Biblio

Discovering the potential vulnerabilities in software plays a crucial role in ensuring the security of computer system. This paper proposes a method that can assist security auditors with the analysis of source code. When security auditors identify new vulnerabilities, our method can be adopted to make a list of recommendations that may have the same vulnerabilities for the security auditors. Our method relies on graph representation to automatically extract the mode of PDG(program dependence graph, a structure composed of control dependence and data dependence). Besides, it can be applied to the vulnerability extrapolation scenario, thus reducing the amount of audit code. We worked on an open-source vulnerability test set called Juliet. According to the evaluation results, the clustering effect produced is satisfactory, so that the feature vectors extracted by the Graph2Vec model are applied to labeling and supervised learning indicators are adopted to assess the model for its ability to extract features. On a total of 12,000 small data sets, the training score of the model can reach up to 99.2%, and the test score can reach a maximum of 85.2%. Finally, the recommendation effect of our work is verified as satisfactory.

With the progressive development of web applications and the urgent requirement of web security, vulnerability scanner has been particularly emphasized, which is regarded as a fundamental component for web security assurance. Various scanners are developed with the intention of that discovering the possible vulnerabilities in advance to avoid malicious attacks. However, most of them only focus on the vulnerability detection with single target, which fail in satisfying the efficiency demand of users. In this paper, an effective web vulnerability scanner that integrates the information collection with the vulnerability detection is proposed to verify whether the target web application is vulnerable or not. The experimental results show that, by guiding the detection process with the useful collected information, our tool achieves great web vulnerability detection capability with a large scanning scope.

The drawbacks of traditional methods of cybernetic vulnerability detection relate to the required time to identify new threats, to register them in the Common Vulnerabilities and Exposures (CVE) records, and to score them with the Common Vulnerabilities Scoring System (CVSS). These problems can be mitigated by early vulnerability detection systems relying on social media and open-source data. This paper presents a model that aims to identify emerging cybernetic vulnerabilities in cybersecurity news articles, as part of a system for automatic detection of early cybernetic threats using Open Source Intelligence (OSINT). Three machine learning models were trained on a novel dataset of 1000 labeled news articles to create a strong baseline for classifying cybersecurity articles as relevant (i.e., introducing new security threats), or irrelevant: Support Vector Machines, a Multinomial Naïve Bayes classifier, and a finetuned BERT model. The BERT model obtained the best performance with a mean accuracy of 88.45% on the test dataset. Our experiments support the conclusion that Natural Language Processing (NLP) models are an appropriate choice for early vulnerability detection systems in order to extract relevant information from cybersecurity news articles.

Ahstract-Vulnerability detection is an important topic of software engineering. To improve the effectiveness and efficiency of vulnerability detection, many traditional machine learning-based and deep learning-based vulnerability detection methods have been proposed. However, the impact of different factors on vulnerability detection is unknown. For example, classification models and vectorization methods can directly affect the detection results and code replacement can affect the features of vulnerability detection. We conduct a comparative study to evaluate the impact of different classification algorithms, vectorization methods and user-defined variables and functions name replacement. In this paper, we collected three different vulnerability code datasets. These datasets correspond to different types of vulnerabilities and have different proportions of source code. Besides, we extract and analyze the features of vulnerability code datasets to explain some experimental results. Our findings from the experimental results can be summarized as follows: (i) the performance of using deep learning is better than using traditional machine learning and BLSTM can achieve the best performance. (ii) CountVectorizer can improve the performance of traditional machine learning. (iii) Different vulnerability types and different code sources will generate different features. We use the Random Forest algorithm to generate the features of vulnerability code datasets. These generated features include system-related functions, syntax keywords, and user-defined names. (iv) Datasets without user-defined variables and functions name replacement will achieve better vulnerability detection results.

In this paper, we assess the vulnerability of network anomaly detection systems that use machine learning methods. Although the performance of these network anomaly detection systems is high in comparison to that of existing methods without machine learning methods, the use of machine learning methods for detecting vulnerabilities is a growing concern among researchers of image processing. If the vulnerabilities of machine learning used in the network anomaly detection method are exploited by attackers, large security threats are likely to emerge in the near future. Therefore, in this paper we clarify how vulnerability detection of machine learning network anomaly detection methods affects their performance.

