In the Semantic Web, users can attach (semantic) meta data to Web resources. One problem is the trustworthiness of these annotations. This paper identifies two main steps towards a Semantic Web trust framework and provides possible solutions. The first step is to authenticate users and annotations. This is done in our approach by introducing Liberty Alliance identity management into Semantic Web annotations. Users have to login using Liberty Alliance accounts to create or modify annotations and their credentials are stored with the annotations. The author of an annotation can also define optional access rights to her annotations, for example to restrict access to the participants of a lecture course. We present our implementation which is an extension of the Annozilla plugin for the Mozilla Web browser. One advantage of using the Liberty Alliance specification for authentication is possible interoperability with other Web applications. Users can thereby reuse accounts for different services under their control (identity management). The second step is a framework for expressing trust in a (Web) resource. Users can rate resources (or other trust ratings) using a trust ontology. The rating is always done with regard to a distinct facet or aspect. For example, a user can express high trust in the w3c.org Web site regarding the aspect “Web technology”. Our trust ontology also integrates (optional) references to domainspecific text representations of trust values.
H.5.3 [Group and Organization Interfaces] and K.6.5 [Security and Protection] - Authentication Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.
Conference’04, Month 1–2, 2004, City, State, Country.
Copyright 2004 ACM 1-58113-000-0/00/0004…$5.00.
In the Semantic Web, users can attach (semantic) meta data to Web resources. These annotations can then be displayed to other users or used by computer programs to improve the understanding of content. For example, in an E-Learning scenario, users could annotate (Web) resources with meta data such as “difficult” or “important”, or rate resources with regard to a certain topic or course.
One problem is the trustworthiness of annotations. In the traditional WWW, users have to manually decide whether a Web page is trustworthy or not. For example, an annotation by the lecturer of a course would constitute high trust to her students. In a Semantic Web, we need a framework for representing trust in Web resources which can be used by programs to enhance search results or information personalization, for example.
Annotea provides a standardized framework for the creation and
management of annotations [
In this paper, we identity two main steps towards a trust framework for (Semantic Web) annotations. First of all, we want to motivate the goal of our research by explaining an example scenario in section 2. In section 3, we describe how to authenticate users applying the Liberty Alliance standard for distributed and federated identity management (section 3). In section 4, we then explain the expression of trust in resources using a trust ontology. The paper finishes with a short discussion of related work and conclusion.
Consider an E-Learning scenario, where a lecturer provides content in form of various Web pages. Users (e.g. lecturer, students, anonymous guests) could annotate and rate these ELearning Web resources. For example, a user wants to annotate a Web page (or part of it) as “difficult” or “important” (using a Semantic Web ontology, for instance), or express her trust in the Web page. These annotations could then be used to adapt content to a learner’s needs or preferences, or used by other users to evaluate resources. If annotations are written by the lecturer of the course, students certainly would have higher confidence in the reliability of the information.
Students can also make annotations but other users might find all information as completely trustworthy. In this case, it is useful to be able to express trust not only in resources such as Web pages but also in other annotations itself. If a student annotation is rated as trustworthy by several other users, this would constitute a higher level of trust.
Furthermore, when annotations are linked to user accounts, users might want to control access to their annotations. For example, a lecturer might want to restrict access to some annotations to the registered students of her course.
Requirements for trust for Semantic Web annotations can be summarized as follows:
Securely authenticating users who want to create or manage annotations Associating identity information with annotations so that other users can be certain who created an annotation
Rating resources using trust expressions in annotations
Using the existing protocols such as Annotea and Liberty Alliance without much modification, if possible These requirements roughly fall into these two categories: 1.Identity management: mechanism for securely authenticating users without necessarily relinquishing privacy considerations 2.Trust framework: an ontology and a framework for trust expressions and rating of resources In the following two sections we will outline our ideas for a solution regarding these requirements.
Our approach for identity management and user authentication is to use the standards of the Liberty Alliance project.
The Liberty Alliance project is an initiative by more than 150
companies and organization to provide standards for secure Web
authentication (www.projectliberty.org). The Liberty Alliance
specification offers a scheme to single sign on, authentication and
controlled sharing of user identities among services. It uses the
standard Security Assertion Markup Language (SAML)
(xml.coverpages.org/saml.html) for the secure exchange of
authentication information [
The integration of Liberty Alliance identity management in Annotea in our approach can be outlined according to the following procedure: In the Liberty Alliance approach, identity providers take care of the authentication of users, not the service providers. Advantages of using Liberty Alliance for Semantic Web annotations are as follows: (www.sourceid.org). We also use SourceID to provide the identity provider system.
