Adversarial Model Research Articles

3-D Skeleton-Based Human Motion Prediction With Manifold-Aware GAN

In this work we propose a novel solution for 3D skeleton-based human motion prediction. The objective of this task consists in forecasting future human poses based on a prior skeleton pose sequence. This involves solving two main challenges still present in recent literature; (1) discontinuity of the predicted motion which results in unrealistic motions and (2) performance deterioration in long-term horizons resulting from error accumulation across time. We tackle these issues by using a compact manifold-valued representation of 3D human skeleton motion. Specifically, we model the temporal evolution of the 3D poses as trajectory, what allows us to map human motions to single points on a sphere manifold. Using such a compact representation avoids error accumulation and provides robust representation for long-term prediction while ensuring the smoothness and the coherence of the whole motion. To learn these non-Euclidean representations, we build a manifold-aware Wasserstein generative adversarial model that captures the temporal and spatial dependencies of human motion through different losses. Experiments have been conducted on CMU MoCap and Human 3.6M datasets and demonstrate the superiority of our approach over the state-of-the-art both in short and long term horizons. The smoothness of the generated motion is highlighted in the qualitative results.

Domain and Patient Adversarial Multi-Task Learning for Arrhythmia Classification.

Manual screening of electrocardiograms (ECGs) for heart arrhythmias by clinicians is time-consuming and labor-intensive. A machine learning model for the automated diagnosis of heart arrhythmia from ECG signals can facilitate improved diagnosis, greater accessibility and earlier intervention for patients. The potential of such models is limited however by the small size of clinical datasets available for training. Methods that can be trained with multiple datasets to classify heart arrhythmia are needed to overcome this problem.In this paper, we propose using adversarial multi-task learning (AMTL) to extract domain and patient invariant features from two electrocardiogram databases. We further investigated the influence of beat segmentation location and beat normalization on domain invariance. Our proposed methods were tested on the MIT-BIH Arrhythmia and the St Petersburg INCART 12-lead Arrhythmia Databases. The domain adversarial models achieved a higher accuracy and average F1 score than their counterparts without domain adversarial learning. In particular, the patient and domain adversarial model improved the F1 scores on the two tested databases from 70% and 74% to 77% each.Clinical Relevance-This establishes that adversarial multitask learning with multiple datasets and multiple adversarial tasks can improve the F1 score of arrhythmia classification.

A provable secure and lightweight ECC-based authenticated key agreement scheme for edge computing infrastructure in smart grid

With the gradual maturity of Smart Grid (SG), security challenges become the most important issues that need to be addressed urgently. In recent years, many schemes adopt mutual authentication and key agreement to ensure secure communication in SG. However, most existing methods have their own shortcomings in either security or efficiency which make them difficult to satisfy the security requirements of SG. In this paper, we propose a provable secure and lightweight authenticated key agreement scheme based on the Elliptic Curve Cryptosystem (ECC) for the cloud-edge-end collaboration SG communication network. The proposed scheme can guarantee the security of the communication and provide anonymity for both the edge server and the smart meters. The anonymity of the smart meters can be provided by a pseudonym mechanism. By rigorous security analysis, the proposed scheme can resist to the typical attacks including replay attacks and identity forgery attacks. The security properties are also evaluated by using the Canetti and Krawczyk (CK) adversary model. The performance simulation results denotes that the proposed scheme not only owns stronger security functions but also improves computation efficiency by 24.5% on average than other four schemes. By simulation results, the proposed scheme has been shown to hold high efficiency.

