Biblio

Machine learning (ML) algorithms provide a good solution for many security sensitive applications, they themselves, however, face the threats of adversary attacks. As a key problem in machine learning, how to design robust feature selection algorithms against these attacks becomes a hot issue. The current researches on defending evasion attacks mainly focus on wrapped adversarial feature selection algorithm, i.e., WAFS, which is dependent on the classification algorithms, and time cost is very high for large-scale data. Since mRMR (minimum Redundancy and Maximum Relevance) algorithm is one of the most popular filter algorithms for feature selection without considering any classifier during feature selection process. In this paper, we propose a novel adversary-aware feature selection algorithm under filter model based on mRMR, named FAFS. The algorithm, on the one hand, takes the correlation between a single feature and a label, and the redundancy between features into account; on the other hand, when selecting features, it not only considers the generalization ability in the absence of attack, but also the robustness under attack. The performance of four algorithms, i.e., mRMR, TWFS (Traditional Wrapped Feature Selection algorithm), WAFS, and FAFS is evaluated on spam filtering and PDF malicious detection in the Perfect Knowledge attack scenarios. The experiment results show that FAFS has a better performance under evasion attacks with less time complexity, and comparable classification accuracy.

The network robustness is defined by how well its vertices are connected to each other to keep the network strong and sustainable. The change of network robustness may reveal events as well as periodic trend patterns that affect the interactions among vertices in the network. The evaluation of network robustness may be helpful to many applications, such as event detection, disease transmission, and network security, etc. There are many existing metrics to evaluate the robustness of networks, for example, node connectivity, edge connectivity, algebraic connectivity, graph expansion, R-energy, and so on. It is a natural and urgent problem how to choose a reasonable metric to effectively measure and evaluate the network robustness in the real applications. In this paper, based on some general principles, we design and implement a benchmark, namely BMNR, for the metrics of network robustness. The benchmark consists of graph generator, graph attack and robustness metric evaluation. We find that R-energy can evaluate both connected and disconnected graphs, and can be computed more efficiently.

AI-synthesized face-swapping videos, commonly known as DeepFakes, is an emerging problem threatening the trustworthiness of online information. The need to develop and evaluate DeepFake detection algorithms calls for datasets of DeepFake videos. However, current DeepFake datasets suffer from low visual quality and do not resemble DeepFake videos circulated on the Internet. We present a new large-scale challenging DeepFake video dataset, Celeb-DF, which contains 5,639 high-quality DeepFake videos of celebrities generated using improved synthesis process. We conduct a comprehensive evaluation of DeepFake detection methods and datasets to demonstrate the escalated level of challenges posed by Celeb-DF.

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.

Consensus is a fundamental approach to implementing fault-tolerant services through replication. It is well known that there exists a tradeoff between the cost and the resilience. For instance, Crash Fault Tolerant (CFT) protocols have a low cost but can only handle crash failures while Byzantine Fault Tolerant (BFT) protocols handle arbitrary failures but have a higher cost. Hybrid protocols enjoy the benefits of both high performance without failures and high resiliency under failures by switching among different subprotocols. However, it is challenging to determine which subprotocols should be used. We propose a moving target approach to switch among protocols according to the existing system and network vulnerability. At the core of our approach is a formalized cost model that evaluates the vulnerability and performance of consensus protocols based on real-time Intrusion Detection System (IDS) signals. Based on the evaluation results, we demonstrate that a safe, cheap, and unpredictable protocol is always used and a high IDS error rate can be tolerated.

In recent years, artificial intelligence has been widely used in the field of network security, which has significantly improved the effect of network security analysis and detection. However, because the power industrial control system is faced with the problem of shortage of attack data, the direct deployment of the network intrusion detection system based on artificial intelligence is faced with the problems of lack of data, low precision, and high false alarm rate. To solve this problem, we propose an anomaly traffic detection method based on cross-domain knowledge transferring. By using the TrAdaBoost algorithm, we achieve a lower error rate than using LSTM alone.

