstruct the whole range currently covered by this tree. A new range token for the gap between the range covered by the tree and the boundary points of the new range token lying outside the tree range is created and added to the tree's leaf. Furthermore, the tree's inner nodes (up to the root) are updated. That is, the indexed range of all inner nodes are replaced by an extended version. Algorithm 5 provides a formal description. In some examples, because at least one leaf got a new entry, the resulting tree is rebalanced after tree extension.
In some implementations, two trees are merged, if they both intersect with the newly queried range. In some examples, these two trees should not have a value gap between them. In some implementations, the end point covered by one tree is directly followed by the start point covered by the other tree. This can be achieved using tree extension as discussed above. In some examples, to merge trees in logarithmic time, the tree {tilde over (Γ)} is integrated with the lower height into the tree Γ with greater height. A new entry in an inner node of Γ is created pointing to the root of {tilde over (Γ)}. This newly covered range is propagated through the inner nodes up to the root. Algorithm 6 provides a formal description. In some examples, the resulting tree is rebalanced.
= r(s)
= r(e);
← RPE - Enc(k,
);
∈ Ri and τ{tilde over (R)}
In some implementations, if a range token has been queried where multiple trees fall within, tree extension and tree merging can be combined. In some examples, all roots in , Γ(s), Γ(e) and the newly queried range token τQ are sent to the client. The client decrypts all roots and gets ranges Ri=[ri(s), ri(e)] covered by tree Γi, sorted according to their range start point ri(s). The client chooses two trees Γj, Γj+1 that are to be merged. In some examples, it can be assumed that Γj has greater height. Consequently, Γj is extended to cover [rj(s), rj+1(s)−1] using Algorithm 5. Subsequently, Γj and Γj+1 can be merged using Algorithm 6, and the number of different trees is reduced by one. This is done repeatedly until exactly one search tree is left covering the complete queried range.
Implementations of the present disclosure are described in further detail herein with reference to an example runtime (e.g., execution). As a starting point in the example runtime, it is assumed that a range has not been queried multiple times so far. As a result, every token contains new information the server can use for updating index γ. Given a value domain with D elements and n indexed items, there exist
different coherent ranges that can be queried. In some examples, after D2 different queries, all possible ranges have been queried and γ includes of exactly one tree containing all possible ranges.
In this state, any repeated range query can be answered in logarithmic time. However, assuming that there are repeated queries before the search index γ contains exactly one tree, such repeated queries may raise problems. Furthermore, these repeated queries do not contain new information, so the server is not able to update index γ. As a result, there are search patterns that result in linear search time: First, O(n) different, not coherent ranges are queried and indexed
Now these ranges are repeatedly queried—in average half of all indexed queries are checked before an answer is provided.
In accordance with implementations of the present disclosure, a cache is used for already queried ranges. In this manner, the search time for such cases is reduced. In some examples, using a hash table keyed with deterministic range identifiers (e.g., Π2=(Gen, Enc, Dec) is a deterministic encryption that is part of every search token), the search time is reduced for repeated range queries to constant time O(1).
The runtime for one search operation is the sum of the actual search time ts and the update time tu. The height of the tree is bound by log(D) and the size of an operation on one predicate-encrypted ciphertext is also O(log(D)). Consequently, merging two trees, extending one tree, refining one tree or rebalancing one tree can be done in O(log2(D)), and r trees can be merged in O(r·log2 (D)). Furthermore, because any update operation adds at least one new boundary element, there can be at most n trees. As a result, the expected update time is bounded by tu=O(n·log2 (D)).
In some implementations, search time depends on the newly queried range Q (i.e., if the newly queried range Q is completely covered by exactly one tree). The probability of this is denoted by Pr[Q⊆Γi]. If this is the case, search can be performed in O(log2(D)), because searching one tree is sufficient for learning the result set. Otherwise, the complete point list is scanned and potentially updated, resulting in search time of O(n log2(D)). As a result, the expected search time can be provided as:
ts=Pr[Q⊆Γi]·O(log2(D))+(1−Pr[Q⊆Γi])·O(n log2(D)).
In some examples, any time a range is not completely covered by a single tree, at least one element in D is added to a search tree. Accordingly, the size of the set Γi increases by at least 1. Consequently, a search complexity of O(n log2 (D)) is conducted n times at most. The maximum total time spent for these searches is n·n log2 (D). In some examples, this time can be amortized over the events Q⊆Γi. In some examples, x is the total number of searches until amortization occurs, where:
In accordance with implementations of the present disclosure, after n2 searches, at the latest, poly-logarithmic search time is achieved on average.
