Unseen Attacks Research Articles

Efficient Attention Branch Network with Combined Loss Function for Automatic Speaker Verification Spoof Detection.

Many endeavors have sought to develop countermeasure techniques as enhancements on Automatic Speaker Verification (ASV) systems, in order to make them more robust against spoof attacks. As evidenced by the latest ASVspoof 2019 countermeasure challenge, models currently deployed for the task of ASV are, at their best, devoid of suitable degrees of generalization to unseen attacks. A joint improvement of components of ASV spoof detection systems including the classifier, feature extraction phase, and model loss function may lead to a better detection of attacks by these systems. Accordingly, the present study proposes the Efficient Attention Branch Network (EABN) architecture with a combined loss function to address the model generalization to unseen attacks. The EABN is based on attention and perception branches. The attention branch provides an attention mask that improves the classification performance and at the same time is interpretable from a human point of view. The perception branch, is used for our main purpose which is spoof detection. The new EfficientNet-A0 architecture was optimized and employed for the perception branch, with nearly ten times fewer parameters and approximately seven times fewer floating-point operations than the SE-Res2Net50 as the best existing network. The proposed method on ASVspoof 2019 dataset achieved EER = 0.86% and t-DCF = 0.0239 in the Physical Access (PA) scenario using the logPowSpec as the input feature extraction method. Furthermore, using the LFCC feature, and the SE-Res2Net50 for the perception branch, the proposed model achieved EER = 1.89% and t-DCF = 0.507 in the Logical Access (LA) scenario, which to the best of our knowledge, is the best single system ASV spoofing countermeasure method.

DEFAEK: Domain Effective Fast Adaptive Network for Face Anti-Spoofing

Existing deep learning based face anti-spoofing (FAS) or deepfake detection approaches usually rely on large-scale datasets and powerful networks with significant amount of parameters to achieve satisfactory performance. However, these make them resource-heavy and unsuitable for handheld devices. Moreover, they are limited by the types of spoof in the dataset they train on and require considerable training time. To produce a robust FAS model, they need large datasets covering the widest variety of predefined presentation attacks possible. Testing on new or unseen attacks or environments generally results in poor performance. Ideally, the FAS model should learn discriminative features that can generalize well even on unseen spoof types. In this paper, we propose a fast learning approach called Domain Effective Fast Adaptive nEt-worK (DEFAEK), a face anti-spoofing approach based on the optimization-based meta-learning paradigm that effectively and quickly adapts to new tasks. DEFAEK treats differences in an environment as domains and simulates multiple domain shifts during training. To further improve the effectiveness and efficiency of meta-learning, we adopt the metric learning in the inner loop update with careful sample selection. With extensive experiments on the challenging CelebA-Spoof and FaceForensics++ datasets, the evaluation results show that DEFAEK can learn cues independent of the environment with good generalization capability. In addition, the resulting model is lightweight following the design principle of modern lightweight network architecture and still generalizes well on unseen classes. In addition, we also demonstrate our model’s capabilities by comparing the numbers of parameters, FLOPS, and model performance with other state-of-the-art methods.

Open Set Recognition for Malware Traffic via Predictive Uncertainty

Existing machine learning-based malware traffic recognition techniques can effectively detect abnormal behaviors in the network. However, almost all of them focus on a closed-set scenario in which the data used for training and testing come from the same label space. Since sophisticated malware and advanced persistent threats are evolving, it is impossible to exhaust all attacks to train a complete recognition model under the existing technical conditions. Therefore, recognition in the real network is an open-set problem, i.e., the recognition system should identify unknown and unseen attacks at test time. In this paper, we propose an uncertainty-aware method to identify known malicious traffic accurately and handle unknown traffic effectively. This method employs predictive uncertainty in deep learning as an indicator for unknown class detection. The predictive uncertainty represents the confidence in neural network predictions. In particular, the Deep Evidence Malware Traffic Recognition (DEMTR) model is presented to provide the multi-classification probability and predictive uncertainty in open-set scenarios using evidential deep learning. We demonstrate the performance of DEMTR on the MCFP dataset. Experimental results indicate that the proposed model outperforms the baseline methods in accuracy and F1-score.

