Advances in Artificial Intelligence (AI) enable the creation of manipulated digital videos with unprecedented quality and minimum user intervention, empowering even non-skilled users to generate plausible fake content and potentially pursue malevolent uses. With the goal of preserving the overall dependability of digital media data, the research field of MultiMedia Forensics (MMF) develops techniques to gather information on their digital history, detect manipulations, and output indications on their authenticity.
In this context, the objective of PREMIER is to devise new techniques capable of distinguishing fake from original videos. The project will develop improved MMF detectors based on deep learning, overcoming the main shortcomings of existing methodologies. Looked-for solutions will merge data-driven learning techniques with signal-processing-oriented methods, with the goal of i) decreasing the amount of data needed for training, ii) providing reliable ways to interpret the results of the forensic analysis, and iii) improving the security of the developed tools in the presence of an informed forger aiming at evading fake media detection.
PREMIER will contribute to boost video forensics technology and provide citizens with advanced publicly available tools to assess the authenticity of their own videos. Data collection of original and fake videos will be performed and updated in the course of the project, and will be made available to the international research community.
News
Grand finale for the PREMIER project!
January 2024
TheFinal PREMIER Workshop has taken place in Andalo (TN) on January 23 2024, in conjuction with theGTTI Thematic Meeting on Multimedia Signal Processing, which gathers every year the italian research community in multimedia signal processing.
The event involved members of PREMIER consortium, as well as renowned multimedia forensics researchers and practitioners at national and international level.
The scientific program included a keynote speech and six technical talks delivered by PREMIER researchers and invited contributors:
Keynote speech:"Synthetic Realities and the Future of Digital Forensics"
by Prof. Anderson Rocha (University of Campinas, Brazil)
Technical talks:
''Reliability Estimation for Synthetic Speech Detection in Data Mismatch Scenario'' by Paolo Bestagini (Polytechnic University of Milan)
"Trustworthy and interpretable approaches in media forensics" by Luisa Verdoliva (University of Naples, Federico II)
''Secure Detection of Image Manipulation via Random Feature selection and Gradient Orthogonalization'' by Benedetta Tondi (University of Siena)
''Floreview and TrueFace: two datasets for real world case scenario forensics applications'' by Massimo Iuliani and Andrea Montibeller (University of Florence and University of Trento)
''An Analysis of the Impact of Super Resolution on Deepfake Detection'' by Roberto Caldelli (CNIT, University Mercatorum)
''AI Compression: A New Challenge for Forensic Research'' by Christian Riess (University of Erlangen-Nürnberg)
Thanks to all the presenters and participants for the excellent contributions and the intriguing discussions!
March 2022
Mauro Barni is invited speaker at thepanel discussion on “Dis/Misinformation Analysis and Countermeasures – A Computational Viewpoint” organized by the APSIPA US Local Chapter.
Check out the programhere!
January 2022
Paolo Bestagini is co-organizer of the IEEE Signal Processing Cup 2022 on "Synthetic speech attribution"!
Check outhere for more information on the competition!
December 2021
Check out Cecilia Pasquini interview with Amped Software srl!
October 2021
Giulia Boato is giving a talk at the "Informatici senza Frontiere" Festival in Rovereto. Check out theprogram!
August 2021
Annalisa Verdoliva held thetutorial"Deepfake generation and detection" at the European Signal Processing Conference (EUSIPCO)
June 2021
Giulia Boato has co-held a webinar on fake news and cybersecurity organized by the Associazione Stampa Romana.
The event is targeted to professional journalists and aims at raising awareness on the threats posed by the use of internet in their work.
The speakers line-up included:
On. Sandra Cioffi, presidente Consiglio Nazionale degli Utenti
Generale Leonardo Tricarico, presidente ICSA
Marco Braccioli, capodivisione Cyber fondazione ICSA
Giulia Boato, Responsabile Unità di Ricerca MEDIA LAB Universita di Trento
Corrado Giustozzi, giornalista, consulente strategico, docente, divulgatore di cybersecurity e sicurezza delle informazioni
Carlo Picozza, giornalista, Ordine dei Giornalisti del Lazio
Lazzaro Pappagallo, giornalista, segretario associazione stampa romana
April 2021
Annalisa Verdoliva, Giulia Boato, and Dasara Shullani, discuss the opportunities and threats of deepfakes atRadio CICAP!
Listen to thefull interview.
February 2021
ThePREMIER PRIN project consortiumorganizes a scientific session at the GTTI-MMSP Thematic Meeting!
We'll show latest developments on how GAN-generated content and deepfakes can be created and recognized by machines and humans.
Check out theprogram!
December 2020
The IEEE Workshop on Information Forensics and Security 2020 is taking place online from December 6 to December 11!
The PREMIER consortium is actively contributing by presenting papers and offering tutorial sessions.
Check out the completetechnical program of the workshop!
May 2020
Mauro Barni and Benedetta Tondi talking on fake images and videos, and how to detect them!
April 2020
Project meeting in the time of Covid-19!
April 2020
Luisa Verdoliva is part of the task force set up by the italian government to monitor and contrast the diffusion of fake news on the COVID-19 pandemics.
Read more! (in italian)
February 2020
Luisa Verdoliva talking to theNew York Times on image manipulation detection!
February 2020
Our kick-off meeting took place at Cima Sappada, Italy, in conjunction with theGTTI MMSP annual thematic meeting. Going over the foundations and objectives of PREMIER, and planning upcoming activities!