Patent Grant
8312276
The present invention generally relates to a network reputation system and its controlling method thereof.
The Internet has transformed the world into a global marketplace. For example, a person anywhere in the world may buy or sell goods to anyone else. Given such convenience, online applications such as auctions, multi-player games and social networks engender a huge potential market and consequently attract a lot of attention. Unfortunately, malicious behavior, such as account hijacking and non-delivery of goods in online auctions, introduces risks.
To prevent such risks, online applications employ reputation systems that rely on user-assigned scores to judge on other users' legitimate behavior. For example, a buyer in an online auction system may use the seller's score of the reputation system to judge whether the seller is trustworthy. However, a serious problem of existing reputation systems is the simplicity to create a highly trustworthy score by a seller himself or a seller's friend, or vice versa, create an untrustworthy score for a competitor. Since adversaries may apply several accounts to complete several faked transactions, foiling existing reputation systems is often straightforward.
Many approaches for addressing these challenges have been discussed. For example, Zhang et al. and Pandit et al. use a transaction network model to analyze auction fraud. In their model, they establish an undirected transaction graph where each node denotes a registered user and each edge indicates a completed transaction between the two users. By cooperating with other mathematical models and algorithms, they may identify potentially fraudulent users. Kuo et al. use reputation network constructed from buyer's feedbacks to identify fraudulent users.
Pujol et al. and Sabater et al. extract reputation by means of social network. Also, Hogg et al. summarize the reputation systems constructed via online social network. The online social network is automatically constructed from any observable internet interaction, such as the referred web links and emails in the homepages, and the relationships established in the social networking websites (e.g., facebook). These online relationships contribute online social network. This type of reputation system may enhance the effectiveness of the rating. Swamynathan et al. also shows that this type of reputation system is more trustful and satisfied than feedback reputation systems. Besides that, the notion to evaluate a user using social network may be applicable to other applications, such as Voice over IP (VoIP), Internet Message (IM), peer to Peer (P2P) and Reliable Email (RE). RE proposed by Freeman et al. considers privacy issue.
Although reputation systems have been proposed for solving these problems, most reputation systems in applications are still unreliable. Reputation systems also pose some privacy problems. For example, they may reveal private information of users. They would not achieve real reliability or being trustworthy if the online social relationship is not related to personal social relationship. Also, a forging problem may occur if an online social relationship is also not cryptographically verifiable. On the other hand, most of these systems reveal private personal information.
Social network represents relationships within a community. Several types of social network may be established according to different social relationship such as kinship, friendship, cooperation, etc.
Recently, research on private matching receives a lot of attention and several schemes and applications are proposed. Assuming there are two databases A and B, one query QεA and one matching protocol computes P=Q∩B. The scheme is secure and preserving privacy if it satisfies the following requirements. (1) Privacy: Each party can know only P and it's input to the matching protocol. Except for this information, each party learns nothing. (2) Non-spoofable: Items in A and B are really authorized by item owners. This means that the user may make query Q only if the owners of these items authorize and give the user these items. In other words, the user can not generate the queried items without authorization of the item owners. Besides that, the user should have some proofs to demonstrate that the item owner authorizes the user.
In Hash Protocol (HP), a person, who wants to query the common items in the other's database, computes hash values of items in his own database. Then he and the target exchange these hash values. By this way, they may find the common items without revealing the information of the un-matched items. On the other hand, Agrawal et al. proposed AgES which uses commutative encryption to achieve private matching. Freedman et al. proposed a polynomial-based private matching scheme. They use the property of homomorphic encryption to achieve better privacy. A variant of their scheme, set cardinality private matching, let databases A know only the cardinality of Q∩B, not the actual items in this set. After that, Kissner and Song extend FNP scheme to support more functionality. HP, AgES and Freedman et al's schemes are categorized to asymmetric exchange of information, different from symmetric exchange in which both parties know the same information in the protocols.
Besides those, Li et al. proposed Data Ownership Certificate (DoC) to ensure non-spoofable. If the user does not obtain the item and the corresponding DoC, he can not make the query and convince the other.
The exemplary embodiments of present invention may provide a network reputation system and its controlling method thereof.
