This application claims priority to U.S. provisional patent application Ser. No. 61/078,068, filed Jul. 3, 2008, titled METHOD FOR QUANTIFYING TRUST, which is incorporated herein by reference.
This application claims priority to U.S. provisional patent application Ser. No. 61/094,861, filed Sep. 5, 2008, titled TRUST AND COLLABORATION, which is incorporated herein by reference.
As digital communications, networks, and transactions increase, the need became apparent for ways in which computer users could “trust” each other. Digital trust systems, such as Public Key Infrastructure (PKI), build trust hierarchies, such as the “Web of Trust”, so that users could securely communicate with each other, authenticate the identities of each other, and determine the integrity of the messages received from each other. In order to establish the trust characteristic of each user, trust systems rely on the certification or revocation of a user by one or more trusted third parties. However, the certification or revocation of a user does not explain what “trust” exactly is or how to quantify it. Instead, each user's trust characteristic is defined in a binary trust form consisting of 1 (trustworthy) or 0 (not trustworthy). The trust characteristics may be established through logical rules of inference from the certification or revocation actions of one or more trusted third parties. The trusted third parties may determine that a user is trustworthy if he has a certain set of credentials and if he complies with a local security policy, but this method does not allow the third party to explain why the user is trusted or how much the third party trusts him.
Described herein are implementations of various technologies for quantifying trust in a computing network. In one implementation, a computer program may be configured to establish a trust value for an agent (user) in a computing network. The trust value may be established by comparing an agent's expected behavior to his actual behavior in past transactions. The computer program may analyze the “gap” or discrepancy between the agent's expected behavior and his actual behavior to establish an initial trust value. Trust values for each agent in the computing network may be evaluated using a similar type of analysis.
After trust values are established for each agent, the computer program may form one or more trust groups, or cliques, containing agents with similar trust values in each other. Each trust group may be created such that trust values of each agent may be within a specific tolerance (“q”) of each other. If an agent's trust value is not within the specified tolerance (“q”) of other agents in a trust group, the computer program may split the trust group into one or more sub-groups such that each agent within the sub-group may have similar trust values in each other but with a smaller tolerance than that of the larger trust group. If the agent's trust value is not within the specified tolerance (“q”) or does not have a similar trust value with other agents in sub-groups, he may be rejected from the small and large trust groups.
After trust groups and sub-groups of agents have been created, new candidates, or new agents, may be granted entry into a trust group or sub-group based on an evaluation performed by each agent in the trust group or sub-group. Each agent within the trust group or sub-group may then assess their trust in the new candidate and create his own trust value for the new candidate. The trust value assigned to the candidate may be quantified into a value between zero and one. If the candidate meets the trust group's trust requirement, i.e., the trust value of the candidate is within the trust group's tolerance “q”, he may be granted access into the trust group. If the candidate does not meet the trust requirement of each agent in a trust group, he may be accepted into a sub-group or he may be rejected from the trust group altogether.
The above referenced summary section is provided to introduce a selection of concepts in a simplified form that are further described below in the detailed description section. The summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter. Furthermore, the claimed subject matter is not limited to implementations that solve any or all disadvantages noted in any part of this disclosure.
In general, one or more implementations described herein are directed to quantifying trust in an agent and grouping agents based on their trust values. One or more implementations of various techniques for quantifying trust will be described in more detail with reference to
Implementations of various technologies described herein may be operational with numerous general purpose or special purpose computing system environments or configurations. Examples of well known computing systems, environments, and/or configurations that may be suitable for use with the various technologies described herein include, but are not limited to, personal computers, server computers, hand-held or laptop devices, multiprocessor systems, microprocessor-based systems, set top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments that include any of the above systems or devices, and the like.
The various technologies described herein may be implemented in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that performs particular tasks or implement particular abstract data types. The various technologies described herein may also be implemented in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network, e.g., by hardwired links, wireless links, or combinations thereof. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.
The computing system 100 may include a central processing unit (CPU) 21, a system memory 22 and a system bus 23 that couples various system components including the system memory 22 to the CPU 21. Although only one CPU is illustrated in
The computing system 100 may further include a hard disk drive 27 for reading from and writing to a hard disk, a magnetic disk drive 28 for reading from and writing to a removable magnetic disk 29, and an optical disk drive 30 for reading from and writing to a removable optical disk 31, such as a CD ROM or other optical media. The hard disk drive 27, the magnetic disk drive 28, and the optical disk drive 30 may be connected to the system bus 23 by a hard disk drive interface 32, a magnetic disk drive interface 33, and an optical drive interface 34, respectively. The drives and their associated computer-readable media may provide nonvolatile storage of computer-readable instructions, data structures, program modules and other data for the computing system 100.
