Fast Gradient Research Articles

Roadmap of Adversarial Machine Learning in Internet of Things-Enabled Security Systems.

Machine learning (ML) represents one of the main pillars of the current digital era, specifically in modern real-world applications. The Internet of Things (IoT) technology is foundational in developing advanced intelligent systems. The convergence of ML and IoT drives significant advancements across various domains, such as making IoT-based security systems smarter and more efficient. However, ML-based IoT systems are vulnerable to lurking attacks during the training and testing phases. An adversarial attack aims to corrupt the ML model's functionality by introducing perturbed inputs. Consequently, it can pose significant risks leading to devices' malfunction, services' interruption, and personal data misuse. This article examines the severity of adversarial attacks and accentuates the importance of designing secure and robust ML models in the IoT context. A comprehensive classification of adversarial machine learning (AML) is provided. Moreover, a systematic literature review of the latest research trends (from 2020 to 2024) of the intersection of AML and IoT-based security systems is presented. The results revealed the availability of various AML attack techniques, where the Fast Gradient Signed Method (FGSM) is the most employed. Several studies recommend the adversarial training technique to defend against such attacks. Finally, potential open issues and main research directions are highlighted for future consideration and enhancement.

Defence against adversarial attacks on IoT detection systems using deep belief network

An Adversarial attack is a technique used to deceive machine learning models to make incorrect predictions by providing slightly modified inputs from the original. Intrusion detection system (IDS) is a crucial tool in computer network security for the detection of adversarial attacks. Deep learning is a trending method in both research and industry, and this study proposes the use of a deep belief network (DBN). DBN can recognize data with small differences, but is also vulnerable to adversarial attacks. Therefore, this research suggests an internet of things-intrusion detection system (IoT-IDS) architecture using a DBN that can counter adversarial attacks. The chosen adversarial attack for this study is the fast gradient sign method (FGSM) used to evaluate the IoT IDS using the DBN model. Testing was conducted in two scenarios: first, the model was trained without adversarial attacks; second, the model was trained with adversarial attacks. The test results indicate that the DBN model struggles to detect FGSM attacks, achieving an accuracy of only 46% when it is not trained with adversarial attacks. However, after training with the FGSM dataset, the DBN model successfully detected adversarial attacks with an accuracy of 97%.

Negative order sobolev cubatures: preconditioners of partial differential equation learning tasks circumventing numerical stiffness

We present a variational approach aimed at enhancing the training of physics-informed neural networks (PINNs) and more general surrogate models for learning partial differential equations (PDE). In particular, we extend our formerly introduced notion of Sobolev cubatures to negative orders, enabling the approximation of negative order Sobolev norms. We mathematically prove the effect of negative order Sobolev cubatures in improving the condition number of discrete PDE learning problems, providing balancing scalars that mitigate numerical stiffness issues caused by loss imbalances. Additionally, we consider polynomial surrogate models (PSMs), which maintain the flexibility of PINN formulations while preserving the convexity structure of the PDE operators. The combination of negative order Sobolev cubatures and PSMs delivers well-conditioned discrete optimization problems, solvable via an exponentially fast convergent gradient descent for λ-convex losses. Our theoretical contributions are supported by numerical experiments, addressing linear and non-linear, forward and inverse PDE problems. These experiments show that the Sobolev cubature-based PSMs emerge as the superior state-of-the-art PINN technique.

VQCNN: variational quantum convolutional neural networks based on quantum filters and fully connected layers

Abstract Classical machine learning is more susceptible to adversarial examples due to its linear and non-robust nature, which results in a severe degradation of the recognition accuracy of classical machine learning models. Quantum techniques are shown to have a higher robustness advantage and are more resistant to attacks from adversarial examples than classical machine learning. Inspired by the robustness advantage of quantum computing and the feature extraction advantage of convolutional neural networks, this paper proposes a novel variational quantum convolutional neural network model (VQCNN), whose quantum fully connected layer consists of a combination of a quantum filter and a variational quantum neural network to increase the model’s adversarial robustness. The network intrusion detection model based on VQCNN is verified on KDD CUP99 and UNSW-NB datasets. The results show that under the attack of Fast Gradient Sign Method, the decline values of accuracy, precision, and recall rate of the intrusion detection model based on VQCNN are less than those of the other four models, and it has higher adversarial robustness.

