## Towards a Formal Framework for Computational Trust (Extended Abstract)

Citations: | 4 - 0 self |

### BibTeX

@MISC{Sassone_towardsa,

author = {Vladimiro Sassone and Mogens Nielsen},

title = {Towards a Formal Framework for Computational Trust (Extended Abstract)},

year = {}

}

### OpenURL

### Abstract

Abstract. We define a mathematical measure for the quantitative comparison of probabilistic computational trust systems, and use it to compare a well-known class of algorithms based on the so-called beta model. The main novelty is that our approach is formal, rather than based on experimental simulation. 1

### Citations

4643 | A tutorial on Hidden Markov Models and selected applications in speech recognition
- Rabiner
- 1989
(Show Context)
Citation Context ...state assumptions182 V. Sassone, K. Krukow, and M. Nielsen and the Markovian transition probabilities, we can then incorporate unobservable states in the model by using so-called Hidden Markov Models =-=[15]-=-. A discrete Hidden Markov Model (HMM) is a tuple � � (S���t� O� s)whereS is a finite set of states; � is a distribution on S ,theinitial distribution; t : S ¢ S � [0� 1] is the transition matrix, wit... |

590 | Probability Theory: The Logic of Science
- Jaynes
- 2003
(Show Context)
Citation Context ...context in which it a«ects his own action.” The contribution of this paper is inspired by such a predictive view of trust, and follows the Bayesian approach to probability theory as advocated in e.g. =-=[8]-=- and exploited in works such as [13, 6, 17]. In particular, we borrow ideas from information theory to measure the quality of the behaviour-approximation algorithms and, therefore, suggest a formal fr... |

238 | The beta reputation system
- Ismail, Jøsang
- 2002
(Show Context)
Citation Context ...a 10 b 2 ) 1 �h� 2 � 11�14 � �80 However, if a state-change has indeed occurred, that probability would be as low as �05. Suppose now exponential decay is used, e.g., as in the Beta reputation system =-=[9]-=-, with a factor of Æ � �5. This means that the last observation weighs approximately the same as the rest of the history; in such a case, the algorithm would adapt quickly, and assign probability À(a ... |

178 | The role of trust management in distributed systems security
- Blaze, Feigenbaum, et al.
- 1999
(Show Context)
Citation Context ...s for computational trust and their fundamental properties. In the area of computational trust one common classification distinguishes between ‘probabilistic’ and ‘non-probabilistic’ models (cf. e.g. =-=[1, 2, 3, 11]-=- for the latter and [6, 16, 10, 14] for the former). The non-probabilistic systems vary considerably and need further classification (e.g., as social networks or cognitive); in contrast, the probabili... |

130 | Review on computational trust and reputation models
- Sabater, Sierra
- 2005
(Show Context)
Citation Context ...ir fundamental properties. In the area of computational trust one common classification distinguishes between ‘probabilistic’ and ‘non-probabilistic’ models (cf. e.g. [1, 2, 3, 11] for the latter and =-=[6, 16, 10, 14]-=- for the former). The non-probabilistic systems vary considerably and need further classification (e.g., as social networks or cognitive); in contrast, the probabilistic systems usually have common ob... |

103 | A Formal Model for Trust in Dynamic Networks
- Carbone, Nielsen, et al.
- 2003
(Show Context)
Citation Context ...s for computational trust and their fundamental properties. In the area of computational trust one common classification distinguishes between ‘probabilistic’ and ‘non-probabilistic’ models (cf. e.g. =-=[1, 2, 3, 11]-=- for the latter and [6, 16, 10, 14] for the former). The non-probabilistic systems vary considerably and need further classification (e.g., as social networks or cognitive); in contrast, the probabili... |

56 | Anonymity protocols as noisy channels
- Chatzikokolakis, Palamidessi, et al.
- 2007
(Show Context)
Citation Context ...e theory, we present theoretical results within the field, regarding a whole class of existing probabilistic trust algorithms. To our knowledge, no such approach has been proposed previously (but cf. =-=[4]-=- for an application of similar concepts to anonymity), and these presents the first formal results ever in way of comparison of computational trust algorithms. Structure of the paper. The paper is org... |

