190 likes | 248 Views
Workshop on Metadata for Security ( W-MS) International Federated Conferences (OTM '03). Business to Consumer Markets on the Semantic Web.
E N D
Workshop on Metadata for Security (W-MS)International Federated Conferences (OTM '03) Business to ConsumerMarkets on the Semantic Web Prof. Dr.-Ing. Robert Tolksdorf, Dipl.-Kfm. Christian BizerFreie Universität Berlin Dr. Rainer Eckstein, Dipl.-Inform. Ralf HeeseHumboldt-Universität zu Berlin
Interval - Knowledge Nets • Project Outline • Participants: Freie Universität Berlin, Humboldt-Universität zu Berlin • Funding: German Ministry of Education and Research (BMBF) • Duration: 7/03 – 12/07 • Project Goal • Predict economic impacts of the Semantic Web as a global data interoperability framework by using scenario analysis • Business Viewpoint • Predict impacts on markets and value chains already in the current early stage of the technological development • Estimate realization chances based on participant's interests and business models • Technological Viewpoint • Derive requirements for the technological infrastructure from the scenarios • Evaluate current technological developments
Agenda • B2C Markets on the Semantic Web • Potential Impacts and Obstacles • Trust and Security Requirements • Comparison of different Trust Mechanisms • Conclusion
B2C Commerce on the WWW • Most data relevant for purchasing decisions is available online • Product descriptions on the manufacturer's web sites • Offers by shops, electronic marketplaces and online actions • Product and vendor ratings on sites like epinions.com or bizrate.com • Challenges for a customer • Finding all relevant information sources for a specific product • Finding similar products • Integrating the information available in order to compare products and vendors • Purchasing decisions are based only on parts of the relevant data available
B2C Commerce on the Semantic Web • Semantic Web technologies offer a framework for the integration of all relevant purchasing data: • Uniform Resource Identifiers (URIs) as global identification mechanism for products and market participants • Web Ontology Language (OWL) for the definition of common terms and concepts needed to describe products and market participants • Resource Description Framework (RDF) data model together with its XML-based serialization syntax for the direct publication of data on the Web
Domain Ontology Information Providers Domain Ontologies Information Consumers Manufacturers Manufacturers Manufacturers Manufacturers Manufacturers Manufacturers Merchants Manufacturers Rating Services Shopping Portal Shopping Portal Shopping Portal B2C Commerce on the Semantic Web n:m Data Interchange Directory Services
Domain Ontologie Information Providers Domain Ontologies Information Consumers Manufacturers Manufacturers Manufacturers Manufacturers Manufacturers Manufacturers Merchants Manufacturers Rating Services Shopping Portal Shopping Portal Shopping Portal B2C Commerce on the Semantic Web n:m Data Interchange Directory Services
Advantages for a Customer • Portals allow product searches and comparisons based on all market data published • Increased market transparency • Querying instead of browsing • Concept based queries • Query per article or product category and not per shop • Domain knowledge supported queries • Query extension based on concept relations • Dissolving vague searches with domain knowledge • Result rendering according to user preferences • Level of detail, device, language
Economic Implications • The usage of Semantic Web technologies would lead to an increased market transparency. • Central Questions: Who would benefit? Is there a win-win situation for all participants? • Customers: Benefit from the increased market transparency. • Manufacturers: Face higher pressure of competition. • Merchants: Face higher pressure of competition. • Rating Services: Need to change business model. • Shopping Portals: Have a weaker market position.
