Enforcing semantics-aware security in multimedia documents.

摘要

The vision of the Semantic Web [BLHL01] aims at putting machine-understandable data on the WWW where data can be shared and processed by automated tools as well as by people to enable integration and provide various web-based services. There is a growing need to provide security and ensure Quality of Service to such data primarily because of its transient nature and secondly because it traverses between multiple subsystems having varying requirements of security and quality.; SMIL (Synchronized Multimedia Integration Language) [Aya01] is an XML-like language for authoring multimedia documents. Unlike XML decorated text data, SMIL constructs have semantics that predate XML, because of the continuity and synchronization requirements pertaining to continuous media that are used to convey the intended meaning.; But the problem is that different syntactic representations could provide the same meaning at application runtimes. Furthermore, although considerable research has been done on the security of textual XML documents, the current SMIL specification and SMIL Metainformation Module [Mic01] do not define or support security and QoS constructs, nor do they provide syntax to define complex relations within multimedia documents to ensure proper semantic interpretation at application runtime.; Therefore in envisioning the Semantic web this dissertation identifies the protection object and represents it as an normal form that conveys the same semantics invariant to its many syntactic variations. It also defines the runtime operational semantics for SMIL, and a meta-structure and the related metadata to model security and QoS specifications. Traditional access control methods in various paradigms such as mandatory, discretionary and role-based are enforced by decorating SMIL documents in their normal form with the metadata thereby eliminating the dependency on the syntax of the language.; This work defines architectures for web-based surveillance applications (such as physical facilities, highway traffic etc) that require selective distribution of live and unfolding information as multimedia streams to recipients with display devices, be cause of their varying security credentials. Other contributions of this research are improvements in coordinated response mechanisms for facility and traffic emergencies with the use live surveillance feed, pay-per-view movies over the internet and authorization models for digital libraries.; Finally an very important contribution is the definition and usability of Unbreakable SMIL Constructs that in addition with regular SMIL constructs facilitate the creation of robust and self-describing multimedia documents. They avoid unintended or malicious mixing and matching of various media components that could lead to faulty information or wrongful incrimination due to fabrication of evidence.