Ensemble adversarial training based defense against adversarial attacks for machine learning-based intrusion detection system

In this paper, a defence mechanism is proposed against adversarial attacks. The defence is based on an ensemble classifier that is adversarially trained. This is accomplished by generating adversarial attacks from four different attack methods, i.e., Jacobian-based saliency map attack (JSMA), projected gradient descent (PGD), momentum iterative method (MIM), and fast gradient signed method (FGSM). The adversarial examples are used to identify the robust machine-learning algorithms which eventually participate in the ensemble. The adversarial attacks are divided into seen and unseen attacks. To validate our work, the experiments are conducted using NSLKDD, UNSW-NB15 and CICIDS17 datasets. Grid search for the ensemble is used to optimise results. The parameter used for performance evaluations is accuracy, F1 score and AUC score. It is shown that an adversarially trained ensemble classifier produces better results.

Comprehensive Study in Open-Set Iris Presentation Attack Detection

Research in presentation attack detection (PAD) for iris recognition has largely moved beyond evaluation in “closed-set” scenarios, to emphasize ability to generalize to presentation attack types not present in the training data. This paper offers multiple contributions to understand and extend the state-of-the-art in open-set iris PAD. First, it describes the most authoritative evaluation to date of iris PAD. We have curated the largest publicly-available image dataset for this problem, drawing from 26 benchmarks previously released by various groups, and adding 150,000 images being released with this paper, to create a set of 450,000 images representing authentic iris and seven types of presentation attack instrument (PAI). We formulate a leave-one-PAI-out evaluation protocol, and show that even the best algorithms in the closed-set evaluations exhibit catastrophic failures on multiple attack types in the open-set scenario. This includes algorithms performing well in the most recent LivDet-Iris 2020 competition, which may come from the fact that the LivDet-Iris protocol emphasizes sequestered images rather than unseen attack types. Second, we evaluate the accuracy of five open-source iris presentation attack algorithms available today, one of which is newly-proposed in this paper, and build an ensemble method that beats the winner of the LivDet-Iris 2020 by a substantial margin. This paper demonstrates that closed-set iris PAD, when all PAIs are known during training, is a solved problem, with multiple algorithms showing very high accuracy, while open-set iris PAD, when evaluated correctly, is far from being solved. The newly-created dataset, new open-source algorithms, and evaluation protocol, all made publicly available with this paper, provide experimental artifacts that researchers can use to measure progress on this important problem.

Asymmetric Modality Translation for Face Presentation Attack Detection

Face presentation attack detection (PAD) is an essentialmeasure to protect face recognition systems from being spoofed by malicious users and has attracted great attention from both academia and industry. Although most of the existing methods can achieve desired performance to some extent, the generalization issue of face presentation attack detection under cross-domain settings (e.g., the setting of unseen attacks and varying illumination) remains to be solved. In this paper, we propose a novel framework based on asymmetric modality translation for face presentation attack detection in bi-modality scenarios. Under the framework, we establish connections between two modality images of genuine faces. Specifically, a novel modality fusion scheme is presented that the image of one modality is translated to the other one through an asymmetric modality translator, then fused with its corresponding paired image. The fusion result is fed as the input to a discriminator for inference. The training of the translator is supervised by an asymmetric modality translation loss. Besides, an illumination normalization module based on Pattern of Local Gravitational Force (PLGF) representation is used to reduce the impact of illumination variation. We conduct extensive experiments on three public datasets, which validate that our method is effective in detecting various types of attacks and achieves state-of-the-art performance under different evaluation protocols.

Interpolated Joint Space Adversarial Training for Robust and Generalizable Defenses.

Adversarial training (AT) is considered to be one of the most reliable defenses against adversarial attacks. However, models trained with AT sacrifice standard accuracy and do not generalize well to unseen attacks. Recent works show generalization improvement with adversarial samples under unseen threat models such as on-manifold threat model or neural perceptual threat model. However, the former requires exact manifold information while the latter requires algorithm relaxation. Motivated by these considerations, we propose a novel threat model called Joint Space Threat Model (JSTM), which exploits the underlying manifold information with Normalizing Flow, ensuring that the exact manifold assumption holds. Under JSTM, we develop novel adversarial attacks and defenses. Specifically, we propose the Robust Mixup strategy in which we maximize the adversity of the interpolated images and gain robustness and prevent overfitting. Our experiments show that Interpolated Joint Space Adversarial Training (IJSAT) achieves good performance in standard accuracy, robustness, and generalization. IJSAT is also flexible and can be used as a data augmentation method to improve standard accuracy and combined with many existing AT approaches to improve robustness. We demonstrate the effectiveness of our approach on three benchmark datasets, CIFAR-10/100, OM-ImageNet and CIFAR-10-C.