An exemplary embodiment relates to a network reputation system. The network reputation system is in conjunction with one or more estimator's social networks and one estimatee's reputation network, which may comprise: a credentials and exchange component that permits a user to generate credentials and exchange matching items with those persons having a social relationship with the user; a reputation evaluation component that enables other users to make evaluations about an estimatee via the sharing of social network information; and a query and response component that receives a query from a person having a social relationship with the user for requesting an evaluation about the estimatee, and responds with an associated evaluation result to the person having a social relationship with the user, via the sharing of social network information and the evaluations made by the other users about the estimatee.
Another exemplary embodiment relates to a controlling method for the network reputation system. The controlling method may comprise: a user producing master social network information via the user's social network; the user sharing social network information with those persons having a social relationship with the user through the conversion of the master social network information; the user making an evaluation about an estimatee via said social network information; a person having a social relationship with the user requesting an application server for the evaluation about the estimatee; and the application server replying an associated evaluation result about the estimatee to the person having a social relationship with the user, at least including the evaluation value evaluated by the user.
The foregoing and other features, aspects and advantages of the present invention will become better understood from a careful reading of a detailed description provided herein below with appropriate reference to the accompanying drawings.
The disclosed exemplary embodiments may provide a network reputation system and its controlling method thereof, which uses at least a social network model on a transaction model to prevent reputation fraud. A user may judge an estimatee by the scores given by the user's personal social network. The present techniques may enable users to recognize evaluations submitted by their friends as these are more trustworthy. The mechanisms may preserve the privacy of the friendship relationships from the website, and the privacy of the mapping between the real identity of a user and her online identity from other users including her friends. For example, the present techniques may apply private matching methodologies in the system to prevent the disclosure of private social relationships.
In the disclosure, some assumptions are made, for example, assume that the estimator (say E) is part of a social network modeled as a graph G=(V, E), where the vertices V denote the users of the system, and E denotes the relationships of the users. A user trusts all her contacts. Social network represents relationships within a community. Several types of social network may be established according to different social relationship, such as kinship, friendship, cooperation, etc. In the reputation system, social network may be built according to aforementioned relationships except trading relationship. The trading relationship is employed to build another network, which is called reputation network.
In the disclosure, it also assumes that the existence of an application server (AS) to which users register using a service identifier (SID). Online auction websites such as E-BAY and YAHOO, and online gaming sites may play the role of such an application server. The application server maintains for each dealer, the evaluations that the users submitted for her. Every user in the network owns a secret-public key pair (SK, PK). It also assumes the existence of secure and authenticated channels between a user and her friends, and between a user and an application server. However, it is not necessary to assume that the existence of a global public key infrastructure (PKI).
The disclosed exemplary embodiments of the present invention enable an estimator to recognize evaluations that were submitted for the estimatee to the application server by her friends, in a privacy preserving fashion as illustrated in
In order to satisfy the desired properties, the protocol according to the present invention applies asymmetric HP to protect privacy due to its efficiency and security. The protocol runs in four phases, i.e. Initialization (Phase 1), Credential Exchange (Phase 2), Reputation Evaluation (Phase 3), and Query and Response (Phase 4).
To run the four phases under a network reputation system, an exemplary embodiment of its controlling method may be illustrated as in
In step 340, the request for the evaluation about the estimatee may be derived from a formula P of social private matching query, which may utilize the social network information in the reputation system model. In addition, the application server may be requested to reply the evaluation with signature thereon. In step 350, the replied evaluation of the estimatee from the application sever may be obtained from another formula Q of social private matching response, which may utilize the formula P. Moreover, all of the replied evaluation of the estimatee from the application sever may have signature thereon. The followings describe detailed operations for each phase under a reputation system on networks, consistent with certain disclosed embodiments of the present invention.
In phase 1 (Initialization), each user produces his/her own parameters (i.e. master social network information). For the example shown in
In phase 2 (Credential Exchange), each user sends a triple to each of her friends via a secure and authenticated channel. For example, user U sends the triple <pid, pk, item> to each of her friends via a secure and authenticated channel. In practice, the credentials may be exchanged via secure email, Bluetooth, SMS, etc. For the exemplary matching item exchange shown in
In phase 3 (Reputation Evaluation), assume a user U_J had a transaction with a user U_S, and wants to submit an evaluation about the user U_S to the application server. User U_J sends the evaluation triple <sid_S, item_J, Evalu_S_J> to the application server, where Evalu_S_J=sid_S∥scr_S_J∥Sig_sk_J(E_Digst), Sig_sk_J denotes signature using signing key sk_J, E_Digst is defined as a hash function of (sid_S∥item_J∥scr_S_J), and scr_S_J is the score from user U_J to user U_S.