The traditional text classification methods usually follow this process: first, a sentence can be considered as a bag of words (BOW), then transformed into sentence feature vector which can be classified by some methods, such as maximum entropy (ME), Naive Bayes (NB), support vector machines (SVM), and so on. However, when these methods are applied to text classification, we usually can not obtain an ideal result. The most important reason is that the semantic relations between words is very important for text categorization, however, the traditional method can not capture it. Sentiment classification, as a special case of text classification, is binary classification (positive or negative). Inspired by the sentiment analysis, we use a novel deep learning-based recurrent neural networks (RNNs)model for automatic security audit of short messages from prisons, which can classify short messages(secure and non-insecure). In this paper, the feature of short messages is extracted by word2vec which captures word order information, and each sentence is mapped to a feature vector. In particular, words with similar meaning are mapped to a similar position in the vector space, and then classified by RNNs. RNNs are now widely used and the network structure of RNNs determines that it can easily process the sequence data. We preprocess short messages, extract typical features from existing security and non-security short messages via word2vec, and classify short messages through RNNs which accept a fixed-sized vector as input and produce a fixed-sized vector as output. The experimental results show that the RNNs model achieves an average 92.7% accuracy which is higher than SVM.

The existing anonymized differential privacy model adopts a unified anonymity method, ignoring the difference of personal privacy, which may lead to the problem of excessive or insufficient protection of the original data [1]. Therefore, this paper proposes a personalized k-anonymity model for tuples (PKA) and proposes a differential privacy data publishing algorithm (DPPA) based on personalized anonymity, firstly based on the tuple personality factor set by the user in the original data set. The values are classified and the corresponding privacy protection relevance is calculated. Then according to the tuple personality factor classification value, the data set is clustered by clustering method with different anonymity, and the quasi-identifier attribute of each cluster is aggregated and noise-added to realize anonymized differential privacy; finally merge the subset to get the data set that meets the release requirements. In this paper, the correctness of the algorithm is analyzed theoretically, and the feasibility and effectiveness of the proposed algorithm are verified by comparison with similar algorithms.

With the rapid development of computer networks, multimedia information is widely used, and the security of digital media has drawn much attention. The revised photo as a forensic evidence will distort the truth of the case badly tampered pictures on the social network can have a negative impact on the parties as well. In order to ensure the authenticity and integrity of digital media, self-recovery of digital images based on information hiding is studied in this paper. Jarvis half-tone change is used to compress the digital image and obtain the backup data, and then spread the backup data to generate the reference data. Hash algorithm aims at generating hash data by calling reference data and original data. Reference data and hash data together as a digital watermark scattered embedded in the digital image of the low-effective bits. When the image is maliciously tampered with, the hash bit is used to detect and locate the tampered area, and the image self-recovery is performed by extracting the reference data hidden in the whole image. In this paper, a thorough rebuild quality assessment of self-healing images is performed and better performance than the traditional DCT(Discrete Cosine Transform)quantization truncation approach is achieved. Regardless of the quality of the tampered content, a reference authentication system designed according to the principles presented in this paper allows higher-quality reconstruction to recover the original image with good quality even when the large area of the image is tampered.

In this paper, we propose a new method to expose AI-generated fake face images or videos (commonly known as the Deep Fakes). Our method is based on the observations that Deep Fakes are created by splicing synthesized face region into the original image, and in doing so, introducing errors that can be revealed when 3D head poses are estimated from the face images. We perform experiments to demonstrate this phenomenon and further develop a classification method based on this cue. Using features based on this cue, an SVM classifier is evaluated using a set of real face images and Deep Fakes.

The disadvantages and inextensibility of the traditional network require more novel thoughts for the future network architecture, as for ICN (Information-Centric Network), is an information centered and self-caching network, ICN is deeply rooted in the 5G era, of which concept is user-centered and content-centered. Although the ICN enables cache replacement of content, an information distribution scheduling algorithm is still needed to allocate resources properly due to its limited cache capacity. This paper starts with data popularity, information epilepsy and other data related attributes in the ICN environment. Then it analyzes the factors affecting the cache, proposes the concept and calculation method of Gene value. Since the ICN is still in a theoretical state, this paper describes an ICN scenario that is close to the reality and processes a greedy caching algorithm named GDGCA (Gene Driven Greedy Caching Algorithm). The GDGCA tries to design an optimal simulation model, which based on the thoughts of throughput balance and satisfaction degree (SSD), then compares with the regular distributed scheduling algorithm in related research fields, such as the QoE indexes and satisfaction degree under different Poisson data volumes and cycles, the final simulation results prove that GDGCA has better performance in cache scheduling of ICN edge router, especially with the aid of Information Gene value.

Nowadays, Vehicular ad hoc Network as a special class of Mobile ad hoc Network(MANET), provides plenty of services. However, it also brings the privacy protection issues, and there are conflicts between the privacy protection and the services. In this paper, we will propose a privacy protection algorithm based on group signature including two parts, group signature based anonymous verification and batch verification. The anonymous verification is based on the network model we proposed, which can reduce the trust authority burden by dividing the roadside units into different levels, and the batch verification can reduce the time of message verification in one group. We also prove our algorithm can satisfy the demand of privacy protection. Finally, the simulation shows that the algorithm we proposed is better than the BBS on the length of the signature, time delay and packet loss rate.