To find software vulnerabilities using software vulnerability detection technology is an important way to ensure the system security. Existing software vulnerability detection methods have some limitations as they can only play a certain role in some specific situations. To accurately analyze and evaluate the existing vulnerability detection methods, an integrated testing and evaluation system (iTES) is designed and implemented in this paper. The main functions of the iTES are:(1) Vulnerability cases with source codes covering common vulnerability types are collected automatically to form a vulnerability cases library; (2) Fourteen methods including static and dynamic vulnerability detection are evaluated in iTES, involving the Windows and Linux platforms; (3) Furthermore, a set of evaluation metrics is designed, including accuracy, false positive rate, utilization efficiency, time cost and resource cost. The final evaluation and test results of iTES have a good guiding significance for the selection of appropriate software vulnerability detection methods or tools according to the actual situation in practice.

This paper proposes a novel zero-forcing (ZF) beamforming strategy that can simultaneously cope with active and passive eavesdroppers (EDs) in visible light communication systems. A related optimization problem is formulated to maximize the signal-to-noise ratio (SNR) of the legitimate user (UE) while suppressing the SNR of active ED to zero and constraining the average SNR of passive EDs. The proposed beamforming directs the transmission along a particular eigenmode related to the null space of the active ED channel and the intensity of the passive ED point process. An inverse free preconditioned Krylov subspace projection method is used to find the eigenmode. The numerical results show that the proposed ZF beamforming scheme yields better performance relative to a traditional ZF beamforming scheme in the sense of increasing the SNR of the UE and reducing the secrecy outage probability.

Visible light communication (VLC) is a promising technique in the fifth and beyond wireless communication networks. In this paper, a secure multiple-input single-output VLC network is studied, where simultaneous lightwave information and power transfer (SLIPT) is exploited to support energy-limited devices taking into account a practical non-linear energy harvesting model. Specifically, the optimal beamforming design problems for minimizing transmit power and maximizing the minimum secrecy rate are studied under the imperfect channel state information (CSI). S-Procedure and a bisection search is applied to tackle challenging non-convex problems and to obtain efficient resource allocation algorithm. It is proved that optimal beamforming schemes can be obtained. It is found that there is a non-trivial trade-off between the average harvested power and the minimum secrecy rate. Moreover, we show that the quality of CSI has a significant impact on achievable performance.

Optical wireless communications (OWC) and its potential to solve physical layer security (PLS) issues are becoming important research areas in 6G communications systems. In this paper, an overview of PLS in visible light communications (VLC), is presented. Then, two new PLS techniques based on generalized space shift keying (GSSK) modulation with spatial constellation design (SCD) and non-orthogonal multiple access (NOMA) cooperative relaying are introduced. In the first technique, the PLS of the system is enhanced by the appropriate selection of a precoding matrix for randomly activated light emitting diodes (LEDs). With the aid of a legitimate user's (Bob's) channel state information (CSI) at the transmitter (CSIT), the bit error ratio (BER) of Bob is minimized while the BER performance of the potential eavesdroppers (Eves) is significantly degraded. In the second technique, superposition coding with uniform signaling is used at the transmitter and relays. The design of secure beamforming vectors at the relay nodes along with NOMA techniques is used to enhance PLS in a VLC system. Insights gained from the improved security levels of the proposed techniques are used to discuss how PLS can be further improved in future generation communication systems by using VLC.

In security protocol design, Visible Light Communication (VLC) has often been abstracted as an ideal channel that is resilient to eavesdropping, manipulation, and jamming. Camera Communication (CamCom), a subcategory of VLC, further strengthens the level of security by providing a visually verifiable association between the transmitter and the extracted information. However, the ideal security guarantees of visible light channels may not hold in practice due to limitations and tradeoffs introduced by hardware, software, configuration, environment, etc. This paper presents our experience and lessons learned from implementing CamCom for security protocols. We highlight CamCom's security-enhancing properties and security applications that it enables. Backed by real implementation and experiments, we also systematize the practical considerations of CamCom-based security protocols.

