Biblio

K-Nearest Neighbor (k-NN) search is one of the most commonly used approaches for similarity search. It finds extensive applications in machine learning and data mining. This era of big data warrants efficiently scaling k-NN search algorithms for billion-scale datasets with high dimensionality. In this paper, we propose a solution towards this end where we use vantage point trees for partitioning the dataset across multiple processes and exploit an existing graph-based sequential approximate k-NN search algorithm called HNSW (Hierarchical Navigable Small World) for searching locally within a process. Our hybrid MPI-OpenMP solution employs techniques including exploiting MPI one-sided communication for reducing communication times and partition replication for better load balancing across processes. We demonstrate computation of k-NN for 10,000 queries in the order of seconds using our approach on 8000 cores on a dataset with billion points in an 128-dimensional space. We also show 10X speedup over a completely k-d tree-based solution for the same dataset, thus demonstrating better suitability of our solution for high dimensional datasets. Our solution shows almost linear strong scaling.

Online social media plays an important role during real world events such as natural calamities, elections, social movements etc. Since the social media usage has increased, fake news has grown. The social media is often used by modifying true news or creating fake news to spread misinformation. The creation and distribution of fake news poses major threats in several respects from a national security point of view. Hence Fake news identification becomes an essential goal for enhancing the trustworthiness of the information shared on online social network. Over the period of time many researcher has used different methods, algorithms, tools and techniques to identify fake news content from online social networks. The aim of this paper is to review and examine these methodologies, different tools, browser extensions and analyze the degree of output in question. In addition, this paper discuss the general approach of fake news detection as well as taxonomy of feature extraction which plays an important role to achieve maximum accuracy with the help of different Machine Learning and Natural Language Processing algorithms.

This paper presents an anomalous IP address detection algorithm for network traffic logs. It is based on word embedding techniques derived from natural language processing to extract the representative features of IP addresses. However, the features extracted from vanilla word embeddings are not always compatible with machine learning-based anomaly detection algorithms. Therefore, we developed an algorithm that enables the extraction of more compatible features of IP addresses for anomaly detection than conventional methods. The proposed algorithm optimizes the objective functions of word embedding-based feature extraction and anomaly detection, simultaneously. According to the experimental results, the proposed algorithm outperformed conventional approaches; it improved the detection performance from 0.876 to 0.990 in the area under the curve criterion in a task of detecting the IP addresses of attackers from network traffic logs.

Social engineering attacks are well known attacks in the cyberspace and relatively easy to try and implement because no technical knowledge is required. In various online environments such as business domains where customers talk through a chat service with employees or in social networks potential hackers can try to manipulate other people by employing social attacks against them to gain information that will benefit them in future attacks. Thus, we have used a number of natural language processing steps and a machine learning algorithm to identify potential attacks. The proposed method has been tested on a semi-synthetic dataset and it is shown to be both practical and effective.

In modern grid systems which is a typical cyber-physical System (CPS), information space and physical space are closely related. Once the communication link is interrupted, it will make a great damage to the power system. If the service path is too concentrated, the risk will be greatly increased. In order to solve this problem, this paper constructs a route planning algorithm that combines node load pressure, link load balance and service delay risk. At present, the existing intelligent algorithms are easy to fall into the local optimal value, so we chooses the deep reinforcement learning algorithm (DRL). Firstly, we build a risk assessment model. The node risk assessment index is established by using the node load pressure, and then the link risk assessment index is established by using the average service communication delay and link balance degree. The route planning problem is then solved by a route planning algorithm based on DRL. Finally, experiments are carried out in a simulation scenario of a power grid system. The results show that our method can find a lower risk path than the original Dijkstra algorithm and the Constraint-Dijkstra algorithm.

Cyber-Physical Systems (CPS) are monitored and controlled by Supervisory Control and Data Acquisition (SCADA) systems that use advanced computing, sensors, control systems, and communication networks. At first, CPS and SCADA systems were protected and secured by isolation. However, with recent industrial technology advances, the increased connectivity of CPSs and SCADA systems to enterprise networks has uncovered them to new cybersecurity threats and made them a primary target for cyber-attacks with the potential of causing catastrophic economic, social, and environmental damage. Recent research focuses on new methodologies for risk modeling and assessment using game theory and reinforcement learning. This paperwork proposes to frame CPS security on two different levels, strategic and battlefield, by meeting ideas from game theory and Multi-Agent Reinforcement Learning (MARL). The strategic level is modeled as imperfect information, extensive form game. Here, the human administrator and the malware author decide on the strategies of defense and attack, respectively. At the battlefield level, strategies are implemented by machine learning agents that derive optimal policies for run-time decisions. The outcomes of these policies manifest as the utility at a higher level, where we aim to reach a Nash Equilibrium (NE) in favor of the defender. We simulate the scenario of a virus spreading in the context of a CPS network. We present experiments using the MiniCPS simulator and the OpenAI Gym toolkit and discuss the results.