Purifier: Defending Data Inference Attacks via Transforming Confidence Scores

Neural networks are susceptible to data inference attacks such as the membership inference attack, the adversarial model inversion attack and the attribute inference attack, where the attacker could infer useful information such as the membership, the reconstruction or the sensitive attributes of a data sample from the confidence scores predicted by the target classifier. In this paper, we propose a method, namely PURIFIER, to defend against membership inference attacks. It transforms the confidence score vectors predicted by the target classifier and makes purified confidence scores indistinguishable in individual shape, statistical distribution and prediction label between members and non-members. The experimental results show that PURIFIER helps defend membership inference attacks with high effectiveness and efficiency, outperforming previous defense methods, and also incurs negligible utility loss. Besides, our further experiments show that PURIFIER is also effective in defending adversarial model inversion attacks and attribute inference attacks. For example, the inversion error is raised about 4+ times on the Facescrub530 classifier, and the attribute inference accuracy drops significantly when PURIFIER is deployed in our experiment.

BNLoop-GAN: a multi-loop generative adversarial model on brain network learning to classify Alzheimer's disease.

Recent advancements in AI, big data analytics, and magnetic resonance imaging (MRI) have revolutionized the study of brain diseases such as Alzheimer's Disease (AD). However, most AI models used for neuroimaging classification tasks have limitations in their learning strategies, that is batch training without the incremental learning capability. To address such limitations, the systematic Brain Informatics methodology is reconsidered to realize evidence combination and fusion computing with multi-modal neuroimaging data through continuous learning. Specifically, we introduce the BNLoop-GAN (Loop-based Generative Adversarial Network for Brain Network) model, utilizing multiple techniques such as conditional generation, patch-based discrimination, and Wasserstein gradient penalty to learn the implicit distribution of brain networks. Moreover, a multiple-loop-learning algorithm is developed to combine evidence with better sample contribution ranking during training processes. The effectiveness of our approach is demonstrated through a case study on the classification of individuals with AD and healthy control groups using various experimental design strategies and multi-modal brain networks. The BNLoop-GAN model with multi-modal brain networks and multiple-loop-learning can improve classification performance.

TouchKey

Secure device pairing is important to wearables. Existing solutions either degrade usability due to the need of specific actions like shaking, or they lack universality due to the need of dedicated hardware like electrocardiogram sensors. This paper proposes TouchKey, a symmetric key generation scheme that exploits the skin electric potential (SEP) induced by powerline electromagnetic radiation. The SEP is ubiquitously accessible indoors with analog-to-digital converters widely available on Internet of Things devices. Our measurements show that the SEP has high randomness and the SEPs measured at two close locations on the same human body are similar. Extensive experiments show that TouchKey achieves a high key generation rate of 345 bit/s and an average success rate of 99.29%. Under a range of adversary models including active and passive attacks, TouchKey shows a low false acceptance rate of 0.86%, which outperforms existing solutions. Besides, the overall execution time and energy usage are 0.44 s and 2.716 mJ, which make it suitable for resource-constrained devices.

Guest Editorial: Privacy, Trust and Reputation Management in Internet of Vehicles (IoV)

The emergence of the Internet of vehicles (IoV) in recent years holds great potential for enhancing the user experience by introducing more sophisticated services that do everything from guaranteeing the user's safety to enhancing their comfort. Vehicles, people, and roadside infrastructure all play important roles in building I0V ecosystem. In addition, a wide variety of communication methods coexist to guarantee the IoV's connectivity and stability. This multiplies the attack surface of the ecosystem and introduces new privacy, trust, and reputation concerns as a direct result of its variety. Due to the widespread nature of the IoV, it is imperative that services be privy to and able to incorporate concerns over user privacy. To protect passengers, drivers, and pedestrians from the dangers of insider attacks—in which communications from authorized users are tampered with or manufactured by malevolent organizations. Multiple variables are utilized to define peer-to-peer exchanges, and this information is then used to make a well-informed decision about whether or not to trust the vehicle in question. One typical method of indicating how much certain characteristics matter is to use a weighting factor. The quantities of the weighting factors are often set manually because there are no influencing variables taken into consideration when doing so. In accordance with this computed criterion, the peer vehicles are then categorized as either trustworthy or not. The use of adversary models in conjunction with trust management models to counter different kinds of insider attacks has been the subject of study. Adversary models help in identifying potential attackers, while trust management models help in determining the level of access and privileges that an individual should have based on their trustworthiness. The combination of these two models can improve the security of a system against insider attacks. This Special Issue (SI) is an effort to tackle some of these issues. The SI includes contributions from experts in the field and aims to provide insights and solutions to the challenges faced in the area of privacy, trust and reputation in IoV. It also highlights potential solutions and areas for future research.

