Material from Talks

Michael Barnsley (University of Melbourne)

Separation of Image Science, Subjectivity and Mathematics in the Storage, Retrieval and Indexing of Digital Real World Images

A way of thinking about algorithms for the approximate representation of real world images by approximants indexed by finite digital strings is presented. The approach uses and separates the roles of science, subjective judgment and mathematics using ``pixel predictors" P:D->D, idempotent filters Q:D->D and indexing (compression) functions C:D->Z respectively, where D is a space of files representing digital images, and Z denotes the integers. When, for all given A in D, the equation B-P(B)=Q(A-P(B)) has a unique solution, then B is an idempotent transformation of A. The image B can be stored using C(B-P(B)). The overall result is an idempotent lossy and lossless compression scheme. Any compression scheme can either be represented in this form or "improved" by correcting its non-idempotent features; the latter step may be of high complexity. It is described how fractal, wavelet, and DCT based image storage methods fit within the approach. With regard to the final lossless compression step, which may be thought of and implemented by many different approaches, the role of dynamical systems is emphasized.

Edward J. Delp (Purdue University)  ace@ecn.purdue.edu

Multimedia Security: Is Their Hope in Securing our "Digital Future?"

In this talk I will describe some of the issue involved in multimedia security and indicate how I see the future evolving. In particular I will discuss what tools are currently available and how I see them being used in digital libraries. The immediate future looks somewhat bleak.

Edward J. Delp (Purdue University)  ace@ecn.purdue.edu

Tutorial: An Overview of Cryptography

Securing the "digital future" is a theme we have heard a great deal about in the popular press. People want to not only preserve their privacy but also protect digital information they create. This is very much a problem in our networked world. One tool that can be used to help accomplish this is encryption. In this talk I will present an overview of cryptography. This will include a look at some of the historical aspects of encryption. The talk will describe block cipher systems (e.g. DES )and public key systems (e.g. RSA) along with authentication techniques. How cryptography can be used in digital libraries will also be emphasized.

David Du (University of Minnesota)    du@cs.umn.edu

Asset Management Aspects of End-to-End Video Streaming over Internet

As the infrastructure of Internet has been rapidly improved and the last mile problem has been gradually resolved by DSL and Cable Modem connections, broadband video streaming becomes an important enabling technology for many applications like distance learning, tele-medicine and e-commerce. In this talk we examine the issues related to how to deliver millions of streams over Internet with quarantined quality of services. We will especially emphasize on the asset management aspect of the delivery path. The overall framework to enable broadband streaming video will also be discussed.

Sharon Flank (VP, Chief Technology Officer, eMotion, Inc.) sharon.flank@emotion.com http://www.emotion.com     

Metadata Issues

Digital media management relies on the use of descriptive information about media objects, or metadata. Good metadata permit quick search, extensive reuse, and consequently cost savings. If the metadata are inadequate, the file languishes unused and unusable. We propose a standard for metadata that does away with the trouble and expense of mandating terminology and thesaurus standards. We incorporate the thesaurus directly into the search software, so that both cataloguers and users are free to use any terms in English that accurately describe their media. It is then the responsibility of the search software to match them up. This paper describes the natural language processing elements of the search software, discusses how to represent hierarchical relationships efficiently, and considers the use and limitations of XML as a metadata tagging solution.

C. Lee Giles (School of Information Sciences and Technology, Computer Science and Engineering e-Business Research Center The Pennsylvania State University)  giles@ist.psu.edu

Web Search The Role for Niche Search Engines

The world wide web is a revolution in information access. The most common ways for finding information on the web is to either browse or use search services. A recent study of web users by the search engine AltaVista indicated that three out of five people who are online are doing some sort of search. Search services, many of which are known as search engines, are some of the most commonly visited sites on the web; a recent study showed that twenty search engines account for 98% of all web search. There are many types of search engines. General purpose search engines such as Google and AltaVista attempt to crawl and index all of the web. Specialized search engines try to crawl and index only a particular part of the web, e.g. ResearchIndex.org attempts to only find and index papers in computer and information science. Metasearch engines such as MetaCrawler search other search engines and fuse their results. Hierarchical directories such as Yahoo and OpenDirectory provide mostly handcrafted indexes for browsing. AskJeeves permits natural language queries. In this seminar we discuss how we characterize what the web is, whats there and how we use that knowledge to improve web search. Niche search engines become an interesting approach to web search since we know that the large search engines have problems indexing the entire web and giving relevant answers to queries. Niche search engines are specialized engines that only attempt to crawl and index a certain part of the web. We will discuss niche search engines and outline some of the research issues in building them.