Robust Data-Driven Detection of Electricity Theft Adversarial Evasion Attacks in Smart Grids

Existing machine learning-based detectors of electricity theft cyberattacks are trained to detect only simple traditional types of cyberattacks while neglecting complex ones like evasion attacks. This paper analyzes the robustness of electricity theft detectors against evasion attacks. Such attacks decrease the reported electricity reading values and fool the electricity theft detectors by injecting adversarial samples. We propose strong evasion attacks that fool the benchmark detectors by iteratively generating adversarial samples based on an electricity reading and its neighboring readings. We study the impact of evasion attacks using white, gray, and black-box settings based on the attacker’s knowledge about the detector’s parameters or datasets. Our investigations revealed that the performance degradation of benchmark detectors is up to 35.8%, 26.9%, and 22.2% in white, gray, and black-box settings, respectively. To enhance the detection robustness, we propose an ensemble learning-based anomaly detector trained only on benign data to detect unseen attacks (traditional and evasion) by sequentially combining an attentive autoencoder, convolutional-recurrent, and feed forward neural networks. The proposed model offers a stable detection performance where the average degradation is only <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$0.7 - 3\%$ </tex-math></inline-formula> , <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$0.9 - 2.1\%$ </tex-math></inline-formula> , and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$0.4 - 1.7\%$ </tex-math></inline-formula> in white, gray, and black-box settings, respectively, with maximum adversarial sample injection levels.

Reliable Detection of Location Spoofing and Variation Attacks

Location spoofing is a critical attack in mobile communications. While several previous studies investigated the detection of location spoofing attacks, they are limited in their performance and lack the consideration of emerging attack variations. In this paper, we present a data-driven methodology for the reliable detection of location spoofing and its variations. To enhance the performance, we introduce and utilize a new set of features, which is differential in nature and enables the checking of the mobility constraints and inconsistency. Our comparison study with the previous research shows that the presented scheme using the new features significantly improves the accuracy and reliability of the detection against location spoofing attacks. To take the possibility of attack variations into account, we establish a set of scenarios manipulating coordinate data to create attack variants. Our experimental results confirm the feasibility and effectiveness of the new features for identifying diverse types of spoofing attacks and their variations, greatly improving the detection performance by up to 99.1% accuracy. Additionally, we present a profiling-based detection approach (building the detector referring only to legitimate coordinate data), to further extend resilience to previously unseen attacks as a means to zero-day detection. The evaluation result shows the potential of the profiling-based detector with comparable or even better performance than the supervised learning methods (requiring both legitimate and falsified data to construct the detector).

HH-NIDS: Heterogeneous Hardware-Based Network Intrusion Detection Framework for IoT Security

This study proposes a heterogeneous hardware-based framework for network intrusion detection using lightweight artificial neural network models. With the increase in the volume of exchanged data, IoT networks’ security has become a crucial issue. Anomaly-based intrusion detection systems (IDS) using machine learning have recently gained increased popularity due to their generation’s ability to detect unseen attacks. However, the deployment of anomaly-based AI-assisted IDS for IoT devices is computationally expensive. A high-performance and ultra-low power consumption anomaly-based IDS framework is proposed and evaluated in this paper. The framework has achieved the highest accuracy of 98.57% and 99.66% on the UNSW-NB15 and IoT-23 datasets, respectively. The inference engine on the MAX78000EVKIT AI-microcontroller is 11.3 times faster than the Intel Core i7-9750H 2.6 GHz and 21.3 times faster than NVIDIA GeForce GTX 1650 graphics cards, when the power drawn was 18mW. In addition, the pipelined design on the PYNQ-Z2 SoC FPGA board with the Xilinx Zynq xc7z020-1clg400c device is optimised to run at the on-chip frequency (100 MHz), which shows a speedup of 53.5 times compared to the MAX78000EVKIT.

Robustness assessment of complex networks using the idle network

By utilizing the information encoded in the connectivity of the nodes affected by an attack (idle network), the assessment of network robustness is systematically enhanced. Moreover, the idle network information provides generalizability to train models to estimate network robustness, even for previously unseen attacks.

Image-Based Approach to Intrusion Detection in Cyber-Physical Objects

Recently, approaches based on the transformation of tabular data into images have gained a lot of scientific attention. This is explained by the fact that convolutional neural networks (CNNs) have shown good results in computer vision and other image-based classification tasks. Transformation of features without spatial relations to images allows the application of deep neural networks to a wide range of analysis tasks. This paper analyzes existing approaches to feature transformation based on the conversion of the features of network traffic into images and discusses their advantages and disadvantages. The authors also propose an approach to the transformation of raw network packets into images and analyze its efficiency in the task of network attack detection in a cyber-physical object, including its robustness to novel and unseen attacks.

