This document proposes a framework for asserting fine-grained trust values between users in social networks to enable more granular privacy controls. It identifies several social factors that can be used to calculate subjective trust values, including identity, profile similarity, relationships, reputation, and interactions. An ontology and implementation are presented to aggregate these trust values and grant access to personal information based on a requester's overall trust level. Future work is outlined to improve and expand the model.
2. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Introduction
Current Social Networks
provide generic privacy settings for sharing information
do not take user’s trust into account
3. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Introduction
In reality, we only share parts of
our information to whom we trust
4. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Research Contribution
In this work we focus on:
Using various methods to
automatically assert fine-grained
subjective trust values for different
Social Factors to grant or restrict
access to personal information
5. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Use Case
1. Alex sends a request to view John’s personal
information
2. Alex is granted access to the information which
his trust value (asserted by John) satisfies a trust
threshold (assigned by John)
Example: Provide John’s phone number to Alex if
he has a trust level of 0.83
Social Web Platforms
...
1
2
Alex
John
6. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Defining Trust
In our work trust is defined as:
“a person’s subjective belief
that another person will act
responsibly and will not
misuse the information”
7. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Social Factors
Trust judgments are influenced by Social Factors:
Past interactions with a person
Opinions of a person’s actions
Other people’s opinions
Rumours
Psychological factors impacted over time
Life events
and so forth
These can be hard to compute since the
information required is limited and unavailable in
Social Networks
8. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Social Factors
However, we have identified several factors that
trust can be asserted from the information
available in Social Networks:
Identity
Profile Similarity
Relationship type
Reputation in a Web of Trust
Interactions
9. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Trust Assertion Framework
Social Web Platforms
...
Alex
JohnSocial Semantic Web Platform
Trust
Manager
1
2
1
2
3 5
4
6
11. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Trust Model
Trust values are represented in the range of [-1,1]
1 represents absolute trust
0 represents either uncertainty or unknown
-1 represents absolute distrust
Positive values less than 1 represents trust but with an
element of uncertainty
This also applies to negative values
12. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Trust Assertions Model
Identity-based Trust
Relies on authentication
The WebID protocol is used as a Single Sign-On Service for
Semantic Web applications
– WebID provides users to authenticate using FOAF and X.509
certificates over SSL
The subjective trust value is assigned to the requester
after s/he authenticates using WebID
1 if successful, -1 if unsuccessful and 0 if aborted
Definition 1: Identity-based Trust
Certificate(Cert,R) Profile(RP,R) Verify(Cert,RP)∧ ∧ ∧
AssertedBy(R,U) AssignTrust(IDT,R)⇒
13. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Trust Assertions Model
Profile Similarity-based Trust
Asserting trust based on the similarities between the user
profile and the requester
Basic information attributes are not taken into
consideration
Attributes such as work place information, interests,
connected peers and other profile attributes are compared
Profile similarity-based trust calculation:
– denotes profile similarity subjective trust valueτ
– m denotes the matched distinct profile attributes between
the user’s profile and the requester’s profile
– a denotes the user’s distinct profile attributes
14. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Trust Assertions Model
Definition 2: Profile Similarity-based Trust
∀UA(Profile(RP,R) Profile(UP,U)∧ ∧
Contain(RA,RP) Contain(UA,UP) Match(RA,UA)∧ ∧ ∧
AssertedBy(R,U)) AssignTrust(PST,R)⇒
where PST {-1,1}∈
15. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Trust Assertions Model
Relationship-based Trust
The Social Semantic Web provides the user to enter a value
of how much s/he trusts that particular relationship type
Definition 3: Relationship-based Trust
∀URT(Profile(UP,U) Contain(URT,UP)∧ ∧
Relationship(R,URT) AssertedBy(R,U)) AssignTrust(RLP,R)∧ ⇒
where RLP {-1,1}∈
16. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Trust Assertions Model
Reputation-based Trust
Consists of a trust value of a user within a Social graph
based on all trust values given by other users
Reputation-based trust value is the weighted average value
of all trust values given to a user
– The user’s trust values assigned within a network
– The weights denotes the reputation value of the person that
assigned the trust value
Reputation-based trust calculation:
– τ denotes reputation trust
– w denotes the reputation of the user assigning a subjective
trust value to the requester
– v denotes the requester’s subjective trust value assigned by a
user
17. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Trust Assertions Model
Definition 4: Reputation-based Trust
SocialGraph(SG,U) Contain(R,SG) Measure(RV,SG)∧ ∧ ∧
Reputaion(RV,R) AssertedBy(R,U) AssignTrust(RPT,R)∧ ⇒
where RPT {-1,1}∈
18. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Trust Assertions Model
Interactions-based Trust
Consists of users sharing microblog posts, comments,
photos, videos, links and other content with their peers
Interactions-based trust calculation:
– denotes interactions trust valueτ
– r denotes the number of interactions between the requester
and the user
– u denotes the number of all the user’s interactions in the
Social Web platform
19. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Trust Assertions Model
Definition 5: Interactions-based Trust
∀UI(Profile(UP,U) Contain(UI,UP) Interaction(R,UI)∧ ∧ ∧
AssertedBy(R,U)) AssignTrust(INTT,R)⇒
where INTT {-1,1}∈
20. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Trust Assertions Model
Aggregating Subjective Trust Values
To assign a fine-grained user’s subjective trust value to a
requester
An average of all the subjective trust values of a requester
from each social factor assigned by the user
Aggregate subjective trust calculation:
– denotes the aggregated subjective trust valueτ
– s a subjective trust value asserted based on a social factor
21. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Trust Assertion Ontology
(TAO)
http://vocab.deri.ie/tao
22. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Information Confidentiality
http://vocab.deri.ie/ppo
23. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Implementing Trust
Assertions
25. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Future Work
To show how this model is useful for other
scenarios (for example for Recommender Systems)
To calculate automatically Relationship-based Trust
values
To take into account the trust values of Certificate
Authorities when calculating Identity-based Trust
To compare our model with other state-of-the-art
models
26. Digital Enterprise Research Institute www.deri.ie
Enabling Networked Knowledge
Thanks!
Email: owen.sacco@deri.org
Notas del editor
Generic privacy settings – does not provide privacy settings for each part of the information SNs do not cater for trust, neither capture trust not even provide users to enter any trust information Assumes that every person in ones social graph has the same trust level Does not provide users to enter trust values for different user lists Cumbersome to manage contacts with respect to trust
Unlike the state-of-the art that focuses on 1 social factor and assumes that the users manually provide a trust score
We are working with personal subjective trust values Asymmetric, if I give a trust value to a person, it does not mean that person has the same trust value for myself .. Therefore each person calculates personal trust values for people
identity of the requester: trust can be asserted from the credentials exchanged through authentication, profile similarity between the user and the requester: trust can be asserted by matching several profile attributes with one another, the relationship type between the user and the requester: trust can be asserted based on the importance of the relationship type, the reputation of the user within a trusted network: trust can be asserted through reputation information asserted from other entities in a Web of Trust, trust based on interactions between the user and the requester: trust can be asserted based on the number of interactions between the user and the requester over a particular period of time.
“ the more similar two people were, the greater the trust between them” Attributes such as work place information, interests, projects, connected peers and other profile attributes are compared profile-similarity by calculating the relationship between the sum of matched distinct profile attributes between the user’s profile and the requester’s profile, and the total sum of all the distinct attributes within the user’s profile.