Search

Deep models are being applied in numerous fields and have become a new important digital product. Meanwhile, previous studies have shown that deep models are vulnerable to backdoor attacks, in which compromised models return attacker-desired results when a trigger appears. Backdoor attacks severely break the trust-worthiness of deep models. In this paper, we turn this weakness of deep models into a strength, and propose a novel revocable backdoor and deep model trading scenario. Specifically, we aim to compromise deep models without degrading their performance, meanwhile, we can easily detoxify poisoned models without re-training the models. We design specific mask matrices to manage the internal feature maps of the models. These mask matrices can be used to deactivate the backdoors. The revocable backdoor can be adopted in the deep model trading scenario. Sellers train models with revocable backdoors as a trial version. Buyers pay a deposit to sellers and obtain a trial version of the deep model. If buyers are satisfied with the trial version, they pay a final payment to sellers and sellers send mask matrices to buyers to withdraw revocable backdoors. We demonstrate the feasibility and robustness of our revocable backdoor by various datasets and network architectures., Comment: to appear in ECAI 2024

Tip leakage flow in turbomachinery inherently generates intense unsteady features, named self-excited unsteadiness, which has not been well understood. A Zonalised LES (ZLES) is employed around a linear cascade, with wall-modelled Large Eddy Simulation (LES) forced to be active in the tip region. The simulation is well validated and demonstrated the advantages of effectively reducing the computational effort while maintaining an equivalent prediction accuracy in the region of interest. The time-averaged and spatial-spectral characteristics of tip leakage vortex (TLV) structures are systematically discussed. The self-excited unsteady processes of TLV include unsteady vortex separation inside the tip gap, the tip leakage jet-mainstream interaction, the primary tip leakage vortex (PTLV) wandering motion and the induced separation near endwall. The Spectral Proper Orthogonal Decomposition (SPOD) is used to examine the dominant frequencies and their coherent structures. It is found that these unsteady features change from a single high-frequency mode to a multiple lower-frequencies mode due to the PTLV breakdown. The SPOD and correlation analyses reveal that the self-excited unsteadiness is mainly induced by the interactions between unsteady vortex separation, tip leakage jet, and mainstream. The associated unsteady fluctuations are convected along the tip leakage jet trajectory, causing the wandering motion of PTLV core. Based on the cause of the investigated unsteadiness, a micro-offset tip design is proposed and validated for effectively suppressing this unsteadiness, and associated turbulence generation and hence pressure fluctuations. This work improves the understanding of tip-leakage-flow dynamics and informs the control of the associated unsteady fluid oscillation and noise.

Backdoor attack against image classification task has been widely studied and proven to be successful, while there exist little research on the backdoor attack against vision-language models. In this paper, we explore backdoor attack towards image captioning models by poisoning training data. Assuming the attacker has total access to the training dataset, and cannot intervene in model construction or training process. Specifically, a portion of benign training samples is randomly selected to be poisoned. Afterwards, considering that the captions are usually unfolded around objects in an image, we design an object-oriented method to craft poisons, which aims to modify pixel values by a slight range with the modification number proportional to the scale of the current detected object region. After training with the poisoned data, the attacked model behaves normally on benign images, but for poisoned images, the model will generate some sentences irrelevant to the given image. The attack controls the model behavior on specific test images without sacrificing the generation performance on benign test images. Our method proves the weakness of image captioning models to backdoor attack and we hope this work can raise the awareness of defending against backdoor attack in the image captioning field.

Training a deep neural network (DNN) requires a high computational cost. Buying models from sellers with a large number of computing resources has become prevailing. However, the buyer-seller environment is not always trusted. To protect the neural network models from leaking in an untrusted environment, we propose a novel copyright protection scheme for DNN using an input-sensitive neural network (ISNN). The main idea of ISNN is to make a DNN sensitive to the key and copyright information. Therefore, only the buyer with a correct key can utilize the ISNN. During the training phase, we add a specific perturbation to the clean images and mark them as legal inputs, while the other inputs are treated as illegal input. We design a loss function to make the outputs of legal inputs close to the true ones, while the illegal inputs are far away from true results. Experimental results demonstrate that the proposed scheme is effective, valid, and secure.

