Event-based Multimedia Indexing and Retrieval

Kashif Ahmad

Multmedia Analysis, Multimedia Retrieval, Event Detection, Image Classification

The spread of smart devices, such as tablets and smart-phones, as well as the increasing popularity of social networks, such as Twitter and Facebook, have modified the way in which people consume multimedia data. User-generated data, from such sources, poses new challenges in multimedia data analysis especially in terms of developing more efficient strategies for multimedia indexing and retrieval. Usually, user-generated multimedia contents are associated with some individual experiences or collective activities, such as a concert, sports matches or other social events. Recent studies demonstrate that users find it easier to search and browse media archives when they are organized according to underlying events. Although the scientific literature reports several interesting ideas on the subject, still many problems remain unsolved mostly due to the heterogeneity, multi-modality and the unstructured nature of the data. In this research work we aim to propose new perspectives and approaches.




Multimedia forensics for forgery detection

Alain Malacarne

Image Forensics, Multimedia Security, Mutimedia Forensics

Due to the increasingly simpler access to affordable smartphone devices and social networks, there is an always growing need for new approaches to content verification and analysis. Multimedia forensics aims at evaluating the truthfulness and consistency of digital contents, such as images or videos. Several techniques are developed with the purpose of unveiling possible manipulations that compromise the trustworthiness of the content itself.





Digital Camera Forensics on Online Social Networks

Quoc Tin Phan

Camera fingerprint clustering, camera identification, Sensor Pattern Noise, PRNU

Online Social Networks (OSNs) have emerged as the social trend enabling digital multimedia sharing between world-wide users. However, not all of information conveyed by means of digital multimedia are posted correctly and legally, and those falsely used multimedia are so sophisticated that human cannot detect. Consequently, multimedia forensic techniques, for instance, multimedia use verification, image forgery detection, camera identification, are profoundly needed to verify the validity of multimedia. My research focuses on the exploitation of source camera information from images on OSNs. Given a set of images uploaded from a user account, optimal clustering solutions to group images taken by the same camera together are considered. From image clusters, and with a certain confidence, forensic investigators can estimate the number of cameras the user owns, establish camera fingerprints, and link a suspected image to one of clusters.