Security of the SRQ-scheme of the present disclosure is described in further detail with respect to the security of encrypted files using SRQ-Enc, and the security of tokenized queries using SRQ-Token. The following example theorem is provided:
Proof: The security game described above (0 of Definition 4) is modified to show the computational indistinguishability of the original game and the modified game (
1). This modified game is as follows: the only step the challenger does not perform as in the original game is the challenge step. Instead, in the modified security game, the challenger responds c=(c1, c2), where c1=RPE-Enc(mk1,
b) and c2=EncIND-CPA(mk2, 0|ƒ
0 bis provided reaking IND-CPA security using distinguisher
between
0 and
1.
0 gets two values
0,
1 and creates a key mk1=RPE-Setup (1λ, [0, D−1]). All token queries of
are answered directly by
0 using RPE-Token and mk1. For all ciphertext queries (zi, ƒi) sent by the distinguisher,
0 asks its encryption oracle for an (EncIND-CPA) encryption c2 of ƒi, creates c1=RPE-Enc(mk1, zi), and sends c=(c1, c2) to
. On challenge (t0, ƒ0) and (t1,ƒ1) sent by the distinguisher,
0 flips a coin b←{0,1}, creates c1=RPE-Enc(mk1, tb) and submits ƒb and 0|ƒ
0 sends (c1, c*) to
. If b*=0 the view of
is the same as in
0, if b*=1 the view is the same as in
1. Consequently, due to IND-CPA security of Π, the probability of differentiating
0 and
1 is negligible.
An attacker 1 can be provided on range predicate only encryption plaintext privacy of the used RPE scheme using an attacker
that wins
1 with probability ∈.
1 gets two values
0,
1 output by
and creates a key mk2=GenIND-CPA(1λ).
1 wishes to be challenged on these values
0,
1, and thus outputs them to the challenger and sends back the answers. In the query phases,
1 forwards all token queries of
to its own challenger. For all ciphertext queries (zi, ƒi)
1 creates c2=EncIND-CPA(mk2, ƒi), asks its own challenger for c1=RPE-Enc(mk1, zi), and sends back c=(c1, c2). On challenge with files ƒ0, ƒ1,
1 asks its own challenger for a challenge (i.e., the challenger flips a coin b* and sends c*=RPE-Enc(mk1,
b*) to
1.
1 creates {tilde over (c)}2=EncIND-CPA(mk2, 0|ƒ
. The guess b output by
is also returned by
1. It is clear that the probability of
1 winning the range predicate game is the same as the probability of
winning
1, which is negligible. Therefore, the advantage of
winning
0 is also negligible.
Before a security proof according Definition 5 is provided, example leakage functions 1,
2 are defined as follows:
1(M)=((ID(mi),len(ƒi)))i∈[1,ƒ]
2(Q)=(IDQ=(IDQ
where RR(Q) is a q×q range relation matrix, each element is in the set {Ø, ∩, =, ⊂, ⊂=, ⊃, ⊃=}. Here an element in row i and column j indicates the relation of ranges Qi and Qj given in queries i and j. Ø denotes no intersection, = denotes the equality of two ranges, ∩ denotes a intersection but no range is a subrange of the other. ⊂ denotes that range Qi is a subset of Qj but no limiting points are in common, ⊂= denotes a subset relation with one limiting point in common, and the other way round ⊃ denotes that range Qi is a superset of Qj, i.e., if ⊂ is at position (i, j) than ⊃ is at position (j, i).
In Definition 4 above, the challenger only accepts challenges 0,
1 that both occur in the same subset of the access pattern. For example, if file ƒi indexed under
i is in IDQ
1-i is also in IDQ
to win the security game.
In some examples, these restrictions are removed by giving the simulator access to this information in the form of the access pattern and the range relation matrix. This can be used to show security of a real protocol run, where fulfilling the restrictions of the security games cannot be guaranteed. On the other hand, given two range token sequences with the same range relation matrix (for their ranges), no attacker can distinguish between these range token sequences.