Face anti-spoofing using feature distilling and global attention learning

Face anti-spoofing (FAS) is essential to assure the security of face recognition systems. Recently, some deep learning based FAS methods have achieved promising results under intra-dataset testing. However, they often fail in generalizing to unseen attacks due to the failure of extracting intrinsic features from face images. In this paper, we propose an end-to-end FAS method which consists of an anti-interference feature distillation module, a global spatial attention learning module and a pyramid binary mask supervision module. The deep features from the pretrained ResNet34 network are first distilled at multiple levels to capture intrinsic information via removing interference of features. Then, the multi-level distilled features are further refined by using a global spatial learning mechanism. Finally, the pyramid pixel-wise supervision is assembled to boost performance. Extensive experimental results on five benchmark datasets show the superior performance of our proposed method on intra-dataset testing and on cross-dataset testing.

Leveraging siamese networks for one-shot intrusion detection model

The use of supervised Machine Learning (ML) to enhance Intrusion Detection Systems (IDS) has been the subject of significant research. Supervised ML is based upon learning by example, demanding significant volumes of representative instances for effective training and the need to retrain the model for every unseen cyber-attack class. However, retraining the models in-situ renders the network susceptible to attacks owing to the time-window required to acquire a sufficient volume of data. Although anomaly detection systems provide a coarse-grained defence against unseen attacks, these approaches are significantly less accurate and suffer from high false-positive rates. Here, a complementary approach referred to as “One-Shot Learning”, whereby a limited number of examples of a new attack-class is used to identify a new attack-class (out of many) is detailed. The model grants a new cyber-attack classification opportunity for classes that were not seen during training without retraining. A Siamese Network is trained to differentiate between classes based on pairs similarities, rather than features, allowing to identify new and previously unseen attacks. The performance of a pre-trained model to classify new attack-classes based only on one example is evaluated using three mainstream IDS datasets; CICIDS2017, NSL-KDD, and KDD Cup’99. The results confirm the adaptability of the model in classifying unseen attacks and the trade-off between performance and the need for distinctive class representations.

Style selective normalization with meta learning for test-time adaptive face anti-spoofing

Although supervised face anti-spoofing (FAS) methods have shown remarkable performance, they suffer from poor generalization to unseen attacks. Many existing methods employed domain adaptation (DA) or domain generalization (DG) techniques to reduce domain variations. However, previous works have yet to fully explore domain-specific style information within intermediate layers that can give knowledge about face attack styles (e.g., illumination, backgrounds, and materials). In this paper, we present a new framework, Meta Style Selective Normalization (MetaSSN) for test-time domain adaptive FAS. Specifically, we propose style selective normalization (SSN) that statistically estimates the domain-specific image style of individual domains. SSN facilitates adaptation of the network to the target image by selecting the optimal normalization parameters to reduce style discrepancy between source and target domain. Furthermore, we meticulously design the training strategy in a meta-learning pipeline to simulate test-time adaptation using the style selection process with virtual test domain, which can boost the adaptation capability. In contrast to the previous DA approaches, our framework is more practical since it does not necessitate additional auxiliary networks (e.g., domain adaptors) during training. To validate our method, we utilized public FAS datasets: CASIA-FASD, MSU-MFSD, Oulu NPU, and Idiap Replay Attack. In most assessments, our results demonstrate a significant gap in performance relative to conventional FAS methods.

Anomal-E: A self-supervised network intrusion detection system based on graph neural networks

This paper investigates graph neural networks (GNNs) applied for self-supervised intrusion and anomaly detection in computer networks. GNNs are a deep learning approach for graph-based data that incorporate graph structures into learning to generalise graph representations and output embeddings. As traffic flows in computer networks naturally exhibit a graph structure, GNNs are a suitable fit in this context. The majority of current implementations of GNN-based network intrusion detection systems (NIDSs) rely on labelled network traffic. This limits the volume and structure of input traffic and restricts the NIDSs’ potential to adapt to unseen attacks. These systems also rely on the use of node features, which may reduce the detection accuracy of these systems, as important edge (packet-level) information is not leveraged. To overcome these restrictions, we present Anomal-E, a GNN approach to intrusion and anomaly detection that leverages edge features and a graph topological structure in a self-supervised manner. This approach is, to the best of our knowledge, the first successful and practical approach to network intrusion detection that utilises network flows in a self-supervised, edge-leveraging GNN. Experimental results on two modern benchmark NIDS datasets display a significant improvement when using Anomal-E compared to raw features and other baseline algorithms. This additionally posits the potential Anomal-E has for intrusion detection on real-world network traffic.