Fast Gradient Algorithm with Dry-like Friction and Nonmonotone Line Search for Nonconvex Optimization Problems

Fast Gradient Algorithm with Dry-like Friction and Nonmonotone Line Search for Nonconvex Optimization Problems

DBI-Attack: Dynamic Bi-Level Integrated Attack for Intensive Multi-Scale UAV Object Detection

Benefiting from the robust feature representation capability of convolutional neural networks (CNNs), the object detection technology of intelligent high-altitude UAV remote sensing has been developed rapidly. In this field, the adversarial examples (AEs) pose serious security risks and vulnerabilities to deep learning-based systems. Due to the limitation of object size, image degradation, and scene brightness, adding adversarial disturbances to small and dense objects is extremely challenging. To study the threat of AE for UAV object detection, a dynamic bi-level integrated attack (DBI-Attack) is proposed for intensive multi-scale UAV object detection. Firstly, we use the dynamic iterative attack (DIA) method to generate perturbation on the classification level by improving the momentum iterative fast gradient sign method (MIM). Secondly, the bi-level adversarial attack method (BAAM) is constructed to add global perturbation on the decision level for completing the white-box attack. Finally, the integrated black-box attack method (IBAM) is combined to realize the black-box mislabeling and fabrication attacks. We experiment on the real drone traffic vehicle detection datasets to better evaluate the attack effectiveness. The experimental results show that the proposed method can achieve mislabeling and fabrication attacks on the UAV object detectors in black-box conditions. Furthermore, the adversarial training is applied to improve the model robustness. This work aims to call more attention to the adversarial and defensive aspects of UAV target detection models.

A method for recovering adversarial samples with both adversarial attack forensics and recognition accuracy

Adversarial samples deceive machine learning models through small but elaborate modifications that lead to erroneous outputs. The severity of the adversarial sample problem has come to the forefront with the widespread use of machine learning in areas such as security systems, autonomous driving, speech recognition, finance, and medical diagnostics. Malicious attackers can use adversarial samples to circumvent security detection systems, interfere with autonomous driving perception, mislead speech recognition, defraud financial systems, and even cause medical diagnosis errors. The emergence of adversarial samples exposes the vulnerability of existing models and poses challenges for information tracing and forensics after the incident. The main goal of current adversarial sample restoration methods is to improve model robustness. Traditional approaches focus only on improving the model’s classification accuracy, ignoring the importance of adversarial information, which is crucial for understanding the attack mechanism and strengthening future defenses. To address this issue, we propose an adversarial sample restoration method based on the similarity between clean and adversarial sample blocks to balance the needs of adversarial forensics and recognition accuracy. We implement the Fast Gradient Sign Method (FGSM), Basic Iterative Method (BIM), and Momentum Iterative Attack (MIA) attacks on MNIST, F-MNIST, and EMNIST datasets and perform experimental validation. The results demonstrate that our restoration method significantly enhances the model’s classification accuracy across various datasets and attack scenarios. Comparative analysis shows that the restored samples maintain a high similarity with the original adversarial samples, proving the method’s effectiveness. In addition, we performed performance tests on pre- and post-recovery samples. Taking the MNIST dataset as an example, we observed that the model performance metrics, such as MAPE, MAE, RMSE, and VAPE, of the restored samples improved by 88%, 88%, 65%, and 82%, respectively, after using the FGSM attack. This indicates that our restoration method successfully preserves the information of the generation mechanism of the adversarial samples and improves the model’s performance. This approach balances forensic capability and prediction accuracy, demonstrates a new direction in adversarial sample research, and substantially impacts security defense in practical applications.

Exploring the cognitive assessment potential of MRI-based volumetric hippocampal segmentation in Parkinson's disease.

