Biblio

This Innovative Practice full paper describes a technical innovation for scalable teaching of cybersecurity hands-on classes using interactive learning environments. Hands-on experience significantly improves the practical skills of learners. However, the preparation and delivery of hands-on classes usually do not scale. Teaching even small groups of students requires a substantial effort to prepare the class environment and practical assignments. Further issues are associated with teaching large classes, providing feedback, and analyzing learning gains. We present our research effort and practical experience in designing and using learning environments that scale up hands-on cybersecurity classes. The environments support virtual networks with full-fledged operating systems and devices that emulate realworld systems. The classes are organized as simultaneous training sessions with cybersecurity assignments and learners' assessment. For big classes, with the goal of developing learners' skills and providing formative assessment, we run the environment locally, either in a computer lab or at learners' own desktops or laptops. For classes that exercise the developed skills and feature summative assessment, we use an on-premises cloud environment. Our approach is unique in supporting both types of deployment. The environment is described as code using open and standard formats, defining individual hosts and their networking, configuration of the hosts, and tasks that the students have to solve. The environment can be repeatedly created for different classes on a massive scale or for each student on-demand. Moreover, the approach enables learning analytics and educational data mining of learners' interactions with the environment. These analyses inform the instructor about the student's progress during the class and enable the learner to reflect on a finished training. Thanks to this, we can improve the student class experience and motivation for further learning. Using the presented environments KYPO Cyber Range Platform and Cyber Sandbox Creator, we delivered the classes on-site or remotely for various target groups of learners (K-12, university students, and professional learners). The learners value the realistic nature of the environments that enable exercising theoretical concepts and tools. The instructors value time-efficiency when preparing and deploying the hands-on activities. Engineering and computing educators can freely use our software, which we have released under an open-source license. We also provide detailed documentation and exemplary hands-on training to help other educators adopt our teaching innovations and enable sharing of reusable components within the community.

Modern technologies for searching viruses, cloud-edge computing, and also federated algorithms and machine learning architectures are shown. The architectures for searching malware based on the xor metric applied in the design and test of computing systems are proposed. A Federated ML method is proposed for searching for malware, which significantly speeds up learning without the private big data of users. A federated infrastructure of cloud-edge computing is described. The use of signature analysis and the assertion engine for searching malware is shown. The paradigm of LTF-computing for searching destructive components in software applications is proposed.

Natural language processing (NLP) is a division of artificial intelligence. The constructed model's quality is entirely reliant on the training dataset's quality. A data streaming pipeline is an adhesive application, completing a managed connection from data sources to machine learning methods. The recommended NLP pipeline composition has well-defined procedures. The implemented message broker design is a usual apparatus for delivering events. It makes it achievable to construct a robust training dataset for machine learning use-case and serve the model's input. The reconstructed dataset is a valid input for the machine learning processes. Based on the data pipeline's product, the model recreation and redeployment can be scheduled automatically.

With the development of open source projects, a large number of open source codes will be reused in binary software, and bugs in source codes will also be introduced into binary codes. In order to detect the reused open source codes in binary codes, it is sometimes necessary to compare and analyze the similarity between source codes and binary codes. One of the main challenge is that the compilation process can generate different binary code representations for the same source code, such as different compiler versions, compilation optimization options and target architectures, which greatly increases the difficulty of semantic similarity detection between source code and binary code. In order to solve the influence of the compilation process on the comparison of semantic similarity of codes, this paper transforms the source code and binary code into LLVM intermediate representation (LLVM IR), which is a universal intermediate representation independent of source code and binary code. We carry out semantic feature extraction and embedding training on LLVM IR based on natural language processing model. Experimental results show that LLVM IR eliminates the influence of compilation on the syntax differences between source code and binary code, and the semantic features of code are well represented and preserved.

