organised by

« back

Invited Talks

 

Moving Targets

When Data Classes Depend on Subjective Judgement, or they are crafted by an Adversary to Mislead Pattern Analysis Algorithms
The Case of Content-Based Multimedia Retrieval and Adversarial Classification

Giorgio Giacinto
Pattern Recognition and Applications Group
Dept. of Electrical and Electronic Eng., University of Cagliari, Italy

 

Abstract

The vast majority of pattern recognition applications assume that data can be subdivided into a number of data classes on the basis of the values of a set of suitable features. Supervised techniques assume the data classes are given in advance, and the goal is to find the most suitable set of feature and classification algorithm that allows the effective partition of the data. On the other hand, unsupervised techniques allow discovering the "natural" data classes in which data can be partitioned, for a given set of features.
These approaches are showing their limitation to handle the challenges issued by applications where, for each instance of the problem, patterns can be assigned to different data classes, and the definition itself of data classes is not uniquely fixed. As a consequence, the set of features providing for an effective discrimination of patterns, and the related discrimination rule, should be set for each instance of the classification problem. Two applications from different domains share similar characteristics: Content-Based Multimedia Retrieval and Adversarial Classification. The retrieval of multimedia data by content is biased by the high subjectivity of the concept of similarity. On the other hand, in an adversarial environment, the adversary carefully craft new patterns so that they are assigned to the incorrect data class. In either case, the solution should be the use of a very high dimensional feature space, and a very large number of diverse pattern analysis techniques, in order to be able to select the most suitable problem formulation for each instance. This solution is clearly flawed, as the complexity of the resulting problem is much higher than the complexity of the original problem. In this talk, some approaches based on information fusion techniques are presented, and future research and implementation issues are discussed.

 

Vita

Giorgio Giacinto is Associate Professor of Computer Engineering at the University of Cagliari, Italy. He obtained the M.Sc. degree in Electrical Engineering in 1994, and the Ph.D. degree in Computer Engineering in 1999. Since 1995 he joined the research group on Pattern Recognition and Applications of the Dept. of Electrical and Electronic Engineering, University of Cagliari, Italy (http://prag.diee.unica.it). His research interests are in the area of pattern recognition and its application to new and challenging tasks. During his career Giorgio Giacinto has published more than sixty papers on international journals, conferences, and books. The main contributions are in the field of "multiple classifier systems", computer security (intrusion detection systems), and content-based image retrieval (relevance feedback techniques). He also contributes to researches in the fields of biometric personal authentication, remote sensing image classification, and video-surveillance. Giorgio Giacinto is associate editor of the "Information Fusion" journal, and regularly acts as a reviewer for the vast majority of the international journals on pattern recognition and applications. He also serves as program committee member in a number of international conferences on pattern recognition and applications. Giorgio Giacinto is a member of the IAPR (International Association for Pattern Recognition), the IEEE (Computer Society, and Geoscience and Remote Sensing Society), and the ACM.

presentation © Shutterstock presentation © Petra Perner Petra Perner © Petra Perner