Radar Jamming Decision-Making in Cognitive Electronic Warfare: A Review

With the increasingly complex electromagnetic environment and the intelligent development of radar, the jammer, as opposed to radar, urgently needs to improve its ability to recognize threat targets and make jamming decisions. In this paper, we first introduce the concepts and systems of cognitive electronic warfare (CEW) and summarize its research status. Through analysis of the existing CEW systems, we propose a CEW model suitable for the cluster confrontation scenarios. Then, for the radar jamming decision-making (RJDM) namely a crucial part of CEW, we discuss the advantages, disadvantages, and applications of the traditional methods and analyze the machine-learning based methods including Markov decision processing, the newest Q-Learning, Deep Q-Learning (DQN), Double Deep Q-Learning (DDQN), A3C algorithms, and their improved algorithms etc. We build radar adversarial models and verify the effectiveness of reinforcement learning (RL) algorithm and the superiority of deep RL by simulating both the underlying Q-Learning and DQN algorithms. Finally, the research trends of CEW are discussed.

An Authenticated Group Shared Key Mechanism Based on a Combiner for Hash Functions over the Industrial Internet of Things

The Industrial Internet of Things (IIoT) provides internet connectivity for instruments, digital machines, and any other manufactured object to enable intelligent industrial operations to achieve high productivity. Securing communications between IIoT devices remains a critical and challenging issue due to the resource-constrained and processing capabilities of sensing devices. Moreover, the traditional group shared key might implement complex mathematical operations that are not suitable for the limited recourse capability of the IIoT device. Furthermore, the standard Diffie–Hellman (DH) and elliptic curve Diffie–Hellman (ECDH), which are the most suited for tiny devices, only work between a pair of IIoT devices, while they are not designed to work among a group of IIoT devices. This paper proposes an authenticated group shared key (AGSK) mechanism that allows a set of industrial objects to establish a common session key over the IIoT. The proposed AGSK utilizes the combiner for the hash function and digital signature, which is implemented in IIoT devices. Additionally, the random oracle model has been used to prove the security of AGSK, while the IIoT adversary model has been used to analyze the AGSK countermeasures against cyberattacks. The results of the performance evaluation showed that the efficiency of the AGSK was reduced by 41.3% for CPU computation time, 45.7% for storage cost, and 40% less power consumption compared to the baseline group key management algorithms.

Transformer-based image generation from scene graphs

Graph-structured scene descriptions can be efficiently used in generative models to control the composition of the generated image. Previous approaches are based on the combination of graph convolutional networks and adversarial methods for layout prediction and image generation, respectively. In this work, we show how employing multi-head attention to encode the graph information, as well as using a transformer-based model in the latent space for image generation can improve the quality of the sampled data, without the need to employ adversarial models with the subsequent advantage in terms of training stability.The proposed approach, specifically, is entirely based on transformer architectures both for encoding scene graphs into intermediate object layouts and for decoding these layouts into images, passing through a lower dimensional space learned by a vector-quantized variational autoencoder. Our approach shows an improved image quality with respect to state-of-the-art methods as well as a higher degree of diversity among multiple generations from the same scene graph. We evaluate our approach on three public datasets: Visual Genome, COCO, and CLEVR. We achieve an Inception Score of 13.7 and 12.8, and an FID of 52.3 and 60.3, on COCO and Visual Genome, respectively. We perform ablation studies on our contributions to assess the impact of each component. Code is available at https://github.com/perceivelab/trf-sg2im.