Theorem 2: Given SRQ that is built upon a RPE-scheme with selective secure plaintext privacy and selective secure predicate privacy, a domain [0, D−1], two query sequences (Q1, . . . , Qn)=Q≠R=(R1, . . . , Rn) with Qi ⊂[0, D−1], Ri ⊂[0, D−1] and RR(Q)=RR(R), the probability for any mk←SRQ-Setup(1λ, [0, D−1]), TKQ=(SRQ-Token(mk, q1), . . . , SRQ-Token(mk, qn)) and TKR=(SRQ-Token(mk, r1), . . . , SRQ-Token(mk, rn)), provided as:
|Pr[(TKQ)=1]−Pr[
(TKR)=1]|
is negligible for any distinguisher .
Proof: Denote ∈Π as the probability of an attacker breaking the used IND-CPA secure encryption scheme, denote ε1 as the probability of an attacker
winning the RPE-plaintext privacy security game, and ε2 as the probability of an attacker
winning the RPE-predicate privacy game. Given negligible εΠ, ε1 and ε2 it is possible to extend, shrink and move the ranges, such that the probability of any attacker
to distinguish between a token τQ and token τ{tilde over (Q)} that is a extended, shrunk or moved version of Q is negligible.
First, given a range token TQ=(cQ(0), cQ(1), tkQ, cQ), it is possible to extend range Q to range {tilde over (Q)} as long as there is no other range R, for which a token τR is known, with R∩Q=Ø but R∩{tilde over (Q)}≠Ø. In a first step, it can be assumed that no such range R exists. A series of games can be provided to show that the probability of any attacker to distinguish two games is negligible.
In 0, the original token τQ is given.
In 1, replace cQ with a new encryption c{tilde over (Q)}=EncIND-CPA({tilde over (Q)}). Attacker
can distinguish between
0 and
1 with probability εΠ.
In 2, replace tkQ with a new RPE-token tk{tilde over (Q)}=RPE-Token({tilde over (Q)}). Note, that q(s)∈{tilde over (Q)} and q(e)∈{tilde over (Q)} still holds. Hence, attacker
can distinguish between
1 and
2 with probability ε2.
In 3, move the limiting point cQ(i) that encrypts q(e): Replace cQ(i)=RPE-Enc(q(e)) with
=RPE-Enc(
). Attacker
can distinguish between
2 and
3 with probability ε1.
After 3, a valid token τ{tilde over (Q)} is provided for the new range {tilde over (Q)}. Putting it altogether, attacker
can distinguish between these tokens with probability:
{tilde over (ε)}=εΠ+ε2+ε1
Shrinking a range Q to a range {tilde over (Q)} can be done in a similar way, as long as there is no other range R, for which token τR is known, with R∩Q≠Ø but R∩{tilde over (Q)}=Ø. In some examples, 3 and
2 are swapped. As a result, attacker
can distinguish between a token τQ and a token for a shrunk range τ{tilde over (Q)} with probability:
εΠ+ε1+ε2={tilde over (ε)}
Combining these techniques, a range Q=[q(s), q(e)] can move to a new range {tilde over (Q)}=[q(s)+x, q(e)+x], as long as there is no other range R with r(s)>q(s) but r(s)<(q(s)+x) (otherwise, this range R must be moved before). First, extend Q to a range Q′=[q(s), q(e)+x], then shrink Q′ to range {tilde over (Q)}=[q(s)+x, q(e)+x].
Without loss of generality, it can be assumed that maxQ
Given this theorem, the security of SRQ-scheme of the present disclosure can be more formal defined using leakage based Definition 5, provided above.
Theorem 3: If the SRQ-scheme has selective secure plaintext privacy based on an RPE-scheme with selective secure predicate privacy and Π1 is an IND-CPA secure encryption scheme, then the SRQ-scheme is (1,
2)-secure against non-adaptive chosen-range attacks.
Proof: A PPT simulator is provided, for which the advantage of any PPT adversary
to distinguish between the Rea
and Idea
experiments from Definition 5 is negligible. For this,
can be described as setting up the environment and simulating range tokens
and ciphertexts {tilde over (C)} using leakage
1 and
2.
Setting up the environment: internally runs SRQ-Setup (1λ, [0, D−1]) and receives a master key mk.
Simulating :
extracts clusters of ranges that form one big coherent range. Algorithm 7 provides a formal description.