DDoS attacks impose serious threats to many large or small organizations; therefore DDoS attacks have to be detected as soon as possible. In this study, a methodology to detect DDoS attacks is proposed and implemented on online systems. In the scope of the proposed methodology, Multi Layer Perceptron (MLP), Random Forest (RF), K-Nearest Neighbor (KNN), C-Support Vector Machine (SVC) machine learning methods are used with scaling and feature reduction preprocessing methods and then effects of preprocesses on detection accuracy rates of HTTP (Hypertext Transfer Protocol) flood, TCP SYN (Transport Control Protocol Synchronize) flood, UDP (User Datagram Protocol) flood and ICMP (Internet Control Message Protocol) flood DDoS attacks are analyzed. Obtained results showed that DDoS attacks can be detected with high accuracy of 99.2%.

Software Defined Networking (SDN) decouples the control plane and the data plane and solves the difficulty of new services deployment. However, the threat of a single point of failure is also introduced at the same time. The attacker can launch DDoS attacks towards the controller through switches. In this paper, a DDoS attack detection method based on information entropy and deep learning is proposed. Firstly, suspicious traffic can be inspected through information entropy detection by the controller. Then, fine-grained packet-based detection is executed by the convolutional neural network (CNN) model to distinguish between normal traffic and attack traffic. Finally, the controller performs the defense strategy to intercept the attack. The experiments indicate that the accuracy of this method reaches 98.98%, which has the potential to detect DDoS attack traffic effectively in the SDN environment.

In the area of cloud security, detection of DDoS attack is a challenging task such that legitimate users use the cloud resources properly. So in this paper, detection and classification of the attacking packets and normal packets are done by using various machine learning classifiers. We have selected the most relevant features from NSL KDD dataset using five (Information gain, gain ratio, chi-squared, ReliefF, and symmetrical uncertainty) commonly used feature selection methods. Now from the entire selected feature set, the most important features are selected by applying our hybrid feature selection method. Since all the anomalous instances of the dataset do not belong to DDoS category so we have separated only the DDoS packets from the dataset using the selected features. Finally, the dataset has been prepared and named as KDD DDoS dataset by considering the selected DDoS packets and normal packets. This KDD DDoS dataset has been discretized using discretize tool in weka for getting better performance. Finally, this discretize dataset has been applied on some commonly used (Naive Bayes, Bayes Net, Decision Table, J48 and Random Forest) classifiers for determining the detection rate of the classifiers. 10 fold cross validation has been used here for measuring the robustness of the system. To measure the efficiency of our hybrid feature selection method, we have also applied the same set of classifiers on the NSL KDD dataset, where it gives the best anomaly detection rate of 99.72% and average detection rate 98.47% similarly, we have applied the same set of classifiers on NSL DDoS dataset and obtain the average DDoS detection of 99.01% and the best DDoS detection rate of 99.86%. In order to compare the performance of our proposed hybrid method, we have also applied the existing feature selection methods and measured the detection rate using the same set of classifiers. Finally, we have seen that our hybrid approach for detecting the DDoS attack gives the best detection rate compared to some existing methods.

When using deep learning for DDoS attack detection, there is a general degradation in detection performance due to small sample size. This paper proposes a small-sample DDoS attack detection method based on deep transfer learning. First, deep learning techniques are used to train several neural networks that can be used for transfer in DDoS attacks with sufficient samples. Then we design a transferability metric to compare the transfer performance of different networks. With this metric, the network with the best transfer performance can be selected among the four networks. Then for a small sample of DDoS attacks, this paper demonstrates that the deep learning detection technique brings deterioration in performance, with the detection performance dropping from 99.28% to 67%. Finally, we end up with a 20.8% improvement in detection performance by deep transfer of the 8LANN network in the target domain. The experiment shows that the detection method based on deep transfer learning proposed in this paper can well improve the performance deterioration of deep learning techniques for small sample DDoS attack detection.