53 | Coping with inaccurate reputation sources: experimental analysis of a probabilistic trust model
- Teacy, Patel, et al.
- 2005
(Show Context)
Citation Context ...action.” The contribution of this paper is inspired by such a predictive view of trust, and follows the Bayesian approach to probability theory as advocated in e.g. [8] and exploited in works such as =-=[13, 6, 17]-=-. In particular, we borrow ideas from information theory to measure the quality of the behaviour-approximation algorithms and, therefore, suggest a formal framework for the comparison of probabilistic... |

50 |
A computational model of trust and reputation (for ebusinesses
- Mui, Mohtashemi, et al.
- 2002
(Show Context)
Citation Context ...action.” The contribution of this paper is inspired by such a predictive view of trust, and follows the Bayesian approach to probability theory as advocated in e.g. [8] and exploited in works such as =-=[13, 6, 17]-=-. In particular, we borrow ideas from information theory to measure the quality of the behaviour-approximation algorithms and, therefore, suggest a formal framework for the comparison of probabilistic... |

24 |
K.: P2P reputation management: probabilistic estimation vs. social networks
- Despotovic, Aberer
- 2006
(Show Context)
Citation Context ...ir fundamental properties. In the area of computational trust one common classification distinguishes between ‘probabilistic’ and ‘non-probabilistic’ models (cf. e.g. [1, 2, 3, 11] for the latter and =-=[6, 16, 10, 14]-=- for the former). The non-probabilistic systems vary considerably and need further classification (e.g., as social networks or cognitive); in contrast, the probabilistic systems usually have common ob... |

19 | K.: A probabilistic approach to predict peers performance in P2P networks
- Despotovic, Aberer
- 2004
(Show Context)
Citation Context ..., rather than its approximation�(¡�X). An example. In order to exemplify our measure, we compare the ¬-based algorithm of Mui et al [13] with the maximum-likelihood algorithm of Aberer and Despotovic =-=[5]-=-. The comparison is possible as the algorithms share the same fundamental assumptions that: each principal’s behaviour is so that there is a fixed parameter ¢ that at each interaction we have, indepen... |

10 | A Calculus for Trust Management
- Carbone, Nielsen, et al.
- 2004
(Show Context)
Citation Context ...s for computational trust and their fundamental properties. In the area of computational trust one common classification distinguishes between ‘probabilistic’ and ‘non-probabilistic’ models (cf. e.g. =-=[1, 2, 3, 11]-=- for the latter and [6, 16, 10, 14] for the former). The non-probabilistic systems vary considerably and need further classification (e.g., as social networks or cognitive); in contrast, the probabili... |

9 |
A Bayesian Model for Event-based Trust
- Nielsen, Krukow, et al.
(Show Context)
Citation Context ...ir fundamental properties. In the area of computational trust one common classification distinguishes between ‘probabilistic’ and ‘non-probabilistic’ models (cf. e.g. [1, 2, 3, 11] for the latter and =-=[6, 16, 10, 14]-=- for the former). The non-probabilistic systems vary considerably and need further classification (e.g., as social networks or cognitive); in contrast, the probabilistic systems usually have common ob... |

8 | Towards a Theory of Trust for the Global Ubiquitous Computer
- Krukow
- 2006
(Show Context)
Citation Context |

2 |
V.: A logical framework for reputation systems
- Krukow, Nielsen, et al.
- 2007
(Show Context)
Citation Context |

1 |
On information and suÆciency
- Kullback, Leibler
- 1951
(Show Context)
Citation Context ...efinition of a formal measure expressing the quality of probabilistic computational trust systems in various application environments. The measure is based on the so-called KullbackLeibler divergence =-=[12]-=-, also known as information divergence or relative entropy,used in the information theory literature to measure the ‘distance’ from an approximation to a known target probability distribution. Here we... |