Electricity Theft Detection in Smart Grids Using a Hybrid BiGRU-BiLSTM Model with Feature Engineering-Based Preprocessing.

In this paper, a defused decision boundary which renders misclassification issues due to the presence of cross-pairs is investigated. Cross-pairs retain cumulative attributes of both classes and misguide the classifier due to the defused data samples’ nature. To tackle the problem of the defused data, a Tomek Links technique targets the cross-pair majority class and is removed, which results in an affine-segregated decision boundary. In order to cope with a Theft Case scenario, theft data is ascertained and synthesized randomly by using six theft data variants. Theft data variants are benign class appertaining data samples which are modified and manipulated to synthesize malicious samples. Furthermore, a K-means minority oversampling technique is used to tackle the class imbalance issue. In addition, to enhance the detection of the classifier, abstract features are engineered using a stochastic feature engineering mechanism. Moreover, to carry out affine training of the model, balanced data are inputted in order to mitigate class imbalance issues. An integrated hybrid model consisting of Bi-Directional Gated Recurrent Units and Bi-Directional Long-Term Short-Term Memory classifies the consumers, efficiently. Afterwards, robustness performance of the model is verified using an attack vector which is subjected to intervene in the model’s efficiency and integrity. However, the proposed model performs efficiently on such unseen attack vectors.

A Domain Generalization Method Based on Hybrid Meta-Learning for Face Anti-Spoofing

&lt;p&gt;For face anti-spoofing, many methods have been proposed to improve the security of face recognition systems. Due to distribution discrepancies among different domains, it is difficult to seek a generalized space which can generalize well to unseen attacks. In this paper, we propose a framework based on meta-learning method to improve the generalization ability of face anti-spoofing. The feature extractor is trained with forcing the distribution of real faces more compact while the distribution of fake faces is more dispersed among domains. Then we add a hybrid-domain meta learner module to simulate multiple domain shift scenarios. Moreover, we add a refined triplet mining to constrain the distance between real faces and fake ones. Multiple gradient information is integrated to optimize the feature extractor and train the model with good generalization performance to unseen attacks of various scenarios. Extensive experiments on four public datasets show that our proposed method can get better generalization ability to unseen target domain compared with state-of-the-art methods. &lt;/p&gt; &lt;p&gt;&amp;nbsp;&lt;/p&gt;

Robust Network Intrusion Detection Through Explainable Artificial Intelligence (XAI)

In this letter, we present a two-stage pipeline for robust network intrusion detection. First, we implement an extreme gradient boosting (XGBoost) model to perform supervised intrusion detection, and leverage the SHapley Additive exPlanation (SHAP) framework to devise explanations of our model. In the second stage, we use these explanations to train an auto-encoder to distinguish between previously seen and unseen attacks. Experiments conducted on the NSL-KDD dataset show that our solution is able to accurately detect new attacks encountered during testing, while its overall performance is comparable to numerous state-of-the-art works from the cybersecurity literature.

FS-IDS: A framework for intrusion detection based on few-shot learning

Due to the high dependency of traditional intrusion detection method on a fully-labeled large dataset, existing works can hardly be applied in real-world scenarios, especially facing zero-day attacks. In this paper we present a novel intrusion detection framework called “FS-IDS”, including flow data encoding method, feature fusion mechanism and architecture of intrusion detection system based on few-shot learning. We utilize task generator to split the dataset into separate tasks and train model in an episodic way, hoping model to learn general knowledge rather than those specific to a single class. The extraction module and distance metric module are responsible for learning and determining whether the traffic data are benign or not. We conduct three sets of experiments on “FS-IDS”, i.e., comparison study, ablation study and multiclass study. Comparison study firstly determines that the best measure metric for discrimination is Euclidean distance. Based on the optimal implementation, “FS-IDS” achieves comparable performance with existing works by using much fewer malicious samples. Ablation study sets two base models to explore how proposed encoding method and feature fusion mechanism improve detection capacity. Both the image representation and feature fusion achieve more than 2% improvement in accuracy and recall. Finally, to test whether “FS-IDS” can perform well under real-world scenario or not, we design network traffic containing various attacks to simulate complex malicious network environment. Experimental results show that “FS-IDS” maintains more than 90% detection accuracy and recall under the worst circumstances, which composes of various seen or unseen attacks with only a few malicious samples available.