The paper presents a reinforcement learning solution to dynamic resource allocation for 5G radio access network slicing. Available communication resources (frequency-time blocks and transmit powers) and computational resources (processor usage) are allocated to stochastic arrivals of network slice requests. Each request arrives with priority (weight), throughput, computational resource, and latency (deadline) requirements, and if feasible, it is served with available communication and computational resources allocated over its requested duration. As each decision of resource allocation makes some of the resources temporarily unavailable for future, the myopic solution that can optimize only the current resource allocation becomes ineffective for network slicing. Therefore, a Q-learning solution is presented to maximize the network utility in terms of the total weight of granted network slicing requests over a time horizon subject to communication and computational constraints. Results show that reinforcement learning provides major improvements in the 5G network utility relative to myopic, random, and first come first served solutions. While reinforcement learning sustains scalable performance as the number of served users increases, it can also be effectively used to assign resources to network slices when 5G needs to share the spectrum with incumbent users that may dynamically occupy some of the frequency-time blocks.

Cross-site scripting (XSS) is one of the main threats of Web applications, which has great harm. How to effectively detect and defend against XSS attacks has become more and more important. Due to the malicious obfuscation of attack codes and the gradual increase in number, the traditional XSS detection methods have some defects such as poor recognition of malicious attack codes, inadequate feature extraction and low efficiency. Therefore, we present a novel approach to detect XSS attacks based on the attention mechanism of Long Short-Term Memory (LSTM) recurrent neural network. First of all, the data need to be preprocessed, we used decoding technology to restore the XSS codes to the unencoded state for improving the readability of the code, then we used word2vec to extract XSS payload features and map them to feature vectors. And then, we improved the LSTM model by adding attention mechanism, the LSTM-Attention detection model was designed to train and test the data. We used the ability of LSTM model to extract context-related features for deep learning, the added attention mechanism made the model extract more effective features. Finally, we used the classifier to classify the abstract features. Experimental results show that the proposed XSS detection model based on LSTM-Attention achieves a precision rate of 99.3% and a recall rate of 98.2% in the actually collected dataset. Compared with traditional machine learning methods and other deep learning methods, this method can more effectively identify XSS attacks.

Malware refers to any software accessing or being installed in a system without the authorisation of administrators. Various malware has been widely used for cyber-criminals to accomplish their evil intentions and goals. To combat the increasing amount and reduce the threat of malicious programs, a novel deep learning framework, which uses NLP techniques for reference, combines CNN and LSTM neurones to capture the locally spatial correlations and learn from sequential longterm dependency is proposed. Hence, high-level abstractions and representations are automatically extracted for the malware classification task. The classification accuracy improves from 0.81 (best one by Random Forest) to approximately 1.0.

This paper describes a system that comprises of control, safety and security subsystem for industries and homes. The entire system is based on the Bolt IoT platform. Using this system, the user can control the devices such as LEDs, speed of the fan or DC motor, monitor the temperature of the premises with an alert sub-system for critical temperatures through SMS and call, monitor the presence of anyone inside the premises with an alert sub-system about any intrusion through SMS and call. If the system is used specifically in any industry then instead of monitoring the temperature any other physical quantity, which is critical for that industry, can be monitored using suitable sensors. In addition, the cloud connectivity is provided to the system using the Bolt IoT module and temperature data is sent to the cloud where using machine-learning algorithm the future temperature is predicted to avoid any accidents in the future.

For real-time applications of arbitrary style transformation, there is a trade-off between the quality of results and the running time of existing algorithms. Hence, it is required to maintain the equilibrium of the quality of generated artwork with the speed of execution. It's complicated for the present arbitrary style-transformation procedures to preserve the structure of content-image while blending with the design and pattern of style-image. This paper presents the implementation of a network using SANET models for generating impressive artworks. It is flexible in the fusion of new style characteristics while sustaining the semantic-structure of the content-image. The identity-loss function helps to minimize the overall loss and conserves the spatial-arrangement of content. The results demonstrate that this method is practically efficient, and therefore it can be employed for real-time fusion and transformation using arbitrary styles.

