MultiMedia FORensics in the WILD (MMForWILD) 2020

The Workshop MultiMedia FORensics in the WILD (MMForWILD) 2020 will be held on January 11, 2021, from 11:00 to 18:00 CET (Central European Time).

The schedule of the workshop will be as follows:

11:00 – 11:10, Opening & Welcome by Chair

11:10 – 12:30, Technical Session #1 (4 papers)
11:10 – 11:30, Analysis of the Scalability of a Deep-Learning Network for Steganography “Into the Wild” Ruiz, Hugo; Chaumont, Marc; Yedroudj, Mehdi; Oulad Amara, Ahmed; Comby, Frédéric; Subsol, Gérard [Slides]
11:30 – 11:50, Forensics Through Stega Glasses: the Case of Adversarial Images bonnet, benoit; Furon, Teddy; Bas, Patrick [Slides]
11:50 – 12:10, Increased-confidence Adversarial Examples for Deep Learning Counter-Forensics Li, Wenjie; Tondi, Benedetta; Ni, Rongrong; Barni, Mauro [Slides]
12:10 – 12:30, Defending Neural ODE Image Classifiers from Adversarial Attacks with Tolerance Randomization Carrara, Fabio; Caldelli, Roberto; Falchi, Fabrizio; Amato, Giuseppe [Slides]

12:30 – 13:20, Break

13:20 – 15:30, Technical Session #2 (4 papers)
13:20 – 14:10, Keynote 1: The problem of face verification in the forensic context: manual, semi-automatic and automatic approach. - Giovanni Tessitore, PhD, Commissario Capo Tecnico Fisico della P. di S., Rome IT.
Abstract: In the biometric context, face verification is usually addressed by using Automatic Facial Recognition Systems. The outcome of AFRS expressed as score, measuring the "similarity" between questioned and reference face, is compared with a threshold chosen to minimize the probabilities of errors (FP and FN on the basis of specific application). We will see that things are quite different in the forensics context where FIC (Facial Image Comparison) is currently addressed with a subjective evaluation made by an expert with a "manual procedure". Despite the increasing performance of AFRS during the last years, these systems cannot be used for forensic evaluation because they do not compute what it is called strength-of-evidence.
Bio: Giovanni Tessitore, graduated in Computer Science with a PhD in Computational Science at the University of Naples Federico II where he carried out research in the field of Neural Networks Artificial applied to the modeling of brain areas involved in action recognition. Since 2011 he joined the Italian National Police in 2011 and was assigned to the coordination of image and video processing and analysis as an officer in charge of the Electronic Investigations Section of the Scientific Police Service. In 2017 he was appointed director of the Electronic Investigations Section with expertise in the field of wiretapping, audio and video analysis, voice and face comparison, digital forensics and cybersecurity.

14:10 – 14:30, The Forchheim Image Database for Camera Identification in the Wild Hadwiger, Benjamin C; Riess, Christian [Slides]
14:30 – 14:50, LSSD: a Controlled Large JPEG Image Database for Deep-Learning-based Steganalysis “into the Wild” Ruiz, Hugo; Yedroudj, Mehdi; Chaumont, Marc; Comby, Frédéric; Subsol, Gérard [Slides]
14:50 – 15:10, In-Depth DCT Coefficient Distribution Analysis for First Quantization Estimation Battiato, Sebastiano; Giudice, Oliver; Guarnera, Francesco; Puglisi, Giovanni [Slides]
15:10 – 15:30, Nested Attention U-Net: A Splicing Detection Method for Satellite Images Horváth, János; Mas Montserrat, Daniel; Delp, Edward [Slides]

15:30 – 15:50, Break

15:50 – 18:00, Technical Session #3 (4 papers)

15:50 – 16:40, Keynote 2: A walk on the wild side of camera attribution - Fernando Pérez-González - Full Professor of University of Vigo, Vigo (UVIGO).
Abstract: This talk will be a guided tour to the developments made in camera attribution over the past 15 years of research, starting with images and later embracing video. Camera attribution is mainly based on the photoresponse non-uniformity (PRNU), which is a kind of noise that serves as a fingerprint of the device and is present in most digital sensors. We will discuss some of the standard applications with forensic interest and some emerging ones with commercial relevance. Emphasis will be put on the wild side, i.e., cases in which the target device is not available nor flatfield images that allow for a cleaner extraction of the PRNU. We will revise the fundamental mathematical assumptions that have been in place for more than a decade and find out some surprising facts.
Bio: Fernando Pérez-González is a Professor at the Department of Signal Theory and Communications, University of Vigo, Spain, where he has been working in the field of multimedia security for the last 25 years. He has co-authored more than 250 papers in international conferences and peer-reviewed journals, and 15 patents in various fields. During 2009-2012 he was the holder of the Prince of Asturias Endowed Chair on Information Science and Related Technologies at the University of New Mexico (UNM). During 2007-2014 he was the founding Executive Director of the Galician Research and Development Center in Advanced Telecommunications (GRADIANT), a semi-private research center. He leads the Signal Processing and Communications Group (GPSC) at the University of Vigo, with over 20 members, including faculty, postdoctoral researchers and graduate students. Fernando served as Associate Editor of IEEE Signal Processing Letters (2005-2009) and IEEE Trans. on Information Forensics and Security (2006-2010). Currently, he is Editor in Chief of EURASIP International Journal on Information Security and Senior Area Editor of IEEE Trans. on Information Forensics and Security. He is a Fellow of the IEEE for his contributions to Multimedia Security.

16:40 – 17:00, Learning to Decipher License Plates in Severely Degraded Images Kaiser, Paula; Schirrmacher, Franziska; Lorch, Benedikt; Riess, Christian [Slides]
17:00 – 17:20, Neural Network for Denoising and Reading Degraded License Plates Rossi, Gianmaria; Fontani, Marco; Milani, Simone [Slides]
17:20 – 17:40, Differential Morphed Face Detection Using Deep Siamese Networks Soleymani, Sobhan; Chaudhary, Baaria A; Dabouei, Ali; Dawson, Jeremy; Nasrabadi, Nasser [Slides]
17:40 – 18:00, Fingerprint Adversarial Presentation Attack in the Physical Domain Marrone, Stefano; Casula, Roberto; Orrù, Giulia; Marcialis, Gian Luca; Sansone, Carlo [Slides]

18:00 – 18:10, Closing Remarks

Industrial Sponsors