To investigate the potential of magnetic resonance imaging (MRI)-based total and segmental hippocampus volume analysis in the assessment of cognitive status in Parkinson's disease (PD). We divided participants into three groups Group A-Parkinson patients (Pp) with normal cognitive status (n=25), Group B-Pp with dementia (n=17), and Group C-healthy controls (n=37). Three-dimensional T1W Fast Spoiled Gradient Recalled Echo images were used for Volbrain hippocampus subfield segmentation. We used the "Winterburn" protocol, which divides the hippocampus into five segments, Cornu Ammonis (CA),CA2/CA3, CA4/dentate gyrus, stratum radiatum, lacunosum, and moleculare, and subiculum. A total of 79 participants were included in the study, consisting of 42 individuals with PD (64.2% male) and 37 healthy controls (54.1% male). The mean age of PD was 60.9±10.7 years and the mean age of control group was 59.27±12.3 years. Significant differences were found in total hippocampal volumes between Group A and B (p=.047. Statistically significant group differences were found in total, right, and left CA1 volumes (analysis of variance [ANOVA]: F(2,76)=8.098, p=.001; F(2,76)=7.628, p=.001; F(2,76)=5.084, p=.008, respectively), as well as in total subiculum volumes (ANOVA: F(2,76)=4.368, p=.016). Post hoc tests showed that total subiculum volume was significantly lower in individuals with normal cognitive status (0.474±0.116cm3) compared to healthy controls (0.578±0.151cm3, p=.013). Volumetric hippocampal MRI can be used to assess the cognitive status of Pp. Longitudinal studies that evaluate Pp who progress from normal cognition to dementia are required to establish a causal relationship.

Machine learning and magnetic resonance image texture analysis predicts accelerated lung function decline in ex-smokers with and without chronic obstructive pulmonary disease.

Our objective was to train machine-learning algorithms on hyperpolarized magnetic resonance imaging (MRI) datasets to generate models of accelerated lung function decline in participants with and without chronic-obstructive-pulmonary-disease. We hypothesized that hyperpolarized gas MRI ventilation, machine-learning, and multivariate modeling could be combined to predict clinically-relevant changes in forced expiratory volume in 1s ( ) across 3 years. Hyperpolarized MRI was acquired using a coronal Cartesian fast gradient recalled echo sequence with a partial echo and segmented using a k-means clustering algorithm. A maximum entropy mask was used to generate a region-of-interest for texture feature extraction using a custom-developed algorithm and the PyRadiomics platform. The principal component and Boruta analyses were used for feature selection. Ensemble-based and single machine-learning classifiers were evaluated using area-under-the-receiver-operator-curve and sensitivity-specificity analysis. We evaluated 88 ex-smoker participants with months follow-up data, 57 of whom (22 females/35 males, years) had negligible changes in and 31 participants (7 females/24 males, years) with worsening . In addition, 3/88 ex-smokers reported a change in smoking status. We generated machine-learning models to predict decline using demographics, spirometry, and texture features, with the later yielding the highest classification accuracy of 81%. The combined model (trained on all available measurements) achieved the overall best classification accuracy of 82%; however, it was not significantly different from the model trained on MRI texture features alone. For the first time, we have employed hyperpolarized MRI ventilation texture features and machine-learning to identify ex-smokers with accelerated decline in with 82% accuracy.

Motion robust coronary MR angiography using zigzag centric ky-kz trajectory and high-resolution deep learning reconstruction.