A Problem-tailored Adversarial Deep Neural Network-Based Attack Model for Feed-Forward Physical Unclonable Functions

With the exceeding advancement in technology, the sophistication of attacks is considerably increasing. Standard security methods fall short of achieving the security essentials of IoT against physical attacks due to the nature of IoTs being resource-constrained elements. Physical Unclonable Functions (PUFs) have been successfully employed as a lightweight memoryless solution to secure IoT devices. PUF is a device that exploits the integrated circuits’ inherent randomness originated during the fabrication process to give each physical entity a unique identifier. Nevertheless, because PUFs are vulnerable to mathematical clonability, Feed-Forward Arbiter PUF (FF PUF) was introduced to withstand potential attack methods. Motivated by the necessity to expose a critical vulnerability of the standard FF PUFs design, we introduce a problem-tailored adversarial model to attack FF PUF design using a carefully engineered loop-specific neural network-based design calibrated and trained using FPGA-based in-silicon implementation data to exhibit real-world attacking scenarios posed on FF PUFs, in addition to applying simulated data. The empirical results show that the proposed adversarial model adds outperforming results to the existing studies in attacking FF PUFs, manifesting the improved efficiency in breaking FF PUFs. We demonstrate our high-performing results in numerical experiments of language modeling using the deep Neural Networks method.

Deep learning techniques for biomedical data processing

The interest in Deep Learning (DL) has seen an exponential growth in the last ten years, producing a significant increase in both theoretical and applicative studies. On the one hand, the versatility and the ability to tackle complex tasks have led to the rapid and widespread diffusion of DL technologies. On the other hand, the dizzying increase in the availability of biomedical data has made classical analyses, carried out by human experts, progressively more unlikely. Contextually, the need for efficient and reliable automatic tools to support clinicians, at least in the most demanding tasks, has become increasingly pressing. In this survey, we will introduce a broad overview of DL models and their applications to biomedical data processing, specifically to medical image analysis, sequence processing (RNA and proteins) and graph modeling of molecular data interactions. First, the fundamental key concepts of DL architectures will be introduced, with particular reference to neural networks for structured data, convolutional neural networks, generative adversarial models, and siamese architectures. Subsequently, their applicability for the analysis of different types of biomedical data will be shown, in areas ranging from diagnostics to the understanding of the characteristics underlying the process of transcription and translation of our genetic code, up to the discovery of new drugs. Finally, the prospects and future expectations of DL applications to biomedical data will be discussed.

A Lightweight Authentication and Key Agreement Protocol for IoT-Enabled Smart Grid System.

The IoT-enabled Smart Grid uses IoT smart devices to collect the private electricity data of consumers and send it to service providers over the public network, which leads to some new security problems. To ensure the communication security in a smart grid, many researches are focusing on using authentication and key agreement protocols to protect against cyber attacks. Unfortunately, most of them are vulnerable to various attacks. In this paper, we analyze the security of an existent protocol by introducing an insider attacker, and show that their scheme cannot guarantee the claimed security requirements under their adversary model. Then, we present an improved lightweight authentication and key agreement protocol, which aims to enhance the security of IoT-enabled smart grid systems. Furthermore, we proved the security of the scheme under the real-or-random oracle model. The result shown that the improved scheme is secure in the presence of both internal attackers and external attackers. Compared with the original protocol, the new protocol is more secure, while keeping the same computation efficiency. Both of them are 0.0552 ms. The communication of the new protocol is 236 bytes, which is acceptable in smart grids. In other words, with similar communication and computation cost, we proposed a more secure protocol for smart grids.

Dual adversarial models with cross-coordination consistency constraint for domain adaption in brain tumor segmentation.

The brain tumor segmentation task with different domains remains a major challenge because tumors of different grades and severities may show different distributions, limiting the ability of a single segmentation model to label such tumors. Semi-supervised models (e.g., mean teacher) are strong unsupervised domain-adaptation learners. However, one of the main drawbacks of using a mean teacher is that given a large number of iterations, the teacher model weights converge to those of the student model, and any biased and unstable predictions are carried over to the student. In this article, we proposed a novel unsupervised domain-adaptation framework for the brain tumor segmentation task, which uses dual student and adversarial training techniques to effectively tackle domain shift with MR images. In this study, the adversarial strategy and consistency constraint for each student can align the feature representation on the source and target domains. Furthermore, we introduced the cross-coordination constraint for the target domain data to constrain the models to produce more confident predictions. We validated our framework on the cross-subtype and cross-modality tasks in brain tumor segmentation and achieved better performance than the current unsupervised domain-adaptation and semi-supervised frameworks.