Style transfer between Chinese fonts is challenging due to both the complexity of Chinese characters and the significant difference between fonts. Existing algorithms for this task typically learn a mapping between the reference and target fonts for each character. Subsequently, this mapping is used to generate the characters that do not exist in the target font. However, the characters available for training are unlikely to cover all fine-grained parts of the missing characters, leading to the overfitting problem. As a result, the generated characters of the target font may suffer problems of incomplete or even radicals and dirty dots. To address this problem, this paper presents a multi-task adversarial learning approach, termed MTfontGAN, to generate more vivid Chinese characters. MTfontGAN learns to transfer a reference font to multiple target ones simultaneously. An alignment is imposed on the encoders of different tasks to make them focus on the important parts of the characters in general style transfer. Such cross-task interactions at the feature level effectively improve the generalization capability of MTfontGAN. The performance of MTfontGAN is evaluated on three Chinese font datasets. Experimental results show that MTfontGAN outperforms the state-of-the-art algorithms in a single-task setting. More importantly, increasing the number of tasks leads to better performance in all of them.

Universal style transfer methods typically leverage rich representations from deep Convolutional Neural Network (CNN) models (e.g., VGG-19) pre-trained on large collections of images. Despite the effectiveness, its application is heavily constrained by the large model size to handle ultra-resolution images given limited memory. In this work, we present a new knowledge distillation method (named Collaborative Distillation) for encoder-decoder based neural style transfer to reduce the convolutional filters. The main idea is underpinned by a finding that the encoder-decoder pairs construct an exclusive collaborative relationship, which is regarded as a new kind of knowledge for style transfer models. Moreover, to overcome the feature size mismatch when applying collaborative distillation, a linear embedding loss is introduced to drive the student network to learn a linear embedding of the teacher's features. Extensive experiments show the effectiveness of our method when applied to different universal style transfer approaches (WCT and AdaIN), even if the model size is reduced by 15.5 times. Especially, on WCT with the compressed models, we achieve ultra-resolution (over 40 megapixels) universal style transfer on a 12GB GPU for the first time. Further experiments on optimization-based stylization scheme show the generality of our algorithm on different stylization paradigms. Our code and trained models are available at https://github.com/mingsun-tse/collaborative-distillation.

The prevalence of photo processing apps suggests the demands of picture editing. As an implementation of the convolutional neural network, style transfer has been deep investigated and there are supported materials to realize it on PC platform. However, few approaches are mentioned to deploy a style transfer model on the mobile and meet the requirements of mobile users. The traditional style transfer model takes hours to proceed, therefore, based on a Perceptual Losses algorithm [1], we created a feedforward neural network for each style and the proceeding time was reduced to a few seconds. The training data were generated from a pre-trained convolutional neural network model, VGG-19. The algorithm took thousandth time and generated similar output as the original. Furthermore, we optimized the model and deployed the model with TensorFlow Mobile library. We froze the model and adopted a bitmap to scale the inputs to 720×720 and reverted back to the original resolution. The reverting process may create some blur but it can be regarded as a feature of art. The generated images have reliable quality and the waiting time is independent of the content and pattern of input images. The main factor that influences the proceeding time is the input resolution. The average waiting time of our model on the mobile phone, HUAWEI P20 Pro, is less than 2 seconds for 720p images and around 2.8 seconds for 1080p images, which are ten times slower than that on the PC GPU, Tesla T40. The performance difference depends on the architecture of the model.

Advances in deep neural networks (DNNs) have shown tremendous promise in the medical domain. However, the deep learning tools that are helping the domain, can also be used against it. Given the prevalence of fraud in the healthcare domain, it is important to consider the adversarial use of DNNs in manipulating sensitive data that is crucial to patient healthcare. In this work, we present the design and implementation of a DNN-based image translation attack on biomedical imagery. More specifically, we propose Jekyll, a neural style transfer framework that takes as input a biomedical image of a patient and translates it to a new image that indicates an attacker-chosen disease condition. The potential for fraudulent claims based on such generated `fake' medical images is significant, and we demonstrate successful attacks on both X-rays and retinal fundus image modalities. We show that these attacks manage to mislead both medical professionals and algorithmic detection schemes. Lastly, we also investigate defensive measures based on machine learning to detect images generated by Jekyll.

Reinforcement learning (RL) algorithms have been demonstrated to be very attractive tools to train agents to achieve sequential tasks. However, these algorithms require too many training data to converge to be efficiently applied to physical robots. By using a human teacher, the learning process can be made faster and more robust, but the overall performance heavily depends on the quality and availability of teacher demonstrations or instructions. In particular, when these teaching signals are inadequate, the agent may fail to learn an optimal policy. In this paper, we introduce a trust-based interactive task learning approach. We propose an RL architecture able to learn both from environment rewards and from various sparse teaching signals provided by non-expert teachers, using an actor-critic agent, a human model and a trust model. We evaluate the performance of this architecture on 4 different setups using a maze environment with different simulated teachers and show that the benefits of the trust model.