Phase 4 (Query and Response) runs when an estimator U_I wants to recognize the evaluations submitted to the application server by her friends about an estimatee U_S. Assume estimator U_I has a set X={x—1, x—2, . . . , x_ki}, where x_ is an item received by estimator U_I from one of her friends during phase 2, and application server has a set Y={y—1, y—2, . . . , y_ks} where y_is the item of one of the users that submitted an evaluation for user U_S. Note that X is a set of 1-distance items. In other words, the holder of the items is a friend of estimator U_I.
In phase 4, the communication flow between estimator U_I and the application server may include the following steps. (1) User U_I sends a query concatenated with sid_I (service ID) to the application server to request the evaluation submitted to the application server by the friends of User U_I about the estimatee U_S. (2) The application server computes an encryption function H, for example, H may be the concatenation of a plurality of Hash functions, such as H=h—1∥h—2∥ . . . ∥h_ks, where h_I=Hash(y_I), 1≦I≦ks. (3) The application server sends the encryption function H concatenated with a plurality of the evaluations back to the estimator U_I.
For example, the produced evaluation sent from the application server may be defined as follows.
H∥Evalu_S—1∥Evalu_S—2∥ . . . ∥Evalu_S_ks
By using it, only the person who owns/knows the item y_I can decrypt the value Evalu_S_I and know the score inside Evalu_S_I. (4) Estimator U_I compares Hash(x_J), 1≦J≦ki, with h_L, 1≦L≦ks. If h_M is matched, estimator U_I verifies the associated Evalu_S_M by using the public key pk_M which is obtained in phase 2. As a result, estimator U_I learns the scores given by his/her friend and then makes a further decision. However, if none is matched, it means estimator U_I has no friend who has estimation on estimatee U_S. In this case, estimator U_I may learn the scores in Evalu_S_S although it may not be trustworthy.
In other words, evaluation result about the estimatee may be obtained via a concatenation of a plurality of parameters, at least including a service identity of the estimatee, an authentication from other users, and a score evaluated by other users on the estimatee. In addition, the authentication may be through a channel of using signing key with an encrypted function, such as hash function, of the service identity of the estimatee, each matching item of each of other users and the score evaluated by the other users on the estimatee.
According to the present invention, some schema may be taken to increase the matching rate in the protocol of the disclosed exemplary embodiments. One exemplary scheme is to accept more social relationships. If more social relationships are expanded, then there are more items for matching. For example, people working in the same organization may exchange their items. Then, the matching rate may be increased. Another exemplary scheme is to use the items having greater distance. Since the estimator has more items for matching, the matching rate obviously increases.
In practice, estimator U_I may upload the request and download the result via an application connecting to the server. The FNP scheme disclosed in “Efficient Private Matching and Set Intersection” may be chosen over the other private matching protocols because of its flexibility and extensibility. The extension schemes will be discussed in the followings.
The present invention may be extended to achieving more features. For example, the first feature is the present invention may provide variants of private matching algorithm to protect some privacy-sensitive transaction, such as set cardinality private matching which let a user know only the number, instead of who, of his/her friends that had reputation on estimates U_S. In this case, the application server does not send encrypted Evalu_S_I to estimator U_I for protecting the privacy of the items those link identities.
The second feature is the present invention may provide query by using n-distance items. For example, n=2, i.e., if the estimator is a friend's friend, it is possible to match his/her reputation record without revealing any private information. For example, assume Andy's friend F receives the item FF (i.e. the item B in
In the real world, the strength of relationship may not be estimated by just “YES” or “NO”, that means having relationship or not. Different relationships may give different weights, which denote the strength of trust. For example, kinship often has higher weight than friendship. Of course, users may be able to decide the weights.
The third feature is the present invention may provide group-based matching. One may choose to generate two or more items and send one or more of these items to their different friends. The items may be for public, group, personal or private usage which is used in different purposes depending on how private they are or how public they are desired to be revealed. For example, for trades for public use and trades for private use, different items may be used. This may prevent private transaction information from disclosing to unclose friends such as co-workers.