Various methods for face validation-based authentication systems have been applied in a number of access control applications. However, using only one biometric factor such as facial data may limit accuracy and use, and is not practical in a real environment. This paper presents the implementation of a face time attendance system with an additional factor, a QR code to improve accuracy. This two- factor authentication system was developed in the form of a kiosk with a contactless process, which emerged due to the COVID-19 pandemic. The experiment was conducted at a well- known training event in Thailand. The proposed two-factor system was evaluated in terms of accuracy and satisfaction. Additionally, it was compared to a traditional single-factor system using only face recognition. The results confirm that the proposed two-factor scheme is more effective and did not incorrectly identify any users.

Despite recent advances in deep learning approaches, haze mitigation in large satellite images is still a challenging problem. Due to amorphous nature of haze, object detection or image segmentation approaches are not applicable. Also it is practically infeasible to obtain ground truths for training. Bounded memory capacity of GPUs is another constraint that limits the size of image to be processed. In this paper, we propose a style transfer based neural network approach to mitigate haze in a large overhead imagery. The network is trained without paired ground truths; further, perception loss is added to restore vivid colors, enhance contrast and minimize artifacts. The paper also illustrates our use of multiple GPUs in a collective way to produce a single coherent clear image where each GPU dehazes different portions of a large hazy image.

In the process of big data nearest neighbor similar data retrieval, affected by the way of data feature extraction, the retrieval accuracy is low. Therefore, this paper proposes the design of big data nearest neighbor similar data retrieval algorithm based on improved random forest. Through the improvement of random forest model and the construction of random decision tree, the characteristics of current nearest neighbor big data are clarified. Based on the improved random forest, the hash code is generated. Finally, combined with the Hamming distance calculation method, the nearest neighbor similar data retrieval of big data is realized. The experimental results show that: in the multi label environment, the retrieval accuracy is improved by 9% and 10%. In the single label environment, the similar data retrieval accuracy of the algorithm is improved by 12% and 28% respectively.

The traditional KNN algorithm takes insufficient consideration of the spatial distribution of training samples, which leads to low accuracy in processing high-dimensional data sets. Moreover, the generation of k nearest neighbors requires all known samples to participate in the distance calculation, resulting in high time overhead. To solve these problems, a feature subspace based KNN algorithm (Feature Subspace KNN, FSS-KNN) is proposed in this paper. First, the FSS-KNN algorithm solves all the feature subspaces according to the distribution of the training samples in the feature space, so as to ensure that the samples in the same subspace have higher similarity. Second, the corresponding feature subspace is matched for the test set samples. On this basis, the search of k nearest neighbors is carried out in the corresponding subspace first, thus improving the accuracy and efficiency of the algorithm. Experimental results show that compared with the traditional KNN algorithm, FSS-KNN algorithm improves the accuracy and efficiency on Kaggle data set and UCI data set. Compared with the other four classical machine learning algorithms, FSS-KNN algorithm can significantly improve the accuracy.

Based on the third generation neural network spiking neural network, this paper optimizes and improves a classification and coding method, and proposes an image recognition method. Firstly, the read image is converted into a spike sequence, and then the spike sequence is encoded in groups and sent to the neurons in the spike neural network. After learning and training for many times, the quantization standard code is obtained. In this process, the spike sequence transformation matrix and dynamic weight matrix are obtained, and the unclassified data are output through the same matrix for image recognition and classification. Simulation results show that the above methods can get correct coding and preliminary recognition classification, and the spiking neural network can be applied.

In this paper, we propose a deep learning based detection for multiple input multiple output (MIMO) physical-layer network coding (DeepPNC) over two way relay channels (TWRC). In MIMO-PNC, the relay node receives the signals superimposed from the two end nodes. The relay node aims to obtain the network-coded (NC) form of the two end nodes' signals. By training suitable deep neural networks (DNNs) with a limited set of training samples. DeepPNC can extract the NC symbols from the superimposed signals received while the output of each layer in DNNs converges. Compared with the traditional detection algorithms, DeepPNC has higher mapping accuracy and does not require channel information. The simulation results show that the DNNs based DeepPNC can achieve significant gain over the DeepNC scheme and the other traditional schemes, especially when the channel matrix changes unexpectedly.

