Digital forensics is a proven method for collecting, preserving, reporting, analyzing, identifying, and presenting digital evidence from the original data, and it helps find evidence of cybercrimes. Intelligent multimedia forensics is a type of digital forensics and is essential because it is used to identify fake multimedia, including images, videos, audio, and text. This paper conducted a comprehensive survey for intelligent multimedia forensics, categorized into 3 classes: deep learning forensics, multimedia forensics, and network and Internet of Things forensics. In the first class, we provided a survey of the multiple attacks breaking the DL models, such as attacks in the training step, and the testing step, and crafted. In the craft attacks, we survey the three attacks: white box, gray box, and black box. We also provided some defense methods against DL attacks, training step attacks, and testing step attacks. In the second class, we offered a survey of multimedia forensics, including passive and active manipulation. In the third class, we provide a survey of network/IoT forensics, including the attacks in five layers: physical, data link, transport, application, and network. We also provided network/IoT attack detection in the third class using deep learning models. We applied AI models in various datasets and obtained higher accuracy and performance on the datasets from various used models. Finally, we offered the challenges and future direction for the researcher in this scope.
Read full abstract