Adversarial learning of permanent seismic deformation from GNSS coordinate timeseries

Deformation produced by an earthquake has a wide variety of forms. Therefore, there are a variety of models for quantifying the amount of deformation observed on GNSS coordinate timeseries, each of them based on different assumptions about the underlying mechanism that generates the data. Hence, it is of interest to look for methods relying on minimal assumptions about the observed position series.In this work we propose a semiparametric method, based on adversarial learning, to perform inference of the permanent seismic deformation. The only assumption made is that the probability distributions of GNSS fixed-length coordinate series for periods with and without observable seismic deformation differs mainly in the permanent deformation, represented by an additive scaled heaviside function.A dataset based on the series of GNSS coordinates published by the Nevada Geodetic Laboratory was built, and an adversarial model was trained over this dataset. In order to train the algorithm, an initial labeling of the samples was conducted using time, position an magnitude information of seismic events from the USGS database. It was shown that learning was possible with the available real data, and multiple sanity checks were run, showing consistency of the offset estimations compared with a trajectory model based estimator and with published offsets for well studied events on the South American active margin.To assess the capabilities of the method in a more controlled environment, further experiments were conducted on synthetic data. Those experiments confirmed that the presence of postseismic transient signals does not impede learning.As a derivative, our proposal allows to refine imperfect initial estimations for the presence/absence of seismic deformation.

Deep Learning Methods Used in Remote Sensing Images: A Review

Undeniably, Deep Learning (DL) has rapidly eroded traditional machine learning in Remote Sensing (RS) and geoscience domains with applications such as scene understanding, material identification, extreme weather detection, oil spill identification, among many others. Traditional machine learning algorithms are given less and less attention in the era of big data. Recently, a substantial amount of work aimed at developing image classification approaches based on the DL model's success in computer vision. The number of relevant articles has nearly doubled every year since 2015. Advances in remote sensing technology, as well as the rapidly expanding volume of publicly available satellite imagery on a worldwide scale, have opened up the possibilities for a wide range of modern applications. However, there are some challenges related to the availability of annotated data, the complex nature of data, and model parameterization, which strongly impact performance. In this article, a comprehensive review of the literature encompassing a broad spectrum of pioneer work in remote sensing image classification is presented including network architectures (vintage Convolutional Neural Network, CNN; Fully Convolutional Networks, FCN; encoder-decoder, recurrent networks; attention models, and generative adversarial models). The characteristics, capabilities, and limitations of current DL models were examined, and potential research directions were discussed.

HackGAN: Harmonious Cross-Network Mapping Using CycleGAN With Wasserstein–Procrustes Learning for Unsupervised Network Alignment

Network alignment (NA) that identifies equivalent nodes across networks is an effective tool for integrating knowledge from multiple networks. The state-of-the-art NA methods learn inter-network node similarities based on labeled anchor links, which are costly, time-consuming, and difficult to acquire. Therefore, a few unsupervised network alignment (UNA) methods propose solving NA problems without anchor links. However, most existing UNA methods rely on discriminative attributes to capture nodes’ similarities and are hard to obtain optimal one-to-one alignments. Toward these issues, this article proposes a novel method named HackGAN to solve the UNA problem solely based on the structural information. Specifically, HackGAN represents nodes with embeddings based on an unsupervised graph neural network (GNN) to capture their global and local structural features. After that, it initializes mapping functions to transform the embedding spaces of different networks into the same vector space by iteratively solving the Wasserstein–Procrustes problem. The mapping functions are then refined by an adversarial model with cycle-consistency and Sinkhorn distance losses to obtain optimized one-to-one mappings. Based on the distances between mapped embeddings, accurate and robust results are obtained with a collective alignment algorithm. Experimental comparisons on both synthetic and real-world datasets demonstrate the superiority of HackGAN.