Due to the increasing concerns of securing private information, context-aware Internet of Things (IoT) applications are in dire need of supporting data privacy preservation for users. In the past years, game theory has been widely applied to design secure and privacy-preserving protocols for users to counter various attacks, and most of the existing work is based on a two-player game model, i.e., a user/defender-attacker game. In this paper, we consider a more practical scenario which involves three players: a user, an attacker, and a service provider, and such a complicated system renders any two-player model inapplicable. To capture the complex interactions between the service provider, the user, and the attacker, we propose a hierarchical two-layer three-player game framework. Finally, we carry out a comprehensive numerical study to validate our proposed game framework and theoretical analysis.

We report a an experimental study of device-independent quantum random number generation based on an detection-loophole free Bell test with entangled photons. After considering statistical fluctuations and applying an 80 Gb × 45.6 Mb Toeplitz matrix hashing, we achieve a final random bit rate of 114 bits/s, with a failure probability less than 10-5.

Since Gatys et al. proved that the convolution neural network (CNN) can be used to generate new images with artistic styles by separating and recombining the styles and contents of images. Neural Style Transfer has attracted wide attention of computer vision researchers. This paper aims to provide an overview of the style transfer application deep learning network development process, and introduces the classical style migration model, on the basis of the research on the migration of style of the deep learning network for collecting and organizing, and put forward related to gathered during the investigation of the problem solution, finally some classical model in the image style to display and compare the results of migration.

With the rapid development of Internet, the issue of cyber security has increasingly gained more attention. An intrusion Detection System (IDS) is an effective technique to defend cyber-attacks and reduce security losses. However, the challenge of IDS lies in the diversity of cyber-attackers and the frequently-changing data requiring a flexible and efficient solution. To address this problem, machine learning approaches are being applied in the IDS field. In this paper, we propose an efficient scalable neural-network-based hybrid IDS framework with the combination of Host-level IDS (HIDS) and Network-level IDS (NIDS). We applied the autoencoders (AE) to NIDS and designed HIDS using word embedding and convolutional neural network. To evaluate the IDS, many experiments are performed on the public datasets NSL-KDD and ADFA. It can detect many attacks and reduce the security risk with high efficiency and excellent scalability.

The Internet of Things enables interaction between IoT devices and users through the cloud. The cloud provides services such as account monitoring, device management, and device control. As the center of the IoT platform, the cloud provides services to IoT devices and IoT applications through APIs. Therefore, the permission verification of the API is essential. However, we found that some APIs are unverified, which allows unauthorized users to access cloud resources or control devices; it could threaten the security of devices and cloud. To check for unauthorized access to the API, we developed IoT-APIScanner, a framework to check the permission verification of the cloud API. Through observation, we found there is a large amount of interactive information between IoT application and cloud, which include the APIs and related parameters, so we can extract them by analyzing the code of the IoT application, and use this for mutating API test cases. Through these test cases, we can effectively check the permissions of the API. In our research, we extracted a total of 5 platform APIs. Among them, the proportion of APIs without permission verification reached 13.3%. Our research shows that attackers could use the API without permission verification to obtain user privacy or control of devices.

To meet the high requirement of human-machine interaction, quadruped robots with human recognition and tracking capability are studied in this paper. We first introduce a marker recognition system which uses multi-thread laser scanner and retro-reflective markers to distinguish the robot's leader and other objects. When the robot follows leader autonomously, the variant A* algorithm which having obstacle grids extended virtually (EA*) is used to plan the path. But if robots need to track and follow the leader's path as closely as possible, it will trust that the path which leader have traveled is safe enough and uses the incremental form of EA* algorithm (IEA*) to reproduce the trajectory. The simulation and experiment results illustrate the feasibility and effectiveness of the proposed algorithms.

We introduce MobiCeal, the first practical Plausibly Deniable Encryption (PDE) system for mobile devices that can defend against strong coercive multi-snapshot adversaries, who may examine the storage medium of a user's mobile device at different points of time and force the user to decrypt data. MobiCeal relies on "dummy write" to obfuscate the differences between multiple snapshots of storage medium due to existence of hidden data. By incorporating PDE in block layer, MobiCeal supports a broad deployment of any block-based file systems on mobile devices. More importantly, MobiCeal is secure against side channel attacks which pose a serious threat to existing PDE schemes. A proof of concept implementation of MobiCeal is provided on an LG Nexus 4 Android phone using Android 4.2.2. It is shown that the performance of MobiCeal is significantly better than prior PDE systems against multi-snapshot adversaries.