To develop a new MR coronary angiography (MRCA) technique by employing a zigzag fan-shaped centric ky-kz k-space trajectory combined with high-resolutiondeep learning reconstruction (HR-DLR). All imaging data were acquired from 12 healthy subjects and 2 patients using two clinical 3-T MR imagers, with institutional review board approval. Ten healthy subjects underwent both standard 3D fast gradient echo (sFGE) and centric ky-kz k-space trajectory FGE (cFGE) acquisitions to compare the scan time and image quality. Quantitative measures were also performed for signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) as well as sharpness of the vessel. Furthermore, the feasibility of the proposed cFGE sequence was assessed in two patients. For assessing the feasibility of the centric ky-kz trajectory, the navigator-echo window of a 30-mm threshold was applied in cFGE, whereas sFGE was applied using a standard 5-mm threshold. Image quality of MRCA using cFGE with HR-DLR and sFGE without HR-DLR was scored in a 5-point scale (non-diagnostic = 1, fair = 2, moderate = 3, good = 4, and excellent = 5). Image evaluation of cFGE, applying HR-DLR, was compared with sFGE without HR-DLR. Friedman test, Wilcoxon signed-rank test, or paired t tests were performed for the comparison of related variables. The actual MRCA scan time of cFGE with a 30-mm threshold was acquired in less than 5min, achieving nearly 100% efficiency, showcasing its expeditious and robustness. In contrast, sFGE was acquired with a 5-mm threshold and had an average scan time of approximately 15min. Overall image quality for MRCA was scored 3.3 for sFGE and 2.7 for cFGE without HR-DLR but increased to 3.6 for cFGE with HR-DLR and (p < 0.05). The clinical result of patients obtained within 5min showed good quality images in both patients, even with a stent, without artifacts. Quantitative measures of SNR, CNR, and sharpness of vessel presented higher in cFGE with HR-DLR. Our findings demonstrate a robust, time-efficient solution for high-quality MRCA, enhancing patient comfort and increasing clinical throughput.

Mitigating adversarial threats in deep CT image diagnosis models via a dual-stage inference-time defense

Artificial intelligence (AI), particularly deep learning (DL) and machine learning (ML) have revolutionized disease diagnosis using complex medical images such as X-rays and CT scans, significantly improving accuracy in identifying various medical conditions. However, concerns arise regarding the trustworthiness of these models due to their vulnerability to adversarial attacks, potentially leading to inaccurate predictions. In response, we propose an innovative pipeline defensive approach that integrates denoising techniques like Total Variation Minimization (TVM) and Non-local (NL) means, followed by a fuzzy logic-based image transformer called Fuzzy Image Transformations (FIT). Deployed during the inferencing phase as a preprocessing module, this strategy aims to fortify DL medical diagnosis models, addressing adversarial vulnerabilities and ensuring their reliability in critical healthcare applications. Our focus is narrowed down to COVID-19 diagnosis, where we initially developed a model for accurately classifying lung CT images, covering both normal and COVID-19 pneumonia cases. Our experimental procedures, employing the Stratified K-Fold cross-validation method with K=5, systematically evaluate the model’s susceptibility to benchmark adversarial attacks such as the Fast Gradient Sign Method (FGSM), revealing a significant drop in aggregate average metrics across all folds under adversarial conditions. Specifically, precision (P), recall (R), accuracy (A), and F1-score (F) metrics decrease from 95.86 %, 96.16 %, 95.85 %, and 96.00–12.54 %, 13.13 %, 12.63 %, and 12.83 %, respectively. Notably, our proposed pipeline defense approach demonstrates substantial efficacy in enhancing diagnostic models, with average P, R, A, and F metrics improving impressively to 92.65 %, 92.39 %, 92.65 %, and 92.51 %, respectively, when applied to the same adversarial images. These promising results underscore the potential of our proposed pipeline defense techniques to enhance the resilience and reliability of AI-based diagnostic systems.

Efficient Adversarial Attack Based on Moment Estimation and Lookahead Gradient

Adversarial example generation is a technique that involves perturbing inputs with imperceptible noise to induce misclassifications in neural networks, serving as a means to assess the robustness of such models. Among the adversarial attack algorithms, momentum iterative fast gradient sign Method (MI-FGSM) and its variants constitute a class of highly effective offensive strategies, achieving near-perfect attack success rates in white-box settings. However, these methods’ use of sign activation functions severely compromises gradient information, which leads to low success rates in black-box attacks and results in large adversarial perturbations. In this paper, we introduce a novel adversarial attack algorithm, NA-FGTM. Our method employs the Tanh activation function instead of the sign which can accurately preserve gradient information. In addition, it utilizes the Adam optimization algorithm as well as the Nesterov acceleration, which is able to stabilize gradient update directions and expedite gradient convergence. Above all, the transferability of adversarial examples can be enhanced. Through integration with data augmentation techniques such as DIM, TIM, and SIM, NA-FGTM can further improve the efficacy of black-box attacks. Extensive experiments on the ImageNet dataset demonstrate that our method outperforms the state-of-the-art approaches in terms of black-box attack success rate and generates adversarial examples with smaller perturbations.