For the intra-pulse modulation classification of radar signal, traditional deep learning algorithms have poor recognition performance without numerous training samples. Meanwhile, the receiver may intercept few pulse radar signals in the real scenes of electronic reconnaissance. To solve this problem, a structure which is made up of signal pretreatment by Smooth Pseudo Wigner-Ville (SPWVD) analysis algorithm, convolution neural network (CNN) and relation network (RN) is proposed in this study. The experimental results show that its classification accuracy is 94.24% under 20 samples per class training and the signal-to-noise ratio (SNR) is -4dB. Moreover, it can classify the novel types without further updating the network.

The exponential increase in the use of Internet of Things (IoT) devices has been accompanied by a spike in cyberattacks on IoT networks. In this research, we investigate the Bot-IoT dataset with a focus on classifying IoT reconnaissance attacks. Reconnaissance attacks are a foundational step in the cyberattack lifecycle. Our contribution is centered on the building of predictive models with the aid of Random Undersampling (RUS) and ensemble Feature Selection Techniques (FSTs). As far as we are aware, this type of experimentation has never been performed for the Reconnaissance attack category of Bot-IoT. Our work uses the Area Under the Receiver Operating Characteristic Curve (AUC) metric to quantify the performance of a diverse range of classifiers: Light GBM, CatBoost, XGBoost, Random Forest (RF), Logistic Regression (LR), Naive Bayes (NB), Decision Tree (DT), and a Multilayer Perceptron (MLP). For this study, we determined that the best learners are DT and DT-based ensemble classifiers, the best RUS ratio is 1:1 or 1:3, and the best ensemble FST is our ``6 Agree'' technique.

Most of existing signal modulation recognition methods attempt to establish a machine learning mechanism by training with a large number of annotated samples, which is hardly applied to the real-world electronic reconnaissance scenario where only a few samples can be intercepted in advance. Few-Shot Learning (FSL) aims to learn from training classes with a lot of samples and transform the knowledge to support classes with only a few samples, thus realizing model generalization. In this paper, a novel FSL framework called Attention Relation Network (ARN) is proposed, which introduces channel and spatial attention respectively to learn a more effective feature representation of support samples. The experimental results show that the proposed method can achieve excellent performance for fine-grained signal modulation recognition even with only one support sample and is robust to low signal-to-noise-ratio conditions.

Anomaly detection is the primary method of detecting intrusion. Unsupervised models, such as auto-encoders network, auto-encoder, and GMM, are currently the most widely used anomaly detection techniques. In reality, the samples used to train the unsupervised model may not be pure enough and may include some abnormal samples. However, the classification effect is poor since these approaches do not completely understand the association between reconstruction errors, reconstruction characteristics, and irregular sample density distribution. This paper proposes a novel intrusion detection system architecture that includes data collection, processing, and feature extraction by integrating data reconstruction features, reconstruction errors, auto-encoder parameters, and GMM. Our system outperforms other unsupervised learning-based detection approaches in terms of accuracy, recall, F1-score, and other assessment metrics after training and testing on multiple intrusion detection data sets.

In recent years, with the continuous development of artificial intelligence algorithms, their applications in network intrusion detection have become more and more widespread. However, as the network speed continues to increase, network traffic increases dramatically, and the drawbacks of traditional machine learning methods such as high false alarm rate and long training time are gradually revealed. CNN(Convolutional Neural Networks) can only extract spatial features of data, which is obviously insufficient for network intrusion detection. In this paper, we propose an intrusion detection model that combines CNN and BiSRU (Bi-directional Simple Recurrent Unit) to achieve the goal of intrusion detection by processing network traffic logs. First, we extract the spatial features of the original data using CNN, after that we use them as input, further extract the temporal features using BiSRU, and finally output the classification results by softmax to achieve the purpose of intrusion detection.

With the ubiquitous network applications and the continuous development of network attack technology, all social circles have paid close attention to the cyberspace security. Intrusion detection systems (IDS) plays a very important role in ensuring computer and communication systems security. Recently, deep learning has achieved a great success in the field of intrusion detection. However, the high computational complexity poses a major hurdle for the practical deployment of DL-based models. In this paper, we propose a novel approach based on a lightweight deep neural network (LNN) for IDS. We design a lightweight unit that can fully extract data features while reducing the computational burden by expanding and compressing feature maps. In addition, we use inverse residual structure and channel shuffle operation to achieve more effective training. Experiment results show that our proposed model for intrusion detection not only reduces the computational cost by 61.99% and the model size by 58.84%, but also achieves satisfactory accuracy and detection rate.