Although the computing system 100 is described herein as having a hard disk, a removable magnetic disk 29 and a removable optical disk 31, it should be appreciated by those skilled in the art that the computing system 100 may also include other types of computer-readable media that may be accessed by a computer. For example, such computer-readable media may include computer storage media and communication media. Computer storage media may include volatile and non-volatile, and removable and non-removable media implemented in any method or technology for storage of information, such as computer-readable instructions, data structures, program modules or other data. Computer storage media may further include RAM, ROM, erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other solid state memory technology, CD-ROM, digital versatile disks (DVD), or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can be accessed by the computing system 100. Communication media may embody computer readable instructions, data structures, program modules or other data in a modulated data signal, such as a carrier wave or other transport mechanism and may include any information delivery media. The term “modulated data signal” may mean a signal that has one or more of its characteristics set or changed in such a manner as to encode information in the signal. By way of example, and not limitation, communication media may include wired media such as a wired network or direct-wired connection, and wireless media such as acoustic, RF, infrared and other wireless media. Combinations of any of the above may also be included within the scope of computer readable media.
A number of program modules may be stored on the hard disk 27, magnetic disk 29, optical disk 31, ROM 24 or RAM 25, including an operating system 35, one or more application programs 36, a trust quantification application 60, program data 38, and a database system 55. The operating system 35 may be any suitable operating system that may control the operation of a networked personal or server computer, such as Windows® XP, Mac OS® X, Unix-variants (e.g., Linux® and BSD®), and the like. The trust quantification application 60 will be described in more detail with reference to
A user may enter commands and information into the computing system 100 through input devices such as a keyboard 40 and pointing device 42. Other input devices may include a microphone, joystick, game pad, satellite dish, scanner, or the like. These and other input devices may be connected to the CPU 21 through a serial port interface 46 coupled to system bus 23, but may be connected by other interfaces, such as a parallel port, game port or a universal serial bus (USB). A monitor 47 or other type of display device may also be connected to system bus 23 via an interface, such as a video adapter 48. In addition to the monitor 47, the computing system 100 may further include other peripheral output devices such as speakers and printers.
Further, the computing system 100 may operate in a networked environment using logical connections to one or more remote computers The logical connections may be any connection that is commonplace in offices, enterprise-wide computer networks, intranets, and the Internet, such as local area network (LAN) 51 and a wide area network (WAN) 52.
When using a LAN networking environment, the computing system 100 may be connected to the local network 51 through a network interface or adapter 53. When used in a WAN networking environment, the computing system 100 may include a modem 54, wireless router or other means for establishing communication over a wide area network 52, such as the Internet. The modem 54, which may be internal or external, may be connected to the system bus 23 via the serial port interface 46. In a networked environment, program modules depicted relative to the computing system 100, or portions thereof, may be stored in a remote memory storage device 50. It will be appreciated that the network connections shown are exemplary and other means of establishing a communications link between the computers may be used.
It should be understood that the various technologies described herein may be implemented in connection with hardware, software or a combination of both. Thus, various technologies, or certain aspects or portions thereof, may take the form of program code (i.e., instructions) embodied in tangible media, such as floppy diskettes, CD-ROMs, hard drives, or any other machine-readable storage medium wherein, when the program code is loaded into and executed by a machine, such as a computer, the machine becomes an apparatus for practicing the various technologies. In the case of program code execution on programmable computers, the computing device may include a processor, a storage medium readable by the processor (including volatile and non-volatile memory and/or storage elements), at least one input device, and at least one output device. One or more programs that may implement or utilize the various technologies described herein may use an application programming interface (API), reusable controls, and the like. Such programs may be implemented in a high level procedural or object oriented programming language to communicate with a computer system. However, the program(s) may be implemented in assembly or machine language, if desired. In any case, the language may be a compiled or interpreted language, and combined with hardware implementations.
At step 210, the trust quantification program 60 may calculate a trust value for each agent present in the computing network. In one implementation, the trust value, or a-priori trust value, may be generated based on previous knowledge about an agent. Such previous knowledge may include information about the agents expected behavior and his corresponding actual behavior in past transactions. The “gap” or discrepancy between the expected and actual behavior of the agent may be quantified and normalized per bit to create a discrepancy value in the interval between 0 and 1. In one implementation, the discrepancy value may correspond to a conditional entropy that may be normalized to a value in the interval between 0 and 1. The discrepancy value, or average uncertainty, of the agent based on another agent's previous knowledge about the agent. Alternatively, symmetric measure based on conditional entropy may be used to determine the discrepancy value (D(x,y)) such that D(x,y)=H_x(y)/H(x,y)+H_y(x)/H(x,y).