James P. Hughes (Storage Technology Corporation)   jim@Network.com

The Secure Distribution of Information in a Corporate Environment

This presentation attacks a basic problem of how to store and share sensitive information when storage locations are independently managed. The methods today of protected networks, firewalls, link encryptors and individual file encryption does not generalize or scale to large centralized storage solutions.

Cryptographic separation of information (knowledge contained in the data) from the management of the data (loss of the bits) reduces the need for trusted storage systems. This is based on the fact that file systems that manage data do not need to "know" what the information that it stores "means."

In general, the protection of information has been the responsibility of the file system and as these systems grow and become more complex, distributed and independently managed this responsibility is becoming more troublesome. There are cryptographic means to transparently supplant the existing access control mechanisms making information protection secure and easy to implement.

Lyman Hurd (Iterated Systems, Inc.)   lhurd@mediabin.com

Protecting Property Algorithms to Solutions

In this talk I will outline some of the state of the art in algorithms designed to protect digital assets, with an emphasis on multimedia. I will examine digital asset protection from an algorithmic standpoint but also discuss social and legal implications. I will put the discussion into the context of systems and protocols, and examine as a case study the increased security gained by a centralized asset management system designed to effect repurposing of image assets.

Dhiraj Kacker (Shutterfly)    kacker@ecn.purdue.edu

Watermarking Printed Images

In this talk I will present a framework for jointly watermarking and halftoning a continuous-tone image. I'll begin by discussing a Human Visual System (HVS) model based halftoning technique called Direct Binary Search (DBS). DBS produces high-fidelity halftones by minimizing a HVS based cost function between the original continuous-tone image and the halftone to be printed. To this cost minimization framework a constraint of maximizing the output of some watermark detector is added. In particular, I'll present an algorithm in which the continuous-tone image is watermarked with a spread-spectrum watermark and the output of the corresponding correlation detector is jointly maximized with the HVS based measure of visual quality. The resulting halftone has both good visual quality and good watermark detection. The halftone is then printed and the watermark is detected from the scanned image.

Gerhard O. Michler (Universität Gesamthochschule Essen)

A Prototype of a Combined Digital and Retrodigitized Searchable Mathematical Journal

We will describe a prototype digital library of the journal "Archiv der Mathematik," Recent issues are linked to retro-digitized back issues. The resulting digital library is:

a) retrievable,
b) searchable

Referenced journal citations are recognized and their bibliographic data including ISSN (International Standard Serial Number) are produced in XML format. These bibliographic data allow the integration of the retrodigitized texts into the digital library database MILESS of the IT Center of Essen University.

Another main ingredient of our retrodigitization program system if the multivalent document format (MVD) developed by T. Phelps (Berkeley).

In this lecture also the recognition problem of mathematical formulas will be addressed and full credit will be given to the work of Professsor Okamoto's study group (University of Nagano).

Florian Pestoni (IBM Almaden Research Center)  fpestoni@almaden.ibm.com

Content Protection for Recordable Media

I will present a solution developed by IBM and its partners in 4C Entity to address the needs of copyright owners to maintain control over their digital assets. The scheme has been designed specifically for removable media, and has been announced for Secure Digital Card, DVD-RAM, DVD-R, DVD-RW and DVD-Audio, with other media formats in the works. CPRM is based on the concept of broadcast encryption and allows for revocation of circumvention devices.

Avni Rambhia (e-Vue Inc.)

Intellectual Property Management and Protection and MPEG-4

Digital media is an attractive format to disseminate multimedia content, from a cost and convenience point of view. However, digital content is susceptible to unlimited generations of piracy. The challenge to making digital media freely available is facilitating authorized use, while strongly hindering unauthorized used (whether intentional or otherwise). The definition of "authorized" must remain with the owner of the content, and it may change with time and circumstance. The process of defining and persistently enforcing rights and policies on digital content is called DRM or Digital Rights Management. This is also referred to as Intellectual Property Management and Protection, or IPMP.