To improve the safety precaution level of network system, a combined network intrusion detection method is proposed based on Supervised Locality Preserving Projections (SLPP) feature extraction and Extreme Learning Machine (ELM). In this method, the feature extraction capability of SLPP is first used to reduce the dimensionality of the original network connection and system audit data, and get a feature set, then, based on this, the advantages of ELM in pattern recognition is adopted to build a network intrusion detection model for detecting and determining intrusion behavior. Simulation results show that, under the same experiment conditions, compared with traditional neural networks and support vector machines, the proposed method has more advantages in training efficiency and generalization performance.

This work seeks to advance the state of the art in HPC I/O performance analysis and interpretation. In particular, we demonstrate effective techniques to: (1) model output performance in the presence of I/O interference from production loads; (2) build features from write patterns and key parameters of the system architecture and configurations; (3) employ suitable machine learning algorithms to improve model accuracy. We train models with five popular regression algorithms and conduct experiments on two distinct production HPC platforms. We find that the lasso and random forest models predict output performance with high accuracy on both of the target systems. We also explore use of the models to guide adaptation in I/O middleware systems, and show potential for improvements of at least 15% from model-guided adaptation on 70% of samples, and improvements up to 10 x on some samples for both of the target systems.

Recent advances in natural language generation have introduced powerful language models with high-quality output text. However, this raises concerns about the potential misuse of such models for malicious purposes. In this paper, we study natural language watermarking as a defense to help better mark and trace the provenance of text. We introduce the Adversarial Watermarking Transformer (AWT) with a jointly trained encoder-decoder and adversarial training that, given an input text and a binary message, generates an output text that is unobtrusively encoded with the given message. We further study different training and inference strategies to achieve minimal changes to the semantics and correctness of the input text.AWT is the first end-to-end model to hide data in text by automatically learning -without ground truth- word substitutions along with their locations in order to encode the message. We empirically show that our model is effective in largely preserving text utility and decoding the watermark while hiding its presence against adversaries. Additionally, we demonstrate that our method is robust against a range of attacks.

To enhance the security of digital circuits, there is often a desire to dynamically generate, rather than statically store, random values used for identification and authentication purposes. Physically Unclonable Functions (PUFs) provide the means to realize this feature in an efficient and reliable way by utilizing commonly overlooked process variations that unintentionally occur during the manufacturing of integrated circuits (ICs) due to the imperfection of fabrication process. When given a challenge, PUFs produce a unique response. However, PUFs have been found to be vulnerable to modeling attacks where by using a set of collected challenge response pairs (CRPs) and training a machine learning model, the response can be predicted for unseen challenges. To combat this vulnerability, researchers have proposed techniques such as Challenge Obfuscation. However, as shown in this paper, this technique can be compromised via modeling the PUF's power side-channel. We first show the vulnerability of a state-of-the-art Challenge Obfuscated PUF (CO-PUF) against power analysis attacks by presenting our attack results on the targeted CO-PUF. Then we propose two countermeasures, as well as their hybrid version, that when applied to the CO-PUFs make them resilient against power side-channel based modeling attacks. We also provide some insights on the proper design metrics required to be taken when implementing these mitigations. Our simulation results show the high success of our attack in compromising the original Challenge Obfuscated PUFs (success rate textgreater 98%) as well as the significant improvement on resilience of the obfuscated PUFs against power side-channel based modeling when equipped with our countermeasures.

