A Behavioural Model for Consumer Reputation Anirban Basu, Ian Wakeman and Dan Chalmers Department of Informatics, University of Sussex {A.Basu, ianw and D.Chalmers}@sussex.ac.uk

In recent years, substantial academic research has focussed on solving the problem of email spam using reputation mechanisms (e.g., [1] and [2]) while others [3] have discussed the generalised problem of any unsolicited network transaction in client-server scenarios. We propose a reputation model based on behavioural history as a solution to this problem. We use behavioural history of long-lived network identities to develop local and global reputation. This helps the provider to make decisions regarding future network transactions. The decisions are either binary (e.g., provide service or cut-off) or more fine grained where the provider can vary levels of service. A declarative specification of behaviour quality of network identities in relation to relevant terms in the service contract is being developed, which will relate network activities to discrete units of good or bad behaviour. In most cases, clients have their own identities or they are identified as members of a group (e.g., customers of a particular Internet Service Provider). In situations where the identity of the client is anonymous and cannot be associated with any group, best effort service will be provided as a fall-back option. The term score is used to denote reputation value of the consumer, while the term rank is used for the provider. We have experimented with mathematical functions to best represent the reputation response to behaviour. We are using exponential saturation, exponential decay and linear patterns with adjustable slope (change) parameters. A local reputation response (ordinate) in relation to cumulative behaviour (abscissa) is illustrated in Fig. 1. The score saturates at positive or negative saturation levels (solid curves) with good and bad behaviour respectively. Good score drops with bad behaviour, and bad score gets better with good behaviour (dotted curves). There is also a score expiring mechanism, which ‘decays’ good reputation to positive default and bad reputation to negative default, respectively, over time. This is represented using quadratic decay patterns. The local scores are globally aggregated for collaborative inferences. A global score can initialise a local score if no prior local score is available. The local scores are securely submitted to and stored in a score aggregation system. The global score of a consumer is calculated, at query-time, over a window of scores submitted by providers. The impact of each submission is weighed by the rank of the provider. Newer score submission from the same provider overrides the corresponding previous submission. There is time decay, which helps scavenging older scores. We are investigating an interesting application of the declarative mapping of terms from the service contract as well as the reputation scheme on a front-end load balancer to filter network connections to back-end service endpoints.

1.5

Theoretical positive saturation

1

Reputation response (r)

0.5

Positive default 0 Negative default

−0.5

−1

−1.5 −1000

Theoretical negative saturation

−800

−600

−400

−200

0 Behaviour (b)

200

400

600

800

1000

Fig. 1. Graph of local reputation response to behaviour

We are developing a long-lived identity mechanism [4] and also exploring the use of public key infrastructure in the context of our proposed model. We are looking into formal ways of developing provider ranking. The model presented here will be simulated with test data-set from real network logs (e.g., The Internet Traffic Archive1 ). We are looking into an interesting experiment on how to use the consumer score as an incentive mechanism in a peer-to-peer content distribution system. We are investigating the defense against possible attacks (e.g., denial of service, source spoofing) on the model.

References 1. Garriss, S., Kaminsky, M., Freedman, M., Karp, B., Mazi`eres, D., Yu, H.: Re: Reliable Email. Proceedings of the 3rd Symposium of Networked Systems Design and Implementation (NSDI ’06) (2006) 2. Golbeck, J., Hendler, J.: Reputation Network Analysis for Email Filtering. Proceedings of Conference on Email and Anti-Spam (CEAS) (2004) 3. Allman, M., Blanton, E., Paxson, V.: An Architecture for Developing Behavioral History. Proc. Workshop on Steps to Reducing Unwanted Traffic on the Internet (2005) 4. Wakeman, I., Chalmers, D., Fry, M.: Reconciling privacy and security in pervasive computing: The case for pseudonymous group membership. Submitted for publication (June 2007) 1

http://ita.ee.lbl.gov/

A Behavioural Model for Consumer Reputation - Semantic Scholar

transaction in client-server scenarios. We propose a reputation model based ... This helps the provider to make decisions regarding future network transactions.

539KB Sizes 0 Downloads 61 Views

Recommend Documents

A Behavioural Model for Consumer Reputation - Semantic Scholar
This helps the provider to make decisions regarding future network transactions. ... group (e.g., customers of a particular Internet Service Provider). In situations.

A Behavioural Model for Consumer Reputation - Semantic Scholar
problem of email spam using reputation mechanisms (e.g., [1] and [2]) while ... cut-off) or more fine grained where the provider can vary levels of service.

A Behavioural Model for Consumer Reputation
We will use the terms score and rank to denote reputation of a consumer and of a ... cay with no activity over time helps a saturated bad reputation to recover; and .... any available real world input data (e.g., The Internet Traffic Archive ... Emai

A Reputation Framework for Behavioural History
2.6 Related work on behavioural history and client reputation . . . 24. 2.6.1 Allman's ...... Cisco's IronPort Reputation Filter (Cisco Systems, 2009a) technology.

ACTIVE MODEL SELECTION FOR GRAPH ... - Semantic Scholar
Experimental results on four real-world datasets are provided to demonstrate the ... data mining, one often faces a lack of sufficient labeled data, since labeling often requires ..... This work is supported by the project (60675009) of the National.

A Taxonomy of Model-Based Testing for ... - Semantic Scholar
ensure a repeatable and scientific basis for testing. ... is insufficient to achieve a desired level of test coverage, different test methods are ... The test specification can take the form of a model, executable model, script, or computer ..... Onl

Link-PLSA-LDA: A new unsupervised model for ... - Semantic Scholar
The output of the new model on blog data reveals very inter- ... modeling topics and topic specific influence of blogs. Introduction ..... 07/04/2005 and 07/24/2005.

A Taxonomy of Model-Based Testing for ... - Semantic Scholar
1. A Taxonomy of Model-Based Testing for Embedded. Systems from Multiple Industry Domains. Justyna Zander, Ina Schieferdecker, and Pieter J. Mosterman.

Model of dissipative dielectric elastomers - Semantic Scholar
Feb 3, 2012 - View online: http://dx.doi.org/10.1063/1.3680878. View Table of Contents: ... As a result of their fast response time and high energy density, dielectric ... transducer is an electromechanical system with two degrees of freedom.

Model Interoperability in Building Information ... - Semantic Scholar
Abstract The exchange of design models in the de- sign and construction .... that schema, a mapping (StepXML [9]) for XML file representation of .... databases of emissions data. .... what constitutes good modelling practice. The success.

Model Construction in Planning - Semantic Scholar
For all but the simplest domains, this technique has obvious deficiencies. ... programming might be a good trick to use, though, if we can develop a planner that can identify parts of a ... It might post a goal to visit the market to buy supplies.

MCGP: A Software Synthesis Tool Based on Model ... - Semantic Scholar
Department of Computer Science, Bar Ilan University. Ramat Gan 52900 .... by following the best generated programs, and by navigating through the chain.

MCGP: A Software Synthesis Tool Based on Model ... - Semantic Scholar
whether) a candidate solution program satisfies a property. The main ... software, a natural challenge is to generate automatically correct-by-design pro- grams.