Boosting transferability of adversarial samples via saliency distribution and frequency domain enhancement

Deep neural networks (DNNs) perform excellently in various vision tasks, however, they are susceptible to the adversarial samples. These samples are crafted by injecting subtle perturbations into benign images that are barely detectable to humans. Due to the differences between the substitute model and target model, the efficacy of black-box attack still yield suboptimal results. To address this issue, we propose a black-box attack approach, SD-FI2M, based on the momentum iterative fast gradient sign method, which can craft potent adversarial samples by inducing diverse inputs in the frequency domain and the saliency distribution of images. To enhance the diversity of input samples in frequency domain, a spectrum transformation technique utilizing the discrete cosine transform was used to craft adversarial samples with higher transferability. Further, Guided Grad-CAM was employed to obtain the saliency distribution of benign images, reducing indiscriminate damage to image features and alleviating the overfitting problem in adversarial samples. Extensive experiments on ImageNet have verified the superior attack performance and transferability of the proposed method.

Minimally Distorted Adversarial Images with a Step-Adaptive Iterative Fast Gradient Sign Method

The safety and robustness of convolutional neural networks (CNNs) have raised increasing concerns, especially in safety-critical areas, such as medical applications. Although CNNs are efficient in image classification, their predictions are often sensitive to minor, for human observers, invisible modifications of the image. Thus, a modified, corrupted image can be visually equal to the legitimate image for humans but fool the CNN and make a wrong prediction. Such modified images are called adversarial images throughout this paper. A popular method to generate adversarial images is backpropagating the loss gradient to modify the input image. Usually, only the direction of the gradient and a given step size were used to determine the perturbations (FGSM, fast gradient sign method), or the FGSM is applied multiple times to craft stronger perturbations that change the model classification (i-FGSM). On the contrary, if the step size is too large, the minimum perturbation of the image may be missed during the gradient search. To seek exact and minimal input images for a classification change, in this paper, we suggest starting the FGSM with a small step size and adapting the step size with iterations. A few decay algorithms were taken from the literature for comparison with a novel approach based on an index tracking the loss status. In total, three tracking functions were applied for comparison. The experiments show our loss adaptive decay algorithms could find adversaries with more than a 90% success rate while generating fewer perturbations to fool the CNNs.

Removing Adversarial Noise in X-ray Images via Total Variation Minimizationand Patch-Based Regularization for Robust DeepLearning-based Diagnosis.

Deep learning has significantly advanced the field of radiology-based disease diagnosis, offering enhanced accuracy and efficiency in detecting various medical conditions through the analysis of complex medical images such as X-rays. This technology's ability to discern subtle patterns and anomalies has proven invaluable for swift and accurate disease identification. The relevance of deep learning in radiology has been particularly highlighted during the COVID-19 pandemic, where rapid and accurate diagnosis is crucial for effective treatment and containment. However, recent research has uncovered vulnerabilities in deep learning models when exposed to adversarial attacks, leading to incorrect predictions. In response to this critical challenge, we introduce a novel approach that leverages total variation minimization to combat adversarial noise within X-ray images effectively. Our focus narrows to COVID-19 diagnosis as a case study, where we initially construct a classification model through transfer learning designed to accurately classify lung X-ray images encompassing no pneumonia, COVID-19 pneumonia, and non-COVID pneumonia cases. Subsequently, we extensively evaluated the model's susceptibility to targeted and un-targetedadversarial attacksby employing the fast gradient sign gradient (FGSM) method. Our findings reveal a substantial reduction in the model's performance, with the average accuracy plummeting from 95.56 to 19.83% under adversarial conditions. However, the experimental results demonstrate the exceptional efficacy of the proposed denoising approach in enhancing the performance of diagnosis models when applied to adversarial examples. Post-denoising, the model exhibits a remarkable accuracy improvement, surging from 19.83 to 88.23% on adversarial images. These promising outcomes underscore the potential of denoising techniques to fortify the resilience and reliability of AI-based COVID-19 diagnostic systems, laying the foundation for their successful deployment in clinical settings.

