zhichao Lian

Qingcao Li, Yipeng Lin, Weichen Lian et al.

Speech deepfake detection has achieved remarkable success in clean environments but faces significant challenges in complex, real-world scenarios where speech is often mixed with background music or noise. Current state-of-the-art methods rely on semantic features from self-supervised learning (SSL) models, which often fail when processing non-speech or mixed-source audio. In this paper, we first introduce MixFake, a large-scale benchmark dataset designed to simulate diverse acoustic environments with varying SNR levels and mixed authenticity components. To address the "semantic-centric" limitation, we propose a Multi-stream Prompt Tuning framework that injects signal-level priors into SSL backbones. By integrating base, frequency, and texture streams through deep prompt injection, our model effectively captures acoustic artifacts. Experimental results demonstrate that our method significantly outperforms existing baselines, achieving a 0.95% EER in foreground detection and a substantial 7.72% absolute improvement in complex background detection tasks. Our dataset and code are available at https://github.com/saltfish233/MixFake.

Sitong Liu, Zhichao Lian, Shuangquan Zhang et al.

Adversarial attacks meticulously generate minuscule, imperceptible perturbations to images to deceive neural networks. Counteracting these, adversarial purification methods seek to transform adversarial input samples into clean output images to defend against adversarial attacks. Nonetheless, extent generative models fail to effectively eliminate adversarial perturbations, yielding less-than-ideal purification results. We emphasize the potential threat of residual adversarial perturbations to target models, quantitatively establishing a relationship between perturbation scale and attack capability. Notably, the residual perturbations on the purified image primarily stem from the same-position patch and similar patches of the adversarial sample. We propose a novel adversarial purification approach named Information Mask Purification (IMPure), aims to extensively eliminate adversarial perturbations. To obtain an adversarial sample, we first mask part of the patches information, then reconstruct the patches to resist adversarial perturbations from the patches. We reconstruct all patches in parallel to obtain a cohesive image. Then, in order to protect the purified samples against potential similar regional perturbations, we simulate this risk by randomly mixing the purified samples with the input samples before inputting them into the feature extraction network. Finally, we establish a combined constraint of pixel loss and perceptual loss to augment the model's reconstruction adaptability. Extensive experiments on the ImageNet dataset with three classifier models demonstrate that our approach achieves state-of-the-art results against nine adversarial attack methods. Implementation code and pre-trained weights can be accessed at \textcolor{blue}{https://github.com/NoWindButRain/IMPure}.

Zijian Ying, Qianmu Li, Tao Wang et al.

Various methods try to enhance adversarial transferability by improving the generalization from different perspectives. In this paper, we rethink the optimization process and propose a novel sequence optimization concept, which is named Looking From the Future (LFF). LFF makes use of the original optimization process to refine the very first local optimization choice. Adapting the LFF concept to the adversarial attack task, we further propose an LFF attack as well as an MLFF attack with better generalization ability. Furthermore, guiding with the LFF concept, we propose an $LLF^{\mathcal{N}}$ attack which entends the LFF attack to a multi-order attack, further enhancing the transfer attack ability. All our proposed methods can be directly applied to the iteration-based attack methods. We evaluate our proposed method on the ImageNet1k dataset by applying several SOTA adversarial attack methods under four kinds of tasks. Experimental results show that our proposed method can greatly enhance the attack transferability. Ablation experiments are also applied to verify the effectiveness of each component. The source code will be released after this paper is accepted.

Sartaj Ahmed Salman, Zhichao Lian, Milad Taleby Ahvanooey et al.

Recently, the application of deep learning models to diagnose neuropsychiatric diseases from brain imaging data has received more and more attention. However, in practice, exploring interactions in brain functional connectivity based on operational magnetic resonance imaging data is critical for studying mental illness. Since Attention-Deficit and Hyperactivity Disorder (ADHD) is a type of chronic disease that is very difficult to diagnose in the early stages, it is necessary to improve the diagnosis accuracy of such illness using machine learning models treating patients before the critical condition. In this study, we utilize the dynamics of brain functional connectivity to model features from medical imaging data, which can extract the differences in brain function interactions between Normal Control (NC) and ADHD. To meet that requirement, we employ the Bayesian connectivity change-point model to detect brain dynamics using the local binary encoding approach and kernel hierarchical extreme learning machine for classifying features. To verify our model, we experimented with it on several real-world children's datasets, and our results achieved superior classification rates compared to the state-of-the-art models.

Yating Ma, Zhichao Lian

With the development of technology rapidly, applications of convolutional neural networks have improved the convenience of our life. However, in image classification field, it has been found that when some perturbations are added to images, the CNN would misclassify it. Thus various defense methods have been proposed. The previous approach only considered how to incorporate modules in the network to improve robustness, but did not focus on the way the modules were incorporated. In this paper, we design a new fusion method to enhance the robustness of CNN. We use a dot product-based approach to add the denoising module to ResNet18 and the attention mechanism to further improve the robustness of the model. The experimental results on CIFAR10 have shown that our method is effective and better than the state-of-the-art methods under the attack of FGSM and PGD.

