/PhD Researcher Image and Video Forgery Detection & Forensics

PhD Researcher Image and Video Forgery Detection & Forensics

PhD - Gent Zwijnaarde | More than two weeks ago

We are looking for a qualified and motivated candidate for a PhD student position in multimedia forensics. One fully funded position as PhD student is available (4 years, with yearly progress evaluation).
 

Ghent University – imec IDLab-Media research team

The IDLAB-MEDIA research team is embedded in IDLab (Internet Technology & Data Science Lab), an imec research group at Ghent University and the University of Antwerp. Bringing together more than 300 experts, we develop technologies for a wide variety of topics, such as communication subsystems, high-speed and low-power networking, distributed computing, multimedia processing, machine learning, artificial intelligence, and web semantics. IDLab has a unique research infrastructure used in numerous national and international collaborations. IDLab also collaborates with many universities and research centers worldwide and jointly develops advanced technologies with industry (R&D centers from international companies, Flanders’ top innovating large companies and SMEs, as well as numerous ambitious start-ups).

Context

Images and videos are spread rapidly on social media, without guarantees on their authenticity. Yet, they could be manipulated (e.g. using Photoshop or new AI-based tools). In fight against disinformation, we need forgery detection and localization tools. As such, we can detect potential image/video manipulations to verify the authenticity of visual evidence and news. Additionally, journalists and digital-forensics investigators can be aided with other multimedia forensic tools as well, such as those that detect synthetically generated media, or those that reveal the digital history of an image or video. Sounds interesting? This job description might be for you!

We are looking for a qualified and motivated candidate for a PhD student position in multimedia forensics. One fully funded position as PhD student is available (4 years, with yearly progress evaluation).

The overall goal of the position is to develop and analyze AI-assisted algorithms and techniques to assess the authenticity of image and video content (e.g., as shared via the internet). Aspects that will be investigated include:

  • Analysis of image and video forgery techniques. Example forgery techniques are traditional editing tools such as Photoshop, or recent AI-based tools such as deepfakes;
  • The detection and localization of multimedia forgeries, using both conventional and more-recent artificial intelligence/machine learning methods, as well as a fusion of those methods;
  • Image and video compression analysis, to be used as traces for forgery detection;
  • Other multimedia forensics topics such as:
    • Synthetic media detection (e.g., was a media file generated by DALL-E?);
    • Provenance analysis (e.g., which social media platform was a media file posted on?);
    • Source acquisition device identification (i.e., which camera captured a media file?).

Our team IDLab-MEDIA focuses on multimedia representation, processing, compression and forensics. Our team currently consists of 6 researchers, working on a variety of projects. The work of our group can be explored on our team page: IDLab-MEDIA. In particular, the following projects and research results are relevant to this position:

  • FWO project on Fake Video Detection: Compression history discovery through video range decoding;
  • COM-PRESS: Combatting disinformation by equipping journalists with new image manipulation insights and detection methods;
  • Comprint: Image forgery detection and localization using compression fingerprints.

What you will do

As researcher within the IDLab-MEDIA team, you will be assigned to our core activities on multimedia forensics. We will introduce you to the field by starting to analyze still images first. Research you will perform includes, but is not limited to, image forgery detection, image compression analysis, and fusion of existing forgery detection methods. Gradually, you will shift focus towards analyzing videos, first considered as a sequence of still images, and later taking temporal aspects into account. As such, you will develop top-performing video forgery detection methods. Throughout the process, you will get the opportunity to develop yourself toward a highly valued expert in the field of multimedia forensics. Finally, the work you perform will partly be in collaboration with different Belgian and international commercial and media players in the field, improving your industrial relevance along the way. 

What we do for you

  • A full-time, 100% PhD scholarship (4 years, with yearly progress evaluation).
  • A competitive salary according to Ghent University salary scale AAP3 (indexed, minimum net monthly scholarship approximately €2,246.91).
  • 36 annual days of holiday leave.
  • A variety of UGent benefits (webpage in Dutch), such as public transport or bicycle allowance, discounted rate at student restaurants, and more.
  • A friendly and highly international working environment.
  • A variety of optional social activities organized by the research team, group and university. 
  • Publish research results in scientific journals in order to pursue your PhD degree.
  • Allowed to travel abroad to communicate your scientific findings at international conferences.
  • We are open for you to collaborate with (inter)national research partners, when relevant to your project. For this, we are happy to bring you in touch with our broad network.
  • Your supervisors (Dr. Hannes Mareen, Prof. Peter Lambert, and Prof. Glenn Van Wallendael) will guide you on a day-to-day basis.
  • We offer you a flexible starting date, with a preference to start as soon as possible.

Who you are

HARD SKILLS

Requirements

  • A relevant master’s degree (e.g., Master of Science in Computer Science Engineering, Master of Science in Informatics, Master of Science in Computer Science, Master of Science in Mathematics, or similar), at the time of starting the PhD position. 
  • Your English is fluent, both speaking and writing. 
  • Programming experience in Python or other relevant languages such as C++.
  • Basic knowledge on image and video representations, multimedia compression (e.g., JPEG and H.264/AVC) and machine learning (classification, regression, deep learning).

Recommended

  • Expert knowledge and/or hands-on experience with multimedia forensics, multimedia compression, and machine learning approaches or frameworks (e.g., Keras, Tensorflow…) are considered a plus.
  • Demonstrated first-class performance in your education, internship, or previous work experience (e.g., outstanding grades, thesis result, or publications).

SOFT SKILLS

  • You are a team player with good networking and reporting skills.
  • Decent (scientific) presentation skills.
  • You can communicate respectfully and clearly with peers, and are open to supervise master students.
  • You are eager to learn both at the scientific level and at the technological level. Having a 'growth mindset' is more useful than being the perfect match from day one.  

Interested?

Great!

For more information about the topic of the PhD itself, please contact Prof. Glenn Van Wallendael (glenn.vanwallendael@ugent.be). 

Applications may be sent by email to Laura Smekens (laura.smekens@ugent.be), with the following guidelines:

  • Email caption: “Application PhD <first name> <last name>”
  • Include cover letter and CV.
  • Include contact details of 1 to 3 referees. Do not include reference letters, these will be solicited by us.
  • Preferably, include an overview of your study results and a final mark of the master’s degree. Especially the marks of relevant courses and your master’s thesis are of interest.
  • Application deadline: 12 February 2023. Although applications remain welcome until the position is filled.

Suitable candidates will be invited for an interview (live or video call), and may get a small assignment.

Who we are
Accept marketing-cookies to view this content.
Cookie settings
imec's cleanroom
Accept marketing-cookies to view this content.
Cookie settings

Send this job to your email