Development of a high-throughput dual-stream liquid chromatography-tandem mass spectrometry method to screen for inhibitors of glutamate carboxypeptidase II.

Glutamate carboxypeptidase II (GCPII) catalyzes the hydrolysis of N-acetylaspartylglutamate (NAAG) to yield glutamate (Glu) and N-acetylaspartate (NAA). Inhibition of GCPII has been shown to remediate the neurotoxicity of excess Glu in a variety of cell and animal disease models. A robust high-throughput liquid chromatography-tandem mass spectrometry (LC/MS/MS) method was needed to quantify GCPII enzymatic activity in a biochemical high-throughput screening assay. A dual-stream LC/MS/MS method was developed. Two parallel eluent streams ran identical HILIC gradient methods on BEH-Amide (2×30 mm) columns. Each LC channel was run independently, and the cycle time was 2min per channel. Overall throughput was 1min per sample for the dual-channel integrated system. Multiply injected acquisition files were split during data review, and batch metadata were automatically paired with raw data during the review process. Two LC sorbents, BEH-Amide and Penta-HILIC, were tested to separate the NAAG cleavage product Glu from isobaric interference and ion suppressants in the bioassay matrix. Early elution of NAAG and NAA on BEH-Amide allowed interfering species to be diverted to waste. The limit of quantification was 0.1pmol for Glu. The Z-factor of this assay averaged 0.85. Over 36 000 compounds were screened using this method. A fast gradient dual-stream LC/MS/MS method for Glu quantification in GCPII biochemical screening assay samples was developed and validated. HILIC separation chemistry offers robust performance and unique selectivity for targeted positive mode quantification of Glu, NAA, and NAAG.

Fast connectivity gradient approximation: maintaining spatially fine-grained connectivity gradients while reducing computational costs

Brain connectome analysis suffers from the high dimensionality of connectivity data, often forcing a reduced representation of the brain at a lower spatial resolution or parcellation. This is particularly true for graph-based representations, which are increasingly used to characterize connectivity gradients, capturing patterns of systematic spatial variation in the functional connectivity structure. However, maintaining a high spatial resolution is crucial for enabling fine-grained topographical analysis and preserving subtle individual differences that might otherwise be lost. Here we introduce a computationally efficient approach to establish spatially fine-grained connectivity gradients. At its core, it leverages a set of landmarks to approximate the underlying connectivity structure at the full spatial resolution without requiring a full-scale vertex-by-vertex connectivity matrix. We show that this approach reduces computational time and memory usage while preserving informative individual features and demonstrate its application in improving brain-behavior predictions. Overall, its efficiency can remove computational barriers and enable the widespread application of connectivity gradients to capture spatial signatures of the connectome. Importantly, maintaining a spatially fine-grained resolution facilitates to characterize the spatial transitions inherent in the core concept of gradients of brain organization.

Efficient Isolation of Outer Membrane Vesicles (OMVs) Secreted by Gram-Negative Bacteria via a Novel Gradient Filtration Method.

