The current deep generative models have enabled the creation of synthetic facial images with remarkable photorealism, raising significant societal concerns over their potential misuse. Despite rapid advancements in the field of deepfake detection, developing an efficient and effective approach for the generalized deepfake detection of unseen forgery samples remains challenging. To address this challenge, we leverage the rich semantic priors of foundation models and propose a novel side-network-based decoder that extracts spatial and temporal cues using the CLIP image encoder for generalized video-based Deepfake detection. Additionally, we introduce Facial Component Guidance (FCG) to enhance spatial learning generalizability by encouraging the model to focus on key facial regions. By leveraging the generic features of a vision-language foundation model, our approach demonstrates promising generalizability on challenging Deepfake datasets while also exhibiting superiority in training data efficiency, parameter efficiency, and model robustness. The source code is available at: https://github.com/aiiu-lab/DFD-FCG

Although many deepfake detection methods have been proposed to fight against severe misuse of generative AI, none provide detailed human-interpretable explanations beyond simple real/fake responses. This limitation makes it challenging for humans to assess the accuracy of detection results, especially when the models encounter unseen deepfakes. To address this issue, we propose a novel deepfake detector based on a large Vision-Language Model (VLM), capable of explaining manipulated facial regions. We frame the deepfake detection task as Visual Question Answering (VQA) and perform visual instruction tuning to train the model on our collected Explainable Deepfake Face (ExDF) dataset. The dataset consists of fake images from diverse generative adversarial networks (GANs) and diffusion models (DMs), with explanations produced by GPT-4o guided by the corresponding ground-truth masks of the manipulated regions. Moreover, a facial mask encoder is introduced to guide the model to focus on key facial features, thereby improving the detection and explanation performances. Extensive experiments demonstrate that training the proposed model on the full ExDF dataset not only enhances detection accuracy compared to baseline methods but also provides detailed, human-interpretable explanations. To our knowledge, ExDF is the first explainable deepfake face dataset covering both GANs and DMs with comprehensive descriptions of altered facial regions. Our code and dataset are available at https://github.com/aiiu-lab/ExDF.

Vehicle-to-everything (V2X) is a core 5G technology. V2X and its enabler, Device-to-Device (D2D), are essential for the Internet of Things (IoT) and the Internet of Vehicles (IoV). V2X enables vehicles to communicate with other vehicles (V2V), networks (V2N), and infrastructure (V2I). While V2X enables ubiquitous vehicular connectivity, the impact of bursty data on the network’s overall Quality of Service (QoS), such as when a vehicle accident occurs, is often ignored. In this work, we study both 4G and 5G V2X utilizing Evolved Universal Terrestrial Radio Access New Radio (E-UTRA-NR) and propose the use of socially aware 5G NR Dual Connectivity (en-DC1) for traffic differentiation. We also propose localized QoS, wherein high-priority QoS flows traverse 5G road side units (RSUs) and normal-priority QoS flows traverse 4G Base Station (BS). We formulate a max-min fair QoS-aware Non-Orthogonal Multiple Access (NOMA) resource allocation scheme, QoS reclassify. QoS reclassify enables localized QoS and traffic steering to mitigate bursty network traffic’s impact on the network’s overall QoS. We then solve QoS reclassify via Integer Linear Programming (ILP) and derive its approximation. We demonstrate that both optimal and approximation QoS reclassify resource allocation schemes in our socially aware QoS management methodology outperform socially unaware legacy 4G V2X algorithms (no localized QoS support, no traffic steering) and socially aware 5G V2X (no localized QoS support, yet utilizes traffic steering). Our proposed QoS reclassify scheme’s QoS flow end-to-end latency requires only ≈15% of the time legacy 4G V2X requires.

Deepfake detection has attracted extensive attention due to widespread forged images on social media. Recently, self-supervised learning (SSL) based Deepfake detection approaches have outperformed supervised methods in terms of model generalization. However, we notice that most SSL-based methods do not take the manipulation strength levels of synthesized forgery samples into consideration according to different synthesis parameters and result in suboptimal detection performances. To address this issue, we introduce several auxiliary losses to the state-of-the-art SSL-based method based on different synthesis sub-tasks during data generation by inferring their synthesis parameters where the ground-truth labels are obtained from the synthesis pipeline for free. With comprehensive evaluations on various benchmarks, our approach has achieved noticeable performance improvement. Specifically, for the cross-dataset evaluation, the proposed approach outperforms the state-of-the-art method in terms of AUC on various datasets with improvements of 3.4%, 1.47%, 1.56%, and 1.3% on the CDF, DFDC, DFDCP, and FFIW datasets and achieves competitive performance on the DFD dataset. This further demonstrates the effectiveness of the proposed approach in its generalization ability.

Vehicle-to-everything (V2X) is one of the core technologies associated with 5G. It enables vehicles to communicate with other vehicles, networks, and infrastructure. Despite the promise V2X offers, i.e. ubiquitous vehicular connectivity, an often overlooked aspect of V2X is the impact bursty data, such as that encountered when a vehicle accident occurs, has on the network's Quality of Service (QoS). In this work, we study both 4G V2X and 5G V2X utilizing Evolved Universal Terrestrial Radio Access New Radio (E-UTRA-NR) Dual Connectivity (en-DC) Road Side Units (RSUs). By using social groups to determine QoS, versus statically defining QoS, we find that we can leverage 5G RSUs to minimize the impact of localized bursty traffic events on the overall network's QoS, while at the same time enabling regional awareness of localized road traffic conditions via a socially-aware Road Condition Warning Systems (RCWS). We demonstrate that our proposed socially-aware QoS management methodology outperforms legacy socially-unaware 4G V2X and can prove beneficial for 5G V2X and beyond.