At present, mobile terminals are widely used in power system and easy to be the target or springboard to attack the power system. It is necessary to have security assessment of power mobile terminal system to enable early warning of potential risks. In the context, this paper builds the security assessment system against to power mobile terminals, with features from security assessment system of general mobile terminals and power application scenarios. Compared with the existing methods, this paper introduces machine learning to the Rank Correlation Analysis method, which relies on expert experience, and uses objective experimental data to optimize the weight parameters of the indicators. From experiments, this paper proves that weights self-learning method can be used to evaluate the security of power mobile terminal system and improve credibility of the result.

False data injection attacks (FDIA) on state estimators are a kind of imminent cyber-physical security issue. Fortunately, it has been proved that if a set of measurements is strategically selected and protected, no FDIA will remain undetectable. In this paper, the metric Return on Investment (ROI) is introduced to evaluate the overall returns of the alternative measurement protection schemes (MPS). By setting maximum total ROI as the optimization objective, the previously ignored cost-benefit issue is taken into account to derive a realistic MPS for power utilities. The optimization problem is transformed into the Steiner tree problem in graph theory, where a tree pruning based algorithm is used to reduce the computational complexity and find a quasi-optimal solution with acceptable approximations. The correctness and efficiency of the algorithm are verified by case studies.

Aiming at the problem that there is little research on firmware vulnerability mining and the traditional method of vulnerability mining based on fuzzing test is inefficient, this paper proposed a new method of mining vulnerabilities in industrial control system firmware. Based on taint analysis technology, this method can construct test cases specifically for the variables that may trigger vulnerabilities, thus reducing the number of invalid test cases and improving the test efficiency. Experiment result shows that this method can reduce about 23 % of test cases and can effectively improve test efficiency.

Federated learning (FL), as a recent advance of distributed machine learning, is capable of learning a model over the network without directly accessing the client's raw data. Nevertheless, the clients' sensitive information can still be exposed to adversaries via differential attacks on messages exchanged between the parameter server and clients. In this paper, we consider the widely used federating averaging (FedAvg) algorithm and propose to enhance the data privacy by the differential privacy (DP) technique, which obfuscates the exchanged messages by properly adding Gaussian noise. We analytically show that the proposed secure FedAvg algorithm maintains an O(l/T) convergence rate, where T is the total number of stochastic gradient descent (SGD) updates for local model parameters. Moreover, we demonstrate how various algorithm parameters can impact on the algorithm communication efficiency. Experiment results are presented to justify the obtained analytical results on the performance of the proposed algorithm in terms of testing accuracy.

With the rapid development of smart grids and the continuous deepening of informatization, while realizing remote telemetry and remote control of massive data-based grid operation, electricity information network security problems have become more serious and prominent. A method for electricity information network security situation prediction method based on improved deep belief network is proposed in this paper. Firstly, the affinity propagation clustering algorithm is used to determine the depth of the deep belief network and the number of hidden layer nodes based on sample parameters. Secondly, continuously adjust the scaling factor and crossover probability in the differential evolution algorithm according to the population similarity. Finally, a chaotic search method is used to perform a second search for the best individuals and similarity centers of each generation of the population. Simulation experiments show that the proposed algorithm not only enhances the generalization ability of electricity information network security situation prediction, but also has higher prediction accuracy.

Due to the harsh environment and energy limitation, maintaining efficient communication is crucial to the lifetime of Underwater Sensor Networks (UWSN). Named Data Networking (NDN), one of future network architectures, begins to be applied to UWSN. Although Underwater Named Data Networking (UNDN) performs well in data transmission, it still faces some security threats, such as the Denial-of-Service (DoS) attacks caused by Interest Flooding Attacks (IFAs). In this paper, we present a new type of DoS attacks, named as Synergetic Denial-of-Service (SDoS). Attackers synergize with each other, taking turns to reply to malicious interests as late as possible. SDoS attacks will damage the Pending Interest Table, Content Store, and Forwarding Information Base in routers with high concealment. Simulation results demonstrate that the SDoS attacks quadruple the increased network traffic compared with normal IFAs and the existing IFA detection algorithm in UNDN is completely invalid to SDoS attacks. In addition, we analyze the infection problem in UNDN and propose a defense method Trident based on carefully designed adaptive threshold, burst traffic detection, and attacker identification. Experiment results illustrate that Trident can effectively detect and resist both SDoS attacks and normal IFAs. Meanwhile, Trident can robustly undertake burst traffic and congestion.