Bacterial extracellular vesicles (bEVs) secreted by Gram-negative bacteria are referred to as outer membrane vesicles (OMVs) because they originate in the outer membrane. OMVs are membrane-coated vesicles 20-250 nm in size. They contain lipopolysaccharide (LPS), peptidoglycan, proteins, lipids, nucleic acids, and other substances derived from their parent bacteria and participate in the transmission of information to host cells. OMVs have broad prospects in terms of potential application in the fields of adjuvants, vaccines, and drug delivery vehicles. Currently, there remains a lack of efficient and convenient methods to isolate OMVs, which greatly limits OMV-related research. In this study, we developed a fast, convenient, and low-cost gradient filtration method to separate OMVs that can achieve industrial-scale production while maintaining the biological activity of the isolated OMVs. We compared the gradient filtration method with traditional ultracentrifugation to isolate OMVs from probiotic Escherichia coli Nissle 1917 (EcN) bacteria. Then, we used RAW264.7 macrophages as an in vitro model to study the influence on the immune function of EcN-derived OMVs obtained through the gradient filtration method. Our results indicated that EcN-derived OMVs were efficiently isolated using our gradient filtration method. The level of OMV enrichment obtained via our gradient filtration method was about twice as efficient as that achieved through traditional ultracentrifugation. The EcN-derived OMVs enriched through the gradient filtration method were successfully taken up by RAW264.7 macrophages and induced them to secrete pro-inflammatory cytokines such as tumor necrosis factor α (TNF-α) and interleukins (ILs) 6 and 1β, as well as anti-inflammatory cytokine IL-10. Furthermore, EcN-derived OMVs induced more anti-inflammatory response (i.e., IL-10) than pro-inflammatory response (i.e., TNF-α, IL-6, and IL-1β). These results were consistent with those reported in the literature. The related literature reported that EcN-derived OMVs obtained through ultracentrifugation could induce stronger anti-inflammatory responses than pro-inflammatory responses in RAW264.7 macrophages. Our simple and novel separation method may therefore have promising prospects in terms of applications involving the study of OMVs.

Explainable SE-MobileNet for Pneumonia detection integrated with robustness assessment using adversarial examples

Pneumonia is a detrimental disease, especially for children, which is caused due to bacterial infection. X-ray images are frequently observed manually to find out the existence of pneumonia in a patient’s body. However, diagnosing pneumonia using X-ray images through manual observation by different health professionals may lead to different conclusions. Thus, an efficient autonomic system is required to diagnose pneumonia from X-ray images, and deep learning techniques, such as CNN-based approaches are frequently used to create such autonomy. To ease the process of pneumonia diagnosis, in this study, we have proposed the SE-MobileNet approach. We compared the performance of the proposed SE-MobileNet with the default version of MobileNetV2 integrated with transfer learning. Using the publicly available Kaggle dataset, it is observed that SE-MobileNet obtained 97.4% accuracy on a select test set against the 96.4% accuracy of MobileNetV2, and in 10-fold cross-validation, SE-MobileNet achieved an average of 95.92% accuracy against the 92.35% accuracy of MobileNetV2. Further comparison analysis proves that the SE-MobileNet model not only performs much better than the vanilla MobileNetV2 but also performs competitively against the literature. In addition, robustness evaluation has been introduced in this study where Fast Gradient Sign Method (FGSM) is performed to generate adversarial images. It is found that in robustness evaluation, SE-MobileNet also performs better compared to MobileNetV2. Finally, to validate the appropriateness of the learning of the model, Explainable AI (XAI) based techniques have been employed.

Adversarial Attacks against Deep-Learning-Based Automatic Dependent Surveillance-Broadcast Unsupervised Anomaly Detection Models in the Context of Air Traffic Management.

Deep learning has shown significant advantages in Automatic Dependent Surveillance-Broadcast (ADS-B) anomaly detection, but it is known for its susceptibility to adversarial examples which make anomaly detection models non-robust. In this study, we propose Time Neighborhood Accumulation Iteration Fast Gradient Sign Method (TNAI-FGSM) adversarial attacks which fully take into account the temporal correlation of an ADS-B time series, stabilize the update directions of adversarial samples, and escape from poor local optimum during the process of iterating. The experimental results show that TNAI-FGSM adversarial attacks can successfully attack ADS-B anomaly detection models and improve the transferability of ADS-B adversarial examples. Moreover, the TNAI-FGSM is superior to two well-known adversarial attacks called the Fast Gradient Sign Method (FGSM) and Basic Iterative Method (BIM). To the best of our understanding, we demonstrate, for the first time, the vulnerability of deep-learning-based ADS-B time series unsupervised anomaly detection models to adversarial examples, which is a crucial step in safety-critical and cost-critical Air Traffic Management (ATM).