https://www.ronpub.com/ojiot/OJIOT_2021v7i1n03_Rong.html Channel of the paper: Dongzhu Rong, Yan Wang and Qindong Sun: Video Source Forensics for IoT Devices Based on Convolutional Neural Networks, Open Journal of Internet Of Things (OJIOT), 7 (1), pages 23-31, URN: urn:nbn:de:101:1-2021082919330983129965, 2021 en-us https://www.ronpub.com/ojiot/OJIOT_2021v7i1n03_Rong.html http://nbn-resolving.de/urn:nbn:de:101:1-2021082919330983129965 With the wide application of Internet of things devices and the rapid development of multimedia technology, digital video has become one of the important information dissemination carriers among Internet of things devices, and it has been widely used in many fields such as news media, digital forensics and so on. However, the current video editing technology is constantly developing and improving, which seriously threatens the integrity and authenticity of digital video. Therefore, the research on digital video forensics has a great significance. In this paper, a new video source passive forensics algorithm based on Convolutional Neural Networks(CNN) is proposed. CNN is used to classify the maximum information block of specified size in video I frame, and then the classification results are fused to determine the camera to which the video belongs. Experimental results show that the recognition algorithm proposed in this paper has a better performance than other methods in trems of accuracy and ROC curve. And our method still can have a good recognition effect even if a small number of I frames are used for recognition.