MPEG-4 is an emerging multimedia standard that enables the streaming of high-quality interactive multimedia over a wide range of bandwidths. With traditional bottlenecks to streaming media like low quality and limited bandwidth poised to disappear, IPMP is a challenge that must be met before powerful and valuable MPEG-4 media presentations can be distributed freely.

In my talk, I will overview the process of IPMP from a served digital media point of view, from publishing to playback. I will also overview MPEG-4 and cover details of issues with IPMP in the MPEG-4 context. Some interesting problems are scalability of service, QoS under lossy environments, and concurrently providing interoperability of players and security of content.

IMA Public Lecture
Monday, January 12, 2001, 7:00 pm, 2650 Moos Tower
(Free and Open to the Public)

Bruce Schneier (Counterpane Internet Security, Inc.)

Natural Laws of Digital Content: The Folly of Copy Protection on the Internet

Copy protection is not new, but has given a new lease on life with the commercialization of the Internet. Old-world media companies want to sell their wares over digital networks, but want to control their use. Limiting resale, preventing copying, pay-per-use: these are all different ways of limiting what a consumer can do with a digital file after he receives it on his computer. Many different products purport to solve these problems, but they all fail. Making digital files uncopyable goes against the natural laws of digital content. In this talk I will discuss these laws, explain why all copy-protection mechanisms will fail technically, and describe some of the non-technical solutions media companies are pursuing.

Slides (pdf)      Audio   MP3  24k (15.5mb)  64k (41mb)  128k (82mb)

Richard E. Smith (Advanced Technology Division, Secure Computer Corporation)   rsmith@computer.org

Modeling the Intellectual Property Value Chain Based on Computer Security Incidents

To establish the utility of a mechanism for protecting intellectual property, we must evaluate it against a model of the intellectual property value chain. An accurate model should be consistent with real-world events that indicate the nature of the threat, types of attacks, and the structure of intellectual property distribution. The model should also reflect conditions that may set boundaries on both the virulence of the threat and the ability of owners to extract value from their property. While this analysis will not yield a value chain model, it will present and justify certain properties that such a model should have.

Francisco Javier Thayer (MITRE Corporation)

Graphical, Algebraic and Probabilistic Models for Protocols

In this talk we consider how free algebras are useful in strand space models of protocols and how they can be realized probabilistically. on algebraic, graphic and probabilistic models for cryptographic protocols.

H. Joel Trussell (Electrical and Computer Engineering Dept. North Carolina State University)   hjt@unity.ncsu.edu

Color Fidelity in Multimedia

Multimedia applications gain much of their effectiveness from the use of color graphics, images and video. Failing to maintain the accuracy of the color reproduction can result in poor quality and even misleading presentations. The problem of colorimetric accuracy is most apparent when attempting to produce hardcopy of softcopy images. However, the problem of communicating color between video devices is significant in itself. This talk will give a brief review of color science and the problems encountered in the multimedia environment. Some solutions will be suggested and the uses of accurate color for other applications discussed.

Michiel van der Veen (Philips Research Laboratories)   michiel.van.der.veen@philips.com

Secure Delivery of Entertainment Content

Privacy, authentication, security, copyrights, and payment play an important role in secure delivery of digital entertainment content. In this presentation, existing approaches in typical consumer electronic devices (e.g. DVD) will be discussed. Since digital assets may often be copied and distributed easily, ownership and copy protection information should be added. A dominant part of the presentation will focus on the utilization of digital watermarks to attach these data.

Hal R. Varian (Dean, School of Information Management & Systems, University of California-Berkeley)   hal@sims.berkeley.edu   http://www.sims.berkeley.edu/~hal

Buying, Renting, and Sharing Information Goods

Information goods such as books, journals, computer software, videos, etc. can often be various circumstances under which such sharing may increase or decrease producer profits. If a rental market is present, more copies will be sold at a lower price; I derive conditions that illustrate when this is more or less profitable than a sales-only market. When content is viewed only a few times and transactions costs of rental are low, rental may be more attractive than sales to both producers and consumers. Finally, when users have heterogeneous tastes, a rental market provides a nice way to segment high-value and low value users. These effects tend to suggest that rental markets may often increase profits, contrary to widespread views to the contrary.

URL for paper: http://www.sims.berkeley.edu/~hal/Papers/sharing.pdf

Material from Talks

Digital Libraries: Digital Asset Management

2000-2001 Program: Mathematics in Multimedia