The followings give some working examples to illustrate the controlling method for a network reputation system of the present invention.
The first working example may be described as follows. (1) User B's friend or friend's friend A makes an evaluation about an estimatee S. (2) User B derives a formula P by using the social network information of A. (3) User B requests an application server for the evaluation about estimatee S made by A. (4) Through the social network information of A, user B gets the evaluation about the estimatee S from the application server, and verifies its correctness.
Referring to
In the communication protocol of the first working example, such as private matching scheme may be used to achieve the privacy preserving for both user identity and social network information. It may also use a way of signature to achieve the authentication. In other words, private matching and signature are bounded together in the communication protocol.
The first working example gives some features, such as (1) friend B may learn the evaluation value submitted to the application server by only the user A about the estimatee S, but may not learn the evaluation value about the estimatee S from others, this will protect other strangers' privacy; (2) the application server learns nothing about the social network of user A or friend B, therefore, the application server learns nothing about the relationship between user A and friend B; (3) friend B may verify the correctness of the evaluation value Evalu_S_A via the formula Q and the social network information, and guarantee the contents of the evaluation value not being changed; and (4) no one may forge any data successfully.
The second working example may be described as follows. (1) User B's friend or friend's friend A makes an evaluation about an estimatee S. (2) User B requests an application server for the evaluation of the estimatee S made by A. (3) The application server replies with all evaluations about the estimatee S. (4) User B gets the evaluation about the estimatee S from the application server by using A's social network information, and verifies its correctness.
Referring to
In the communication protocol of the second working example, such as sharing bridge identity may be used to achieve the privacy preserving for user identity. It may also use a way of signature to achieve the authentication. The application server may sends (Bridge ID, Signature)/Bridge ID contained in the entire or partial evaluation values to friend B to protect the privacy of social information. In practice, some designs may be employed as follows, during the initialization, the evaluation submission and the evaluation retrieval. (1) During the initialization, a user may use a pseudonym to register at the review site. When exchange credentials with friends, the user may use real name, and the public keys are exchanged with and only given to friends. However, a Bridge ID for the user may be given to both the review site and friends. (2) During the evaluation submission, a user may submit a signed evaluation to the review site after a transaction with a seller. The review site may store the signed evaluation. (3) During the evaluation retrieval, a user may retrieve all evaluations about the seller when considering a transaction with the seller, and distinguish reviews by friends. The privacy of friendship relationships may be preserved through the Bridge ID.
The second working example gives some features, such as (1) friend B may learn all the evaluation values submitted to the application server about the estimatee S, and the contents of the evaluation values may also be encrypted by their own private keys inside their own social network information; (2) the application server learns nothing about the social network of user A or friend B, and this will protect the user A's privacy and friend B's privacy; and (3) friend B may verify the correctness of the evaluation value Evalu_S_A via the social network information, and guarantee the contents of the evaluation value not being changed.
Compared with the second working example, the major feature of the third working example is that before the step 1050, friend B builds or adds a personal social group SG_B via A's social network information, and allows the application server learn the social group SG_B.
Referring to
Continuing now with an exemplary network reputation system by implementing the network reputation system in conjunction with one or more estimator's social networks and one estimatee's reputation network like any of those discussed above, the network reputation system may comprise a credential exchange component 1201, a reputation evaluation component 1203 and a query and response component 1205, as shown in
The query may be further sent to an application server via query and response component 1205, and the associated evaluation result may be also generated by the application server then sent to the person via query and response component 1205.
The exemplary communication protocol depicted in the first working example of
The network reputation system may be embodied in a reputation system model such as depicted in
For estimating the performance of the present invention, a simulation is performed by using maplesoft in the 2Gmhz-CPU with 1G-RAN PC environment.
In summary, the exemplary disclosed embodiments of the present invention provide a trustable network reputation system and its controlling method thereof, which may enable users to recognize evaluations submitted by those persons having social relationships with the users. The present invention may preserve the privacy of the social relationships from the website and the privacy of the mapping between the real identity of a user and her online identity from other users including her friends.