Recent generative models show impressive performance in generating photographic images. Humans can hardly distinguish such incredibly realistic-looking AI-generated images from real ones. AI-generated images may lead to ubiquitous disinformation dissemination. Therefore, it is of utmost urgency to develop a detector to identify AI generated images. Most existing detectors suffer from sharp performance drops over unseen generative models. In this paper, we propose a novel AI-generated image detector capable of identifying fake images created by a wide range of generative models. We observe that the texture patches of images tend to reveal more traces left by generative models compared to the global semantic information of the images. A novel Smash&Reconstruction preprocessing is proposed to erase the global semantic information and enhance texture patches. Furthermore, pixels in rich texture regions exhibit more significant fluctuations than those in poor texture regions. Synthesizing realistic rich texture regions proves to be more challenging for existing generative models. Based on this principle, we leverage the inter-pixel correlation contrast between rich and poor texture regions within an image to further boost the detection performance. In addition, we build a comprehensive AI-generated image detection benchmark, which includes 17 kinds of prevalent generative models, to evaluate the effectiveness of existing baselines and our approach. Our benchmark provides a leaderboard for follow-up studies. Extensive experimental results show that our approach outperforms state-of-the-art baselines by a significant margin. Our project: https://fdmas.github.io/AIGCDetect, Comment: Our project: https://fdmas.github.io/AIGCDetect/

Backdoor attack aims to compromise a model, which returns an adversary-wanted output when a specific trigger pattern appears yet behaves normally for clean inputs. Current backdoor attacks require changing pixels of clean images, which results in poor stealthiness of attacks and increases the difficulty of the physical implementation. This paper proposes a novel physical invisible backdoor based on camera imaging without changing nature image pixels. Specifically, a compromised model returns a target label for images taken by a particular camera, while it returns correct results for other images. To implement and evaluate the proposed backdoor, we take shots of different objects from multi-angles using multiple smartphones to build a new dataset of 21,500 images. Conventional backdoor attacks work ineffectively with some classical models, such as ResNet18, over the above-mentioned dataset. Therefore, we propose a three-step training strategy to mount the backdoor attack. First, we design and train a camera identification model with the phone IDs to extract the camera fingerprint feature. Subsequently, we elaborate a special network architecture, which is easily compromised by our backdoor attack, by leveraging the attributes of the CFA interpolation algorithm and combining it with the feature extraction block in the camera identification model. Finally, we transfer the backdoor from the elaborated special network architecture to the classical architecture model via teacher-student distillation learning. Since the trigger of our method is related to the specific phone, our attack works effectively in the physical world. Experiment results demonstrate the feasibility of our proposed approach and robustness against various backdoor defenses.

Backdoor attacks are rapidly emerging threats to deep neural networks (DNNs). In the backdoor attack scenario, attackers usually implant the backdoor into the target model by manipulating the training dataset or training process. Then, the compromised model behaves normally for benign input yet makes mistakes when the pre-defined trigger appears. In this paper, we analyze the drawbacks of existing attack approaches and propose a novel imperceptible backdoor attack. We treat the trigger pattern as a special kind of noise following a multinomial distribution. A U-net-based network is employed to generate concrete parameters of multinomial distribution for each benign input. This elaborated trigger ensures that our approach is invisible to both humans and statistical detection. Besides the design of the trigger, we also consider the robustness of our approach against model diagnose-based defences. We force the feature representation of malicious input stamped with the trigger to be entangled with the benign one. We demonstrate the effectiveness and robustness against multiple state-of-the-art defences through extensive datasets and networks. Our trigger only modifies less than 1\% pixels of a benign image while the modification magnitude is 1. Our source code is available at https://github.com/Ekko-zn/IJCAI2022-Backdoor., Comment: IJCAI 2022

Chaotic dynamical systems such as turbulent flows are characterized by an exponential divergence of infinitesimal perturbations to initial conditions. Therefore, conventional adjoint/tangent sensitivity analysis methods that are successful with RANS simulations fail in the case of chaotic LES/DNS. In this work, we discuss the limitations of current approaches, including ensemble-based and shadowing-based sensitivity methods, that were proposed as alternatives to conventional sensitivity analysis. We propose a new alternative, called the space-split sensitivity (S3) algorithm, that is computationally efficient and addresses these limitations. In this work, the derivation of the S3 algorithm is presented in the special case where the system converges to a stationary distribution that can be expressed with a probability density function everywhere in phase-space. Numerical examples of low-dimensional chaotic maps are discussed where S3 computation shows good agreement with finite-difference results, indicating potential for the development of the method in more generality., Comment: 10 pages; Center for turbulence research summer program 2018