The popular federated edge learning (FEEL) framework allows privacy-preserving collaborative model training via frequent learning-updates exchange between edge devices and server. Due to the constrained bandwidth, only a subset of devices can upload their updates at each communication round. This has led to an active research area in FEEL studying the optimal device scheduling policy for minimizing communication time. However, owing to the difficulty in quantifying the exact communication time, prior work in this area can only tackle the problem partially by considering either the communication rounds or per-round latency, while the total communication time is determined by both metrics. To close this gap, we make the first attempt in this paper to formulate and solve the communication time minimization problem. We first derive a tight bound to approximate the communication time through cross-disciplinary effort involving both learning theory for convergence analysis and communication theory for per-round latency analysis. Building on the analytical result, an optimized probabilistic scheduling policy is derived in closed-form by solving the approximate communication time minimization problem. It is found that the optimized policy gradually turns its priority from suppressing the remaining communication rounds to reducing per-round latency as the training process evolves. The effectiveness of the proposed scheme is demonstrated via a use case on collaborative 3D objective detection in autonomous driving.

Collaborative learning has gained great popularity due to its benefit of data privacy protection: participants can jointly train a Deep Learning model without sharing their training sets. However, recent works discovered that an adversary can fully recover the sensitive training samples from the shared gradients. Such reconstruction attacks pose severe threats to collaborative learning. Hence, effective mitigation solutions are urgently desired.In this paper, we propose to leverage data augmentation to defeat reconstruction attacks: by preprocessing sensitive images with carefully-selected transformation policies, it becomes infeasible for the adversary to extract any useful information from the corresponding gradients. We design a novel search method to automatically discover qualified policies. We adopt two new metrics to quantify the impacts of transformations on data privacy and model usability, which can significantly accelerate the search speed. Comprehensive evaluations demonstrate that the policies discovered by our method can defeat existing reconstruction attacks in collaborative learning, with high efficiency and negligible impact on the model performance.

Deep learning (DL) models have been shown to be vulnerable to adversarial attacks. DL model security against adversarial attacks is critical to using DL-trained models in forward deployed systems, e.g. facial recognition, document characterization, or object detection. We provide results and lessons learned applying a moving target defense (MTD) strategy against iterative, gradient-based adversarial attacks. Our strategy involves (1) training a diverse ensemble of DL models, (2) applying randomized affine input transformations to inputs, and (3) randomizing output decisions. We report a primary lesson that this strategy is ineffective against a white-box adversary, which could completely circumvent output randomization using a deterministic surrogate. We reveal how our ensemble models lacked the diversity necessary for effective MTD. We also evaluate our MTD strategy against a black-box adversary employing an ensemble surrogate model. We conclude that an MTD strategy against black-box adversarial attacks crucially depends on lack of transferability between models.

The evolution of deep learning has promoted the popularization of smart devices. However, due to the insufficient development of computing hardware, the ability to conduct local training on smart devices is greatly restricted, and it is usually necessary to deploy ready-made models. This opacity makes smart devices vulnerable to deep learning backdoor attacks. Some existing countermeasures against backdoor attacks are based on the attacker’s ignorance of defense. Once the attacker knows the defense mechanism, he can easily overturn it. In this paper, we propose a Trojaning attack defense framework based on moving target defense(MTD) strategy. According to the analysis of attack-defense game types and confrontation process, the moving target defense model based on signaling game was constructed. The simulation results show that in most cases, our technology can greatly increase the attack cost of the attacker, thereby ensuring the availability of Deep Neural Networks(DNN) and protecting it from Trojaning attacks.

As the major factors affecting the safety of deep learning models, corner cases and related detection are crucial in AI quality assurance for constructing safety- and security-critical systems. The generic corner case researches involve two interesting topics. One is to enhance DL models' robustness to corner case data via the adjustment on parameters/structure. The other is to generate new corner cases for model retraining and improvement. However, the complex architecture and the huge amount of parameters make the robust adjustment of DL models not easy, meanwhile it is not possible to generate all real-world corner cases for DL training. Therefore, this paper proposes a simple and novel approach aiming at corner case data detection via a specific metric. This metric is developed on surprise adequacy (SA) which has advantages on capture data behaviors. Furthermore, targeting at characteristics of corner case data, three modifications on distanced-based SA are developed for classification applications in this paper. Consequently, through the experiment analysis on MNIST data and industrial data, the feasibility and usefulness of the proposed method on corner case data detection are verified.