Fig. 1 displays our user interface for creating an annotation. In the top part, the user can enter the text of her annotation. Then she has to select an annotation service. Depending on the service, one or more identity providers are available. In step 3 the user can select one identity provider and login using a Liberty Alliance account (as explained above). When creating several annotations within one session, the user has to login only once. Note that users cannot fill in a “author“ field, because the author information is provided through the Liberty Alliance account. In step 4, the user can choose from several (facultative) options. She can select a type of the annotation, the language and a privacy attribute for the annotation. The privacy option is used to define access rights to the annotations. Because users might not want to specify detailed access rights themselves, profiles with predefined options can be provided in our framework. A simple profile might include the following options: public: (read) access for everybody private: read access only for the author of the annotation anonymous: everybody can read the content of the annotation, but the author information is not shown The third option is interesting, because users have to login to create annotations in our framework and cannot make anonymous annotations by omitting author information or entering fake data.
In chapter 3, we explained how to integrate identity management into Semantic Web annotations to securely authenticating users. The second part now is to provide a framework for expressing trust in Web resources. The idea is that users can rate Web resources. Rating thereby means assigning a trust value to a resource. These ratings (or trust statements) are stored and managed using the Annotea protocol. Our approach is not only designed to rate Web resources but also other trust expressions themselves.
All trust statements are made with regard to a distinct facet or aspect. For example, a user might find the Web site of the World Wide Web Consortium trustworthy regarding “Web technology”. Since users have to login (as explained in section 3) to create trust statements, the ratings are associated with the user identity. Other examples for aspects are “design” or “function” when rating a product.
The core of the approach is a trust ontology that formalizes the
mentioned trust statements. Fig. 2 shows an example of an
instance of the ontology.
<rdf:RDF
... namespace definitions omitted ..>
<tr:TrustStatement
rdf:about="http://annotest.w3.org/annotations/anno
tation/5930134703.359235"
a:context="http://www.dgk.org/#xpointer(/html[
As an example in the E-Learning scenario from section 2, trustRegarding might be “relevance for exam”, the trustOntology could be various degrees of relevance, such as “very relevant” or “not really relevant, but might help”. The lecturer can then annotate material with trust statements, and, since the identity of users is securely authenticated, students can take these annotations for granted. On the other hand, students can create annotations themselves which might then be rated by other students.
The trust statements are integrated into Annotea using a distinct
annotation type. Annotea uses sub classes of a class “Annotation”
to describe the type of an annotation [
A lot of work is done in the field of trust for the Semantic Web.
An overview gives the (slightly outdated) Semantic Web
Trust and Security Resource Guide:
www.wiwiss.fuberlin.de/suhl/bizer/SWTSGuide/
Our trust ontology is similar to the one used by Golbeck, Parsia
and Hendler in [
C.Görn recently proposed an ontology for trust models [
In this paper, we have presented an approach to improve trust for (Semantic Web) annotations. First, we have applied Liberty Alliance identity management to authenticate users. Then, we have shown how users can make trust expression by using a trust ontology.
The rating of Web resources is already done by online shops such as amazon.com, for example. With a general Web trust framework, users could rate products or other resources across different sites. Such an approach would be susceptible to spam or fake ratings. In this case, secure authentication of users is especially important. This raises privacy issues. One solution is to disallow anonymous annotations but optionally leaving out author information when displaying annotations. In our framework, this is possible by associating access rights to annotations. Trust ratings could also be used to compute similarities between users and recommend item based on this user similarity (collaborative filtering). Thereby, the explicit representation of distrust in our model is beneficial.
An idea for improvement in our approach is the introduction of aging of trust annotations. Old ratings might constitute a lesser meaning than new ratings. Another (related) problem is how to handle changes in Web resources. If a Web page is modified, the associated trust statements have to be modified as well. Future work also includes the refinement of our implementation, which include the integration in current Annozilla versions, and the application and evaluation of the approach in the E-Learning scenario.
The author would like to thank his diploma/master students Markus Geiger and Werner Jansen, who worked out the details and designed and partly implemented the two parts of the system, and also Manuel Giuliani, who worked on the demo application.