Trust Requirements derived from the Scenario • Central enabling factor: Trustworthiness of the information published • Trust Situation • Open, dynamic network of information providers • Different intensions • Different levels of knowledge • Different levels of trust and proof required for different transactions • Requirements • Decentral trust architecture • Tolerance to incomplete information • Support for different, subjective trust strategies • Ability to explain trust decisions • Minimal effort for users
Trust and Security on the Semantic Web • Information is treated as argument and not as fact • Truth has to be evaluated by each application that processes the information • by using the context of the information • e.g. who said what and when and what credentials they had to say it • Levels of the Trust Decision • Verification of information origin • Evaluation of trustworthiness
1. Verification of Information Origin • Does information really originate from an author? • W3C Approach: Signing RDF models • Using XML-DSIG signatures and public key infrastructures • Disadvantages • Need for public key infrastructures • Lack of certificates, especially on the customer side • Pragmatic Approach: Information Source = Author • Crawler adds provenance metadata • Advantage • Lower effort for information providers • Disadvantage • Imprecise and less secure
2. Evaluation of Trustworthiness Context Information • Approach 1: Web-Of-Trust using explicit Ratings • Approach 2: Network Analysis using explicit Ratings • Approach 3: Analysis based on Context Information • Approach 4: Statistical Information Analysis • Approach 5: Comparison with Trusted Information Assertion Assertion Other Users trusts trusts Author 1 Assertion Author 1 Assertion Trustworthy? Author Information User Assertion states Trusted Information Context Information Context Information
Web-Of-Trust using Explicit Ratings • W3C Approach • Every users defines other trusted users. • Only statements of these users andusers these users trust are used in a trust evaluation. • Requirements • Relatively close community, where the members know each other • Explicit trust statement for every domain • High quality and currency of the ratings. • Conclusion • Relatively precise, if the paths are not to long • High effort for information providers • Not appropriate for open systems • Use cases • User defines trusted shop directories • User defines trusted test institutions
Network Analysis using Explicit Ratings • Approach: An external nodes use an existing Web-of-Trust. • Advantages • Doesn't require close community • No Web-of-Trust definitions for every user and every domain required • Disadvantages • Requires trust in the subjective statements of unknown users • Sensitivity to attacks like Bad-Mouthing and Ballot-Stuffing • Motivation for the users to make accurate trust statements • Use cases • Product ratings by other consumers • Shop ratings by other consumers
Analysis based on Context Information • Context Information • Assertion Context: Date, Author, Creation Method • Authors Context: Business Role, Membership in Organization • Example Trust Policies • Prefer product descriptions published by the manufacturer to descriptions published by a vendor. • Distrust everything a vendor says about its competitor. • Infer "is_competitor" from published offers. • Advantages • No explicit ratings needed • Transfer of real-world trust policies
Other Approaches • Statistical Information Analysis • Compare information published by different information providers. • Example strategy: Distrust an offer, if the price is 80 % lower than the average price offered by other merchants. • Comparison with Trusted Information • Compare new knowledge for contradictions against a trusted knowledge base. • Example strategy: Prefer own product classifications to classifications published by a manufacturer. • Ontological constraints • Test new knowledge against ontological constraints (cardinalities, ranges, disjunct classes) • Example strategy: A Product cannot belong to the classes "35mm Camera" and "Digital Camera".
Conclusion • The creation of trust will be the critical factor for the vision of the Semantic Web as an open system to become reality. • There won't be a single best approach, but a mixture of the 5 approaches depending on the • level of trust required for a specific application • the context and background information available • There is still a lot of research to do. • Further work • develop more scenarios to derive realistic requirements • develop a language to express trust policies as a combination of the 5 approaches. • Example policy: If there are more then 5 ratings use network analysis, otherwise use the author's role to evaluate the trustworthiness of a statement.
Thanks a lot! • For a collection of papers, ontologies and standards which could be building blocks of a Semantic Web trust architecture refer to: • Semantic Web Trust and Security Resource Guidehttp://www.wiwiss.fu-berlin.de/suhl/bizer/SWTSGuide Contact: Chris Bizer (chris@bizer.de) Project Homepage: http://www.inf.fu-berlin.de/inst/ag-nbi/research/wissensnetze/ Slides online: http://www.wiwiss.fu-berlin.de/suhl/bizer/pub/catania.pdf