Ruiqi Zha, Zhichao Lian, Qianmu Li et al.

Most of previous deepfake detection researches bent their efforts to describe and discriminate artifacts in human perceptible ways, which leave a bias in the learned networks of ignoring some critical invariance features intra-class and underperforming the robustness of internet interference. Essentially, the target of deepfake detection problem is to represent natural faces and fake faces at the representation space discriminatively, and it reminds us whether we could optimize the feature extraction procedure at the representation space through constraining intra-class consistence and inter-class inconsistence to bring the intra-class representations close and push the inter-class representations apart? Therefore, inspired by contrastive representation learning, we tackle the deepfake detection problem through learning the invariant representations of both classes and propose a novel real-centric consistency learning method. We constraint the representation from both the sample level and the feature level. At the sample level, we take the procedure of deepfake synthesis into consideration and propose a novel forgery semantical-based pairing strategy to mine latent generation-related features. At the feature level, based on the centers of natural faces at the representation space, we design a hard positive mining and synthesizing method to simulate the potential marginal features. Besides, a hard negative fusion method is designed to improve the discrimination of negative marginal features with the help of supervised contrastive margin loss we developed. The effectiveness and robustness of the proposed method has been demonstrated through extensive experiments.

Tao Wang, Zijian Ying, Qianmu Li et al.

Adversarial examples generated from surrogate models often possess the ability to deceive other black-box models, a property known as transferability. Recent research has focused on enhancing adversarial transferability, with input transformation being one of the most effective approaches. However, existing input transformation methods suffer from two issues. Firstly, certain methods, such as the Scale-Invariant Method, employ exponentially decreasing scale invariant parameters that decrease the adaptability in generating effective adversarial examples across multiple scales. Secondly, most mixup methods only linearly combine candidate images with the source image, leading to reduced features blending effectiveness. To address these challenges, we propose a framework called Uniform Scale and Mix Mask Method (US-MM) for adversarial example generation. The Uniform Scale approach explores the upper and lower boundaries of perturbation with a linear factor, minimizing the negative impact of scale copies. The Mix Mask method introduces masks into the mixing process in a nonlinear manner, significantly improving the effectiveness of mixing strategies. Ablation experiments are conducted to validate the effectiveness of each component in US-MM and explore the effect of hyper-parameters. Empirical evaluations on standard ImageNet datasets demonstrate that US-MM achieves an average of 7% better transfer attack success rate compared to state-of-the-art methods.

Wei Zhang, Yeying Jin, Xin Li et al.

Image-based virtual try-on (VTON) aims to synthesize photorealistic images of a person wearing specified garments. Despite significant progress, building a universal VTON framework that can flexibly handle diverse and complex tasks remains a major challenge. Recent methods explore multi-task VTON frameworks guided by textual instructions, yet they still face two key limitations: (1) semantic gap between text instructions and reference images, and (2) data scarcity in complex scenarios. To address these challenges, we propose UniFit, a universal VTON framework driven by a Multimodal Large Language Model (MLLM). Specifically, we introduce an MLLM-Guided Semantic Alignment Module (MGSA), which integrates multimodal inputs using an MLLM and a set of learnable queries. By imposing a semantic alignment loss, MGSA captures cross-modal semantic relationships and provides coherent and explicit semantic guidance for the generative process, thereby reducing the semantic gap. Moreover, by devising a two-stage progressive training strategy with a self-synthesis pipeline, UniFit is able to learn complex tasks from limited data. Extensive experiments show that UniFit not only supports a wide range of VTON tasks, including multi-garment and model-to-model try-on, but also achieves state-of-the-art performance. The source code and pretrained models are available at https://github.com/zwplus/UniFit.

Zijian Ying, Qianmu Li, Zhichao Lian et al.

Saliency maps have proven to be a highly efficacious approach for explicating the decisions of Convolutional Neural Networks. However, extant methodologies predominantly rely on gradients, which constrain their ability to explicate complex models. Furthermore, such approaches are not fully adept at leveraging negative gradient information to improve interpretive veracity. In this study, we present a novel concept, termed positive and negative excitation, which enables the direct extraction of positive and negative excitation for each layer, thus enabling complete layer-by-layer information utilization sans gradients. To organize these excitations into final saliency maps, we introduce a double-chain backpropagation procedure. A comprehensive experimental evaluation, encompassing both binary classification and multi-classification tasks, was conducted to gauge the effectiveness of our proposed method. Encouragingly, the results evince that our approach offers a significant improvement over the state-of-the-art methods in terms of salient pixel removal, minor pixel removal, and inconspicuous adversarial perturbation generation guidance. Additionally, we verify the correlation between positive and negative excitations.