For example, suppose that x, y, and z are random variables or random agents in a network. If x is an ideal agent that does not lie or err, y is an agent in a well defined role whose trustworthiness is being evaluated, and z is a random variable describing the evaluator's trustworthiness in x. The average uncertainty or discrepancy value of y's trustworthiness given x's trustworthiness may be represented as a normalized conditional entropy of y given x's previous knowledge of y. The average uncertainty or discrepancy value of y given x may be denoted as Hx(y). Since x may be considered to be an ideal agent that does not lie or err, the absolute trustworthiness value of y may then be determined by subtracting the discrepancy value from 1, such that y's absolute trustworthiness, or ty, may be defined as ty=1−Hx(y). Since z is the evaluator's determination of how agent y should behave, agent z's trustworthiness in agent y, or tzy, may be defined as tzy=1−Hz(y).
For example, in an online marketplace where computer users may buy or sell merchandise on the Internet, previous knowledge pertaining to a purchaser (agent Y) may be used to generate an initial trust value for the purchaser. The purchaser's expected behavior and his corresponding actual behavior may correspond to his promise to pay a specified amount and the actual amount he paid in previous transactions. In one implementation, information pertaining to the purchaser's previous transactions may be provided by a credit card company (agent X). The discrepancy between the purchaser's promise to pay and his actual payment may be used to create a discrepancy value, or average uncertainty that the credit card company may have in the purchaser. The average uncertainty of the credit card company in the purchaser may be defined as Hcredit-card-co(purchaser). The trust value of the purchaser may then be determined by subtracting the discrepancy value from 1, such that the credit card company's trustworthiness in the purchaser may be defined as tcredit-card-co-purchaser=1−Hcredit-card-co(purchaser).
In one implementation, the trust quantification application 60 may help gather the input data for trust evaluation. For example, the trust quantification application 60 may detect truth-in-ads discrepancies (the ad promised price x, and the buyer was charged y>x) made by merchants. Furthermore, the trust quantification application 60 may detect discrepancies in a revocation list, such as complaints about truth-in-ads, and it may gather input data about the trustworthiness of revocation authority.
Although the above example based the gap value of the purchaser on his previous transaction's promise to pay and his subsequent actual payment, it should be noted that the gap value may be based on one or more other factors, such as information pertaining to the date in which the purchaser paid, the manner in which he paid it, and/or combinations of the like. Similarly, previous knowledge pertaining to a merchant (agent) may be used to generate an initial trust value for the merchant. The merchant's expected behavior and his corresponding actual behavior may correlate to his advertised price on a product and the actual amount he charged for the product in previous transactions. In one implementation, information pertaining to the purchaser's and/or the merchant's previous transactions may be provided by one or more credit card companies, banks, peer reviews, or the like.
At step 220, the trust quantification application 60 may group agents with trust values in each other into a trust group. The group of agents within a trust group may have trust values within a specified tolerance ‘q1’ of each other. In one implementation, the specified tolerance ‘q1’ may correspond to a high trust value. The trust values of each agent on each other may naturally converge to the extremes such that the result is the formation of maximal-trust trust groups among peers. For example, in extremely large trust groups, trust values of each agent in each other may converge to the extremes such that each agent may be deemed as either trustworthy or not. At step 230, the trust quantification application 60 may divide or split the trust group formed at step 220 into one or more sub-groups Splits may occur when a new agent does not meet the trust value requirement (specified tolerance ‘q1’) for the whole group but it may meet the trust requirement for a subgroup. If the new agent meets the trust requirements for a subgroup, the trust quantification application 60 may decide to split the group and accept the new agent into a subgroup as opposed to rejecting the new agent altogether. In one implementation, the sub-groups may contain agents with similar trust values in each other but with a smaller tolerance ‘q2’ than those agents in the trust group. The agents in the sub-group may be considered to “trust” each other more than those agents in the original trust group.
In one implementation, the trust quantification application 60 may split the larger trust group into two or more sub-groups based on one or more economic utility functions. Economic utility functions may be used to maximize the utility or purpose for establishing trust groups. Economic utility functions may measure the relative satisfaction of an agent based on a correlation between an agent's economic behavior and his desire for consuming various goods and services. Examples of some economic utility functions may include a trust group of sellers and buyers that may wish to maximize overall market share, a user who may wish to maximize the number of features (plug-in modules) in his machine, assigning various non-uniform weights to various features, or other types of utility functions. Unfortunately, all economic utility functions may not be able to exist in harmony with each other because there may be interdependency as well as conflicts between different economic utilities. Therefore, an agent may switch opportunistically among its cliques when performing distinct tasks that may depend on his economic utility functions. In one implementation, the trust values change as sub-groups grow, therefore, the trust quantification application 60 may have to verify each agent's acceptance into every sub-group. In one implementation, the trust quantification application 60 may gradually evaluate each agent's trustworthiness and allow some tolerance q>0 in the acceptance criteria and then a limit on the clique size may be quantified as n. If the trust quantification application 60 evaluates the agents and cliques instantaneously, there may not be a limit on the clique size. The overall uncertainty may be represented as O(q·exp(n)), hence n may most likely be small, and q<<exp(−n). In that case, the acceptance criteria may require that every pair of agents may have close to mutual maximal trust (>1−q) and that the consensus among agents in the sub-clique may be allowed to oscillate within tolerance δ=O(q·exp(n)), away from the maximal trust.