Deep Neural Networks (DNNs) are susceptible to model stealing attacks, which allows a data-limited adversary with no knowledge of the training dataset to clone the functionality of a target model, just by using black-box query access. Such attacks are typically carried out by querying the target model using inputs that are synthetically generated or sampled from a surrogate dataset to construct a labeled dataset. The adversary can use this labeled dataset to train a clone model, which achieves a classification accuracy comparable to that of the target model. We propose "Adaptive Misinformation" to defend against such model stealing attacks. We identify that all existing model stealing attacks invariably query the target model with Out-Of-Distribution (OOD) inputs. By selectively sending incorrect predictions for OOD queries, our defense substantially degrades the accuracy of the attacker's clone model (by up to 40%), while minimally impacting the accuracy (\textbackslashtextless; 0.5%) for benign users. Compared to existing defenses, our defense has a significantly better security vs accuracy trade-off and incurs minimal computational overhead.

The concept of the adversary model has been widely applied in the context of cryptography. When designing a cryptographic scheme or protocol, the adversary model plays a crucial role in the formalization of the capabilities and limitations of potential attackers. These models further enable the designer to verify the security of the scheme or protocol under investigation. Although being well established for conventional cryptanalysis attacks, adversary models associated with attackers enjoying the advantages of machine learning techniques have not yet been developed thoroughly. In particular, when it comes to composed hardware, often being security-critical, the lack of such models has become increasingly noticeable in the face of advanced, machine learning-enabled attacks. This paper aims at exploring the adversary models from the machine learning perspective. In this regard, we provide examples of machine learning-based attacks against hardware primitives, e.g., obfuscation schemes and hardware root-of-trust, claimed to be infeasible. We demonstrate that this assumption becomes however invalid as inaccurate adversary models have been considered in the literature.

Artificial neural networks in general and deep learning networks in particular established themselves as popular and powerful machine learning algorithms. While the often tremendous sizes of these networks are beneficial when solving complex tasks, the tremendous number of parameters also causes such networks to be vulnerable to malicious behavior such as adversarial perturbations. These perturbations can change a model's classification decision. Moreover, while single-step adversaries can easily be transferred from network to network, the transfer of more powerful multi-step adversaries has - usually - been rather difficult.In this work, we introduce a method for generating strong adversaries that can easily (and frequently) be transferred between different models. This method is then used to generate a large set of adversaries, based on which the effects of selected defense methods are experimentally assessed. At last, we introduce a novel, simple, yet effective approach to enhance the resilience of neural networks against adversaries and benchmark it against established defense methods. In contrast to the already existing methods, our proposed defense approach is much more efficient as it only requires a single additional forward-pass to achieve comparable performance results.

Increasing deep neural networks are applied in solving graph evolved tasks, such as node classification and link prediction. However, the vulnerability of deep models can be revealed using carefully crafted adversarial examples generated by various adversarial attack methods. To explore this security problem, we define the link prediction adversarial attack problem and put forward a novel iterative gradient attack (IGA) strategy using the gradient information in the trained graph autoencoder (GAE) model. Not surprisingly, GAE can be fooled by an adversarial graph with a few links perturbed on the clean one. The results on comprehensive experiments of different real-world graphs indicate that most deep models and even the state-of-the-art link prediction algorithms cannot escape the adversarial attack, such as GAE. We can benefit the attack as an efficient privacy protection tool from the link prediction of unknown violations. On the other hand, the adversarial attack is a robust evaluation metric for current link prediction algorithms of their defensibility.

Machine Learning has revolutionized many fields of computer science. Reinforcement Learning (RL), in particular, stands out as a solution to sequential decision making problems. With the growing complexity of computer networks in the face of new emerging technologies, such as the Internet of Things and the growing complexity of threat vectors, there is a dire need for autonomous network systems. RL is a viable solution for achieving this autonomy. Software-defined Networking (SDN) provides a global network view and programmability of network behaviour, which can be employed for security management. Previous works in RL-based threat mitigation have mostly focused on very specific problems, mostly non-sequential, with ad-hoc solutions. In this paper, we propose ATMoS, a general framework designed to facilitate the rapid design of RL applications for network security management using SDN. We evaluate our framework for implementing RL applications for threat mitigation, by showcasing the use of ATMoS with a Neural Fitted Q-learning agent to mitigate an Advanced Persistent Threat. We present the RL model's convergence results showing the feasibility of our solution for active threat mitigation.