Sitong Liu, Zhichao Lian, Siqi Gu et al.

Deepfake detection methods based on convolutional neural networks (CNN) have demonstrated high accuracy. \textcolor{black}{However, these methods often suffer from decreased performance when faced with unknown forgery methods and common transformations such as resizing and blurring, resulting in deviations between training and testing domains.} This phenomenon, known as overfitting, poses a significant challenge. To address this issue, we propose a novel block shuffling regularization method. Firstly, our approach involves dividing the images into blocks and applying both intra-block and inter-block shuffling techniques. This process indirectly achieves weight-sharing across different dimensions. Secondly, we introduce an adversarial loss algorithm to mitigate the overfitting problem induced by the shuffling noise. Finally, we restore the spatial layout of the blocks to capture the semantic associations among them. Extensive experiments validate the effectiveness of our proposed method, which surpasses existing approaches in forgery face detection. Notably, our method exhibits excellent generalization capabilities, demonstrating robustness against cross-dataset evaluations and common image transformations. Especially our method can be easily integrated with various CNN models. Source code is available at \href{https://github.com/NoWindButRain/BlockShuffleLearning}{Github}.

Sartaj Ahmed Salman, Zhichao Lian, Marva Saleem et al.

These days, computational diagnosis strategies of neuropsychiatric disorders are gaining attention day by day. It's critical to determine the brain's functional connectivity based on Functional-Magnetic-Resonance-Imaging(fMRI) to diagnose the disorder. It's known as a chronic disease, and millions of children amass the symptoms of this disease, so there is much vacuum for the researcher to formulate a model to improve the accuracy to diagnose ADHD accurately. In this paper, we consider the functional connectivity of a brain extracted using various time templates/Atlases. Local-Binary Encoding-Method (LBEM) algorithm is utilized for feature extraction, while Hierarchical- Extreme-Learning-Machine (HELM) is used to classify the extracted features. To validate our approach, fMRI data of 143 normal and 100 ADHD affected children is used for experimental purpose. Our experimental results are based on comparing various Atlases given as CC400, CC200, and AAL. Our model achieves high performance with CC400 as compared to other Atlases

Siqi Gu, Zhichao Lian

In this paper, a novel Unified Multi-Task Learning Framework of Real-Time Drone Supervision for Crowd Counting (MFCC) is proposed, which utilizes an image fusion network architecture to fuse images from the visible and thermal infrared image, and a crowd counting network architecture to estimate the density map. The purpose of our framework is to fuse two modalities, including visible and thermal infrared images captured by drones in real-time, that exploit the complementary information to accurately count the dense population and then automatically guide the flight of the drone to supervise the dense crowd. To this end, we propose the unified multi-task learning framework for crowd counting for the first time and re-design the unified training loss functions to align the image fusion network and crowd counting network. We also design the Assisted Learning Module (ALM) to fuse the density map feature to the image fusion encoder process for learning the counting features. To improve the accuracy, we propose the Extensive Context Extraction Module (ECEM) that is based on a dense connection architecture to encode multi-receptive-fields contextual information and apply the Multi-domain Attention Block (MAB) for concerning the head region in the drone view. Finally, we apply the prediction map to automatically guide the drones to supervise the dense crowd. The experimental results on the DroneRGBT dataset show that, compared with the existing methods, ours has comparable results on objective evaluations and an easier training process.

Jie Song, Liang Xiao, Mohsen Molaei et al.

In this paper, we propose an easily trained yet powerful representation learning approach with performance highly competitive to deep neural networks in a digital pathology image segmentation task. The method, called sparse coding driven deep decision tree ensembles that we abbreviate as ScD2TE, provides a new perspective on representation learning. We explore the possibility of stacking several layers based on non-differentiable pairwise modules and generate a densely concatenated architecture holding the characteristics of feature map reuse and end-to-end dense learning. Under this architecture, fast convolutional sparse coding is used to extract multi-level features from the output of each layer. In this way, rich image appearance models together with more contextual information are integrated by learning a series of decision tree ensembles. The appearance and the high-level context features of all the previous layers are seamlessly combined by concatenating them to feed-forward as input, which in turn makes the outputs of subsequent layers more accurate and the whole model efficient to train. Compared with deep neural networks, our proposed ScD2TE does not require back-propagation computation and depends on less hyper-parameters. ScD2TE is able to achieve a fast end-to-end pixel-wise training in a layer-wise manner. We demonstrated the superiority of our segmentation technique by evaluating it on the multi-disease state and multi-organ dataset where consistently higher performances were obtained for comparison against several state-of-the-art deep learning methods such as convolutional neural networks (CNN), fully convolutional networks (FCN), etc.