Each cluster is a separate R-Tree in the implementation described herein. For every cluster, simulates ranges with the same range relation matrix as the actual given range relation matrix RR(Q). In more detail, for every cluster, simulator
transforms the range relation matrix RR (Q) into a linear program that is solved. Every relation is formulated as in equations. Doing this for all clusters,
gets simulated ranges {tilde over (Q)} with RR(Q)=RR({tilde over (Q)}). Now
sets
=(SRQ-Token(mk,
)i∈[1,q]) which is indistinguishable by Theorem 2. Note that
can restore the simulated range
given a range token SRQ-Token(mk,
), because a component includes an ordinary IND-CPA encrypted value that can be decrypted.
Simulating {tilde over (C)}: Simulator creates a set of leafs L. More particularly,
divides IDQ in a set L consisting of disjoint sets, where L covers the same values as IDQ. Two sets IDQ
stores the indexes of all range queries that contain Li as result set: L(i)={j|Li⊆IDQ
IDQ
can simulate the ciphertexts {tilde over (C)}=(
, . . . ,
) as follows:
iterates over all tuples (ID(ƒi), len (ƒi)) and:
Due to IND-CPA security of Π1, selective secure plaintext privacy of SRQ and Theorem 2, the probability for to distinguish between C and {tilde over (C)} generated by
is negligible.
It is possible for to simulate range queries {tilde over (Q)} from given leakage
2 (Q). Simulator
is able to simulate all update protocols on these tokens
. Because decrypting range token τ
is possible for the simulator,
can run all update queries on the simulated ranges {tilde over (Q)}. Note that these update protocols do not contain new information, but all information is already covered by
1(M) and
2(Q).
A range query token is received (402). For example, a range query token is received by a server-side computing device from a client-side computing device. In some examples, the range query token is encrypted. It is determined whether a tree list is empty (404). In some examples, an encrypted search index stored at the server-side includes the tree list and a point list. In some examples, the tree list is empty, if the tree list does not include at least one search tree. In some examples, a search tree is associated with an already searched range (e.g., a range resulting from a previously submitted range query token). In some examples, the point list includes a list of indexed points of ciphertext that is to be searched based on received queries (e.g., range queries). If the tree list is empty, an encrypted query result is determined based on the point list (406). For example, the server-side computing device queries the encrypted data (ciphertext) based on indexed points provided in the point list to provide the encrypted query result (e.g., a range).
If the tree list is not empty, it is determined whether the range query token corresponds to a previously submitted range query token of a search tree in the tree list (410). In some examples, if the range query token is a sub-range of, or is equal to a previously submitted range query token of a search tree in the tree list, it is determined that the range query token corresponds to the range query token of the search tree in the tree list. In some examples, if the range query token does not intersect a previously submitted range query token of a search tree in the tree list, the range query token only partially intersects a previously submitted range query token of a search tree in the tree list, or values of the range query token correspond to values of multiple search trees in the tree list, it is determined that the range query token does not correspond to a range of a search tree in the tree list.
If it is determined that the range query token does not correspond to a range of a search tree in the tree list, an encrypted query result is determined based on the point list (406). If it is determined that the range query token does correspond to a range of a search tree in the tree list, an encrypted query result is determined based on a search tree (412). For example, the encrypted query result is based on value points indexed by the search tree.
The encrypted search index is updated (414). In some examples, the encrypted search index is updated based on the range query token and a client-server protocol. In some examples, the encrypted search index is updated by adding a search tree to the tree list, refining a search tree (e.g., Algorithm 4), extending a search tree (e.g., Algorithm 5), merging search trees (e.g., Algorithm 6), and/or rebalancing a search tree (Algorithm 3). Encrypted query results are returned (416). For example, the server-side computing device transmits the encrypted query results to the client-side computing device.
Referring now to
The memory 520 stores information within the system 500. In one implementation, the memory 520 is a computer-readable medium. In one implementation, the memory 520 is a volatile memory unit. In another implementation, the memory 520 is a non-volatile memory unit. The storage device 530 is capable of providing mass storage for the system 500. In one implementation, the storage device 530 is a computer-readable medium. In various different implementations, the storage device 530 may be a floppy disk device, a hard disk device, an optical disk device, or a tape device. The input/output device 540 provides input/output operations for the system 500. In one implementation, the input/output device 540 includes a keyboard and/or pointing device. In another implementation, the input/output device 540 includes a display unit for displaying graphical user interfaces.