Although the present invention has been described with reference to the exemplary disclosed embodiments, it will be understood that the invention is not limited to the details described thereof. Various substitutions and modifications have been suggested in the foregoing description, and others will occur to those of ordinary skill in the art. Therefore, all such substitutions and modifications are intended to be embraced within the scope of the invention as defined in the appended claims.
This application claims the benefit of U.S. Provisional Application No. 61/150,324, filed Feb. 6, 2009, which is incorporated by reference.
| Number | Name | Date | Kind |
|---|---|---|---|
| 5870744 | Sprague | Feb 1999 | A |
| 6112186 | Bergh et al. | Aug 2000 | A |
| 6839690 | Foth et al. | Jan 2005 | B1 |
| 7143052 | LaSalle et al. | Nov 2006 | B2 |
| 7472110 | Achlioptas | Dec 2008 | B2 |
| 7707122 | Hull et al. | Apr 2010 | B2 |
| 7716079 | Hanif et al. | May 2010 | B2 |
| 7818200 | Kerschbaum et al. | Oct 2010 | B2 |
| 7877353 | Ahmed et al. | Jan 2011 | B2 |
| 7886334 | Walsh et al. | Feb 2011 | B1 |
| 8010460 | Work et al. | Aug 2011 | B2 |
| 8010619 | Lawler et al. | Aug 2011 | B1 |
| 8015484 | Backer | Sep 2011 | B2 |
| 8200819 | Chung et al. | Jun 2012 | B2 |
| 20020078152 | Boone | Jun 2002 | A1 |
| 20020103999 | Camnisch et al. | Aug 2002 | A1 |
| 20050192958 | Widjojo et al. | Sep 2005 | A1 |
| 20050216300 | Appelman et al. | Sep 2005 | A1 |
| 20060009994 | Hogg et al. | Jan 2006 | A1 |
| 20060253578 | Dixon et al. | Nov 2006 | A1 |
| 20060253584 | Dixon et al. | Nov 2006 | A1 |
| 20060294134 | Berkhim et al. | Dec 2006 | A1 |
| 20070245245 | Blue et al. | Oct 2007 | A1 |
| 20080005223 | Flake et al. | Jan 2008 | A1 |
| 20080077517 | Sappington | Mar 2008 | A1 |
| 20080120411 | Eberle | May 2008 | A1 |
| 20080133657 | Pennington | Jun 2008 | A1 |
| 20080189768 | Callahan et al. | Aug 2008 | A1 |
| 20080313723 | Naono et al. | Dec 2008 | A1 |
| 20090063630 | Obasanjo et al. | Mar 2009 | A1 |
| 20090070679 | Shen et al. | Mar 2009 | A1 |
| 20090178125 | Barber et al. | Jul 2009 | A1 |
| 20090210244 | Koister et al. | Aug 2009 | A1 |
| 20090265198 | Lester et al. | Oct 2009 | A1 |
| 20090299819 | Davis et al. | Dec 2009 | A1 |
| 20090306996 | Tsaparas et al. | Dec 2009 | A1 |
| 20090327054 | Yao et al. | Dec 2009 | A1 |
| 20090327907 | Estrada et al. | Dec 2009 | A1 |
| 20100042511 | Sundaresan et al. | Feb 2010 | A1 |
| 20100042931 | Dixon et al. | Feb 2010 | A1 |
| 20100122347 | Nadler | May 2010 | A1 |
| 20100125630 | Krishnamurthy et al. | May 2010 | A1 |
| 20100132049 | Vernal et al. | May 2010 | A1 |
| 20100199340 | Jonas et al. | Aug 2010 | A1 |
| 20100257183 | Kim et al. | Oct 2010 | A1 |
| 20110071953 | Shen et al. | Mar 2011 | A1 |
| 20110113098 | Walsh et al. | May 2011 | A1 |
| 20110161191 | Shen et al. | Jun 2011 | A1 |
| 20110219073 | Lawler et al. | Sep 2011 | A1 |
| 20110239282 | Svarfvar et al. | Sep 2011 | A1 |
| 20110314111 | Wang et al. | Dec 2011 | A1 |
| Number | Date | Country |
|---|---|---|
| 479182 | Mar 2002 | TW |
| Number | Date | Country | |
|---|---|---|---|
| 20100205430 A1 | Aug 2010 | US |
| Number | Date | Country | |
|---|---|---|---|
| 61150324 | Feb 2009 | US |