ShieldDB: An Encrypted Document Database With Padding Countermeasures

Cloud storage systems have seen a growing number of clients due to the fact that more and more businesses and governments are shifting away from in-house data servers and seeking cost-effective and ease-of-access solutions. However, the security of cloud storage is underestimated in current practice, which resulted in many large-scale data breaches. To change the status quo, this paper presents the design of ShieldDB, an encrypted document database. ShieldDB adapts the searchable encryption technique to preserve the search functionality over encrypted documents without having much impact on its scalability. However, merely realising such a theoretical primitive suffers from real-world threats, where a knowledgeable adversary can exploit the leakage (aka access pattern to the database) to break the claimed protection on data confidentiality. To address this challenge in practical deployment, ShieldDB is designed with tailored padding countermeasures. Unlike prior works, we target a more realistic adversarial model, where the database gets updated continuously, and the adversary can monitor it at an (or multiple) arbitrary time interval(s). ShieldDB’s padding strategies ensure that the access pattern to the database is obfuscated all the time. We present a full-fledged implementation of ShieldDB and conduct intensive evaluations on Azure Cloud.

A novel hybrid distance guided domain adversarial method for cross domain fault diagnosis of gearbox

Distance-based domain adaptation methods have received extensive application in the transfer learning field. Different domain distances have different characteristics due to various data processing principles. Therefore, choosing appropriate domain distance can accomplish transfer tasks more efficiently. Domain adversarial neural networks can extract domain invariant features through game confrontation, but it is not capable of extracting hidden features of gear under speed fluctuations, and only using the adversarial mechanism for domain feature alignment is prone to gradient collapse. To solve the above problems, a novel hybrid distance guided domain adversarial fault diagnosis method of gear is proposed. First, stacked sparse autoencoders is employed in the model to extract the hidden features from the domain data, and the extracted features are input into the corresponding feature classifier and domain discriminator. Then, a mixture of maximum mean discrepancy (MMD) and Wasserstein distance is utilized to reduce the distribution difference. Finally, the domain adversarial mechanism is used to conduct adversarial training for feature alignment. Through two verification experiments of planetary gearboxes, it is verified that the proposed a Wasserstein and MMD distance guided Domain Adversarial model has excellent fault diagnosis performance under gear fluctuating conditions. In addition, the model has higher prediction accuracy and better fault feature extraction ability compared with other methods.

CLONE NODE DETECTION AND RECTIFICATION OF WIRELESS SENSOR NETWORK

Abstract-A wireless sensor network is a collection of nodes organized into a cooperative network. Each node consists of processing capability, multiple types of memory (program, data and flash memories), a RF transceiver, a power source (e.g., batteries and solar cells), and accommodate various sensors and actuators. Sensor nodes that are deployed in hostile environments are vulnerable to capture and compromise. An adversary may obtain private information from these sensors, clone and intelligently deploy them in the network to launch a variety of insider attacks. The defenses against clone attacks are not only very few, but also suffer from selective interruption of detection and high overhead (computation and memory).A few distributed solutions to address this fundamental problem are not satisfactory. First, they are energy and memory demanding to be used in the WSN- resource constrained environment. Further, they are vulnerable to the certain adversary models. Hence we make the solutions of this work in threefold. First, we analyze the desirable properties of a distributed mechanism for the detection of node replication attacks. Second, we show that the known solutions for this problem do not completely meet our requirements. Third, we propose a new self-healing, distributed hash table DHT-Based Protocol for the detection of node replication attacks, and we show that it satisfies the introduced requirements. Finally, extensive simulations show that our protocol is highly efficient in communication, memory, and computation. Keywords: distributed hash table, tamper-proof, Replication attack