The features described can be implemented in digital electronic circuitry, or in computer hardware, firmware, software, or in combinations of them. The apparatus can be implemented in a computer program product tangibly embodied in an information carrier, e.g., in a machine-readable storage device, for execution by a programmable processor; and method steps can be performed by a programmable processor executing a program of instructions to perform functions of the described implementations by operating on input data and generating output. The described features can be implemented advantageously in one or more computer programs that are executable on a programmable system including at least one programmable processor coupled to receive data and instructions from, and to transmit data and instructions to, a data storage system, at least one input device, and at least one output device. A computer program is a set of instructions that can be used, directly or indirectly, in a computer to perform a certain activity or bring about a certain result. A computer program can be written in any form of programming language, including compiled or interpreted languages, and it can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.
Suitable processors for the execution of a program of instructions include, by way of example, both general and special purpose microprocessors, and the sole processor or one of multiple processors of any kind of computer. Generally, a processor will receive instructions and data from a read-only memory or a random access memory or both. Elements of a computer can include a processor for executing instructions and one or more memories for storing instructions and data. Generally, a computer can also include, or be operatively coupled to communicate with, one or more mass storage devices for storing data files; such devices include magnetic disks, such as internal hard disks and removable disks; magneto-optical disks; and optical disks. Storage devices suitable for tangibly embodying computer program instructions and data include all forms of non-volatile memory, including by way of example semiconductor memory devices, such as EPROM, EEPROM, and flash memory devices; magnetic disks such as internal hard disks and removable disks; magneto-optical disks; and CD-ROM and DVD-ROM disks. The processor and the memory can be supplemented by, or incorporated in, ASICs (application-specific integrated circuits).
To provide for interaction with a user, the features can be implemented on a computer having a display device such as a CRT (cathode ray tube) or LCD (liquid crystal display) monitor for displaying information to the user and a keyboard and a pointing device such as a mouse or a trackball by which the user can provide input to the computer.
The features can be implemented in a computer system that includes a back-end component, such as a data server, or that includes a middleware component, such as an application server or an Internet server, or that includes a front-end component, such as a client computer having a graphical user interface or an Internet browser, or any combination of them. The components of the system can be connected by any form or medium of digital data communication such as a communication network. Examples of communication networks include (e.g., a LAN, a WAN), and the computers and networks forming the Internet.
The computer system can include clients and servers. A client and server are generally remote from each other and typically interact through a network, such as the described one. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.
In addition, the logic flows depicted in the figures do not require the particular order shown, or sequential order, to achieve desirable results. In addition, other steps may be provided, or steps may be eliminated, from the described flows, and other components may be added to, or removed from, the described systems. Accordingly, other implementations are within the scope of the following claims.
A number of implementations of the present disclosure have been described. Nevertheless, it will be understood that various modifications may be made without departing from the spirit and scope of the present disclosure. Accordingly, other implementations are within the scope of the following claims.
| Number | Name | Date | Kind |
|---|---|---|---|
| 7395437 | Agrawal et al. | Jul 2008 | B2 |
| 8832427 | Roeder et al. | Sep 2014 | B2 |
| 8856138 | Neubeck | Oct 2014 | B1 |
| 20040243816 | Hacigumus et al. | Dec 2004 | A1 |
| 20090171916 | Kirk et al. | Jul 2009 | A1 |