At step 310, the trust quantification application 60 may receive a request from a new candidate to gain entry into one or more trust groups or sub-groups. For example, with respect to the purchaser/merchant example described in
At step 320, the trust quantification application 60 may provide information pertaining to the new candidate to each agent in the trust group to receive a consensus trust evaluation on the new candidate. In the preceding example, the new purchaser may base its initial trust value on information obtained from a credit card company. Each agent of a trust group in the computing network may be provided the new purchaser's initial trust value which may be determined using information from the credit card company. Each agent may then evaluate the new purchaser's discrepancy value as described in step 210 and assess his own trust value for the new purchaser. In one implementation, each agent in a trust group or sub-group may assess their trust value in the new candidate based on their trust value in the credit card company that assigned the initial trust value in the candidate.
In another implementation, masters or users may program their agents to evaluate new candidates based on his initial trust value and the trust value given by other agents within the trust group. In another implementation, agents may not have the space to hold trust data, so they may request for trust values from their masters. In some implementations, agents may not even analyze the candidate's trust value; instead, they may request their masters to provide conclusions.
At step 330, the trust quantification application 60 may receive from each agent in the computing network a trust value assessment of the new candidate. Based on the received trust values of the agents within the trust group or sub-groups, the trust quantification application 60 may determine the consensus of all of the agents in the trust group at step 340.
At step 340, the trust quantification application 60 may determine a consensus value for the new candidate. In one implementation, at a discrete time (τ=0), the trust values may be defined as ti0(0) where i may represent an agent within the trust group. For each agent i, the consensus values at time (τ+1) may be. In one implementation, the process for calculating the consensus values may be iterated until a stable consensus is reached. Here, the weights may be uniform for each value; however, in some implementations, the weights may not be uniform yet the sum of all of the weights may be equal to one.
In one implementation, an estimation error in the values of a stable consensus trust may be defined as δ(n, N)=O(exp(n−N))where N is the size of message space and n is the total number of agents. Each random variable may be defined over a message space. For example, when evaluating a merchant and his truth in ads, each data point (“message”) may be an advertisement and the actual price charged as reported by many users. When a trust group grows gradually, adding one agent at a time, then the error in the trust quantification application 60 estimations may become exp(n−N), where n=number of agents in the group, and N=the number of messages in the message space used to evaluate the gaps. However, when evaluations are instantaneous the error may be much smaller, ˜2 exp(−N). Based on the consensus, the trust quantification application 60 may reject the new candidate's entry into a trust group (step 350), accept his entry into a trust group (step 360), or accept his entry into a sub-group (step 370).
At step 350, the trust quantification application 60 may reject the new candidate's entry into a trust group based on the consensus values received at step 340. In one implementation, rejection into a trust group may indicate to agents in a network that the candidate is not trustworthy. The new candidate may be rejected from a trust group if one or more members of the trust group do not have trust values within a specified tolerance for the new candidate.
At step 360, the trust quantification application 60 may accept the new candidate's entry into a trust group. In one implementation, if each member of a trust group has high trust value in the candidate, and the candidate thus has a high trust value in each member of the trust group, then the candidate may be accepted into the trust group. Acceptance into the trust group may indicate to agents in a network that the candidate is trustworthy within a certain degree. The trust value between each member in a trust group may be within a specified tolerance.
At step 370, the trust quantification application 60 may accept the new candidate's entry into a sub-group. In one implementation, if the new candidate is accepted into a sub-group, then the trust quantification application 60 may split the larger trust group into two or more sub-groups based on one or more economic utility functions as described in
In addition to the example of quantifying trust in an online purchaser and merchant relationship, the trust quantification application 60 may apply the same method 200 and method 300 to various scenarios such as trust between humans and certification authority, Tit-for-tat strategy in the iterative prisoner's dilemma game, the inter-relations among software modules in a system, the stock market, and other trust oriented applications.
Although the subject matter has been described in language specific to structural features and/or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as example forms of implementing the claims.
Number | Date | Country | |
---|---|---|---|
61078068 | Jul 2008 | US | |
61094861 | Sep 2008 | US |