Remote Attestation (RA) is a security service that detects malware presence on remote IoT devices by verifying their software integrity by a trusted party (verifier). There are three main types of RA: software (SW)-, hardware (HW)-, and hybrid (SW/HW)-based. Hybrid techniques obtain secure RA with minimal hardware requirements imposed on the architectures of existing microcontrollers units (MCUs). In recent years, considerable attention has been devoted to hybrid techniques since prior software-based ones lack concrete security guarantees in a remote setting, while hardware-based approaches are too costly for low-end MCUs. However, one key problem is that many already deployed IoT devices neither satisfy minimal hardware requirements nor support hardware modifications, needed for hybrid RA. This paper bridges the gap between software-based and hybrid RA by proposing a novel RA scheme based on software virtualization. In particular, it proposes a new scheme, called SIMPLE, which meets the minimal hardware requirements needed for secure RA via reliable software. SIMPLE depends on a formally-verified software-based memory isolation technique, called Security MicroVisor (Sμ V). Its reliability is achieved by extending the formally-verified safety and correctness properties to cover the entire software architecture of SIMPLE. Furthermore, SIMPLE is used to construct SIMPLE+, an efficient swarm attestation scheme for static and dynamic heterogeneous IoT networks. We implement and evaluate SIMPLE and SIMPLE+ on Atmel AVR architecture, a common MCU platform.

Machine-to-Machine (M2M) communication is a essential subset of the Internet of Things (IoT). Secure access to communication network systems by M2M devices requires the support of a secure and efficient anonymous authentication protocol. The Direct Anonymous Attestation (DAA) scheme in Trustworthy Computing is a verified security protocol. However, the existing defense system uses a static architecture. The “mimic defense” strategy is characterized by active defense, which is not effective against continuous detection and attack by the attacker. Therefore, in this paper, we propose a Mimic-DAA scheme that incorporates mimic defense to establish an active defense scheme. Multiple heterogeneous and redundant actuators are used to form a DAA verifier and optimization is scheduled so that the behavior of the DAA verifier unpredictable by analysis. The Mimic-DAA proposed in this paper is capable of forming a security mechanism for active defense. The Mimic-DAA scheme effectively safeguard the unpredictability, anonymity, security and system-wide security of M2M communication networks. In comparison with existing DAA schemes, the scheme proposed in this paper improves the safety while maintaining the computational complexity.

False data injection attack (FDIA) destroys the automatic generation control (AGC) system and leads to unstable operation of the power system. Fast and accurate detection can help prevent and disrupt malicious attacks. This paper proposes an improved detection method, which is combined with fuzzy theory and support vector machine (SVM) to identify various types of attacks. The impacts of different types of FDIAs on the AGC system are analyzed, and the reliability of the method is proved by a large number of experimental data. This experiment is simulated on a single-area LFC system and the effects of adding a wind storage system were compared in a dynamic model. Simulation studies also show a higher accuracy of fuzzy support vector machine (FSVM) than traditional SVM and fuzzy pattern trees (FPTs).

Vehicle-to-vehicle (V2V) communication systems are currently being prepared for real-world deployment, but they face strong opposition over privacy concerns. Position beacon messages are the main culprit, being broadcast in cleartext and pseudonymously signed up to 10 times per second. So far, no practical solutions have been proposed to encrypt or anonymously authenticate V2V messages. We propose two cryptographic innovations that enhance the privacy of V2V communication. As a core contribution, we introduce zone-encryption schemes, where vehicles generate and authentically distribute encryption keys associated to static geographic zones close to their location. Zone encryption provides security against eavesdropping, and, combined with a suitable anonymous authentication scheme, ensures that messages can only be sent by genuine vehicles, while adding only 224 Bytes of cryptographic overhead to each message. Our second contribution is an authentication mechanism fine-tuned to the needs of V2V which allows vehicles to authentically distribute keys, and is called dynamic group signatures with attributes. Our instantiation features unlimited locally generated pseudonyms, negligible credential download-and-storage costs, identity recovery by a trusted authority, and compact signatures of 216 Bytes at a 128-bit security level.

Anonymous communication networks (ACNs) are intended to protect the metadata during communication. As classic ACNs, onion mix-nets are famous for strong anonymity, in which the source defines a static path and wraps the message multi-times with the public keys of nodes on the path, through which the message is relayed to the destination. However, onion mix-nets lacks in resilience when the static on-path mixes fail. Mix failure easily results in message loss, communication failure, and even specific attacks. Therefore, it is desirable to achieve resilient routing in onion mix-nets, providing persistent routing capability even though node failure. The state-of-theart solutions mainly adopt mix groups and thus need to share secret keys among all the group members which may cause single point of failure. To address this problem, in this work we propose a hybrid routing approach, which embeds the onion mix-net with hop-by-hop routing to increase routing resilience. Furthermore, we propose the threshold hybrid routing to achieve better key management and avoid single point of failure. As for experimental evaluations, we conduct quantitative analysis of the resilience and realize a local T-hybrid routing prototype to test performance. The experimental results show that our proposed routing strategy increases routing resilience effectively, at the expense of acceptable latency.