| 20100106980 | Kerschbaum et al. | Apr 2010 | A1 |
| 20120078914 | Roeder | Mar 2012 | A1 |
| 20130046974 | Kamara | Feb 2013 | A1 |
| 20130262852 | Roeder | Oct 2013 | A1 |
| Number | Date | Country |
|---|---|---|
| 104408177 | Mar 2015 | CN |
| 104765848 | Jul 2015 | CN |
| 2199935 | Jun 2010 | EP |
| Entry |
|---|
| Communication and extended European Search Report from European Patent Office re Appl. No. 16002387.5-1870; dated Jan. 18, 2017; 6 pages. |
| Bijit Hore et al. “A Privacy-Preserving Index for Range Queries”, Proceedings of the Thirtieth International Conference on Very Large Data Bases, Jan. 1, 2004, pp. 720-731, XP009133028. |
| Peng Wang et al. “Secure and Efficient Range Queries on Outsourced Databases Using Rp-trees”, Data Engineering (ICDE), 2013 IEEE 29th International Conference on, IEEE, Apr. 8, 2013, pp. 314-325, XP032430875. |
| Agrawal et al. “Order Preserving Encryption for Numeric Data,” In Proceedings of the 2004 ACM SIGMOD international conference on management of data, ACM, 2004, pp. 563-574. |
| Bellare et al., “Deterministic and Efficiently Searchable Encryption,” In Proceedings of Crypto, 2007, 535-552. |
| Boldyreva et al., “Order-Preserving Symmetric Encryption,” In Advances in Cryptology (Eurocrypt), 2009, 24 pages. |
| Boldyreva et al., “Order-Preserving Encryption Revisited: Improved Security Analysis and Alternative Solutions,” Advances in Cryptology, Crypto 2011, Computer Science. Springer Berlin Heidelberg, 2011, 36 pages. |
| Boneh et al., “Public Key Encryption with keyword Search,” In Advances in Cryptology—Eurocrypt 2004, pp. 506-522. |
| Boneh et al., “Functional Encryption: Definitions and Challenges,” vol. 6597 of the series Lecture Notes in Computer Science, 2011, pp. 253-273. |
| Boneh et al., “Conjunctive, Subset, and Range Queries on Encrypted Data,” S.P. Vadhan (Ed.): TCC 2007, LNCS 4392, pp. 535-554. |
| Cash et al., Highly-Scalable Searchable Symmetric Encryption with Support for Boolean Queries, Crypto 2013, Part I, LNCS 8042, 2013, pp. 353-373. |
| Chang et al., “Privacy Preserving Keyword Searches on Remote Encrypted Data,” Applied Cryptography and Network Security, 2004, pp. 442-455. |
| Curtmola et al., “Searchable Symmetric Encryption: Improved Definitions and Efficient Constructions,” In Proceedings of ACM CCS'06, 2006, 10 pages. |
| Goh, “Secure Indexes,” Cryptology ePrint Archive Report 2003/216, 2003, 19 pages. |
| Golle et al., “Secure Conjunctive Keyword Search Over Encrypted Data,” In Applied Cryptography and Network Security, ACNS 2004, LNCS 3089, 2004, pp. 31-45. |
| Guttman, “R-Trees: A Dynamic Index Structure for Spatial Searching,” vol. 14. ACM, 1984, pp. 47-57. |
| Hacigumus et al., “Executing SQL over Encrypted Data in the Database-Service-Provider Model,” In Proceedings of ACM SIGMOD'02, 2002, 12 pages. |
| Hore et al., “A Privacy-Preserving Index for Range Queries,” In Proceedings of VLDB'04, 2004, 720-731. |
| Kamara et al., “Parallel and Dynamic Searchable Symmetric Encryption,” In Financial Cryptography and Data Security, pp. 258-274. |
| Kamara et al., “Dynamic Searchable Symmetric Encryption,” In Proceedings of the 2012 ACM conference on Computer and communications security, 2012, pp. 965-976. |
| Katz et al., “Predicate Encryption Supporting Disjunctions, Polynomial Equations, and Inner Products,” In Advances in Cryptology—Eurocrypt 2008, pp. 146-162. |
| Lu, “Privacy-Preserving Logarithmic-Time Search on Encrypted Data in Cloud,” In NDSS'11, Feb. 6, 2011, 15 pages. |
| Naveed et al., “Dynamic Searchable Encryption via Blind Storage,” IEEE Symposium on Security and Privacy (SP), May 2014, pp. 639-654. |
| Popa et al., “CryptDB: Protecting Confidentiality with Encrypted Query Processing,” In Proceedings of the 23rd ACM Symposium on Operating Systems Principles (SOSP), 2011, 16 pages. |
| Shen et al., “Predicate Privacy in Encryption Systems,” In Theory of Cryptography, LNCS 5444, 2009, pp. 457-473. |
| Shi et al., “Multi-Dimensional Range Query Over Encrypted Data,” In Security and Privacy, 2007. SP'07, IEEE Symposium on Security and Privacy, 2007, pp. 350-364. |
| Song et al., “Practical Techniques for Searches on Encrypted Data,” In Proceedings of IEEE Symposium on Security and Privacy, 2000, 12 pages. |
| CN Office Action issued in Chinese Application No. 201610911555.0 dated Apr. 1, 2021, 16 pages (With English Translation). |
| Lu, “Privacy-preserving Logarithmic-time Search on Encrypted Data in Cloud.” NDSS, Feb. 2012, 17 pages. |
| Number | Date | Country | |
|---|---|---|---|
| 20170139985 A1 | May 2017 | US |
