The goal of FOSTERER (Robust Multimedia FOrensic Solutions To Enhance ModeRn MEdia TRustworthiness in Adversarial Conditions) is to develop new forensic tools that prove robust in a modern scenario in which forensic detectors work in a naturally adversarial environment. To do so, we will consider both targeted and non-targeted attacks that can be operated by either AI-based adversarial techniques or operations characteristic of the MMF field (e.g., social media sharing). To solve the cat-and-mouse game in which new defenses trigger new attacks, we will develop MMF defenses following two separate research lines. On one hand, we will develop robust detectors based on external information. These will seek for additional forensic traces in the related modalities of a media (e.g., pixels vs. metadata, audio track vs. visual track of a video, etc.). On the other hand, we will propose intrinsically robust detectors. These will exploit semantic information to get more robust features, and resort to the use of security-enhanced architectures and training.