In vehicular cyber-physical systems, the mounted cameras on the vehicles, together with the fixed roadside cameras, can produce pictorial data for multiple purposes. In this process, ensuring the security and privacy of vehicles while guaranteeing efficient data transmission among vehicles is critical. This motivates us to propose a trusted anonymous authentication scheme for vehicular cyber-physical systems and Internet-of-Things. Our scheme is designed based on a three-tier architecture which contains cloud layer, fog layer, and user layer. It utilizes bilinear-free certificateless signcryption to realize a secure and trusted anonymous authentication efficiently. We verify its effectiveness through theoretical analyses in terms of correctness, security, and efficiency. Furthermore, our simulation results demonstrate that the communication overhead, the computation overhead, and the packet loss rate of the proposed scheme are significantly better than those of the state-of-the-art techniques. Particularly, the proposed scheme can speed up the computation process at least 10× compared to all the state-of-the-art approaches.

As a current disruptive and transformative technology, artificial intelligence is constantly infiltrating all aspects of production and life. However, with the in-depth development and application of artificial intelligence, the security challenges it faces have become more and more prominent. In the real world, attacks against intelligent systems such as the Internet of Things, smart homes, and driverless cars are constantly appearing, and incidents of artificial intelligence being used in cyber-attacks and cybercrimes frequently occur. This article aims to discuss artificial intelligence security issues and propose some countermeasures.

Our research team have devoted to extract internal malicious behavior by monitoring the network traffic for many years. We applied the deep learning approach to recognize the malicious patterns within network, but this methodology may lead to more works to examine the results from AI models production. Hence, this paper addressed the scenario to consider the burden of AI, and proposed an idea for long-term reliable detection in the future work.

Using artificial intelligence technology to help network security has become a major trend. At present, major countries in the world have successively invested R & D force in the attack and defense of automatic network based on artificial intelligence. The U.S. Navy, the U.S. air force, and the DOD strategic capabilities office have invested heavily in the development of artificial intelligence network defense systems. DARPA launched the network security challenge (CGC) to promote the development of automatic attack system based on artificial intelligence. In the 2016 Defcon final, mayhem (the champion of CGC in 2014), an automatic attack team, participated in the competition with 14 human teams and once defeated two human teams, indicating that the automatic attack method generated by artificial intelligence system can scan system defects and find loopholes faster and more effectively than human beings. Japan's defense ministry also announced recently that in order to strengthen the ability to respond to network attacks, it will introduce artificial intelligence technology into the information communication network defense system of Japan's self defense force. It can be predicted that the deepening application of artificial intelligence in the field of network attack and defense may bring about revolutionary changes and increase the imbalance of the strategic strength of cyberspace in various countries. Therefore, it is necessary to systematically investigate the current situation of network attack and defense based on artificial intelligence at home and abroad, comprehensively analyze the development trend of relevant technologies at home and abroad, deeply analyze the development outline and specification of artificial intelligence attack and defense around the world, and refine the application status and future prospects of artificial intelligence attack and defense, so as to promote the development of artificial intelligence attack and Defense Technology in China and protect the core interests of cyberspace, of great significance

Neural networks with deep architectures have been used to construct state-of-the-art classifiers that can match human level accuracy in areas such as image classification. However, many of these classifiers can be fooled by examples slightly modified from their original forms. In this work, we propose a novel approach for generating adversarial examples that makes use of only attribution information of the features and perturbs only features that are highly influential to the output of the classifier. We call this approach Attribution Based Adversarial Generation (ABAG). To demonstrate the effectiveness of this approach, three somewhat arbitrary algorithms are proposed and examined. In the first algorithm all non-zero attributions are utilized and associated features perturbed; in the second algorithm only the top-n most positive and top-n most negative attributions are used and corresponding features perturbed; and in the third algorithm the level of perturbation is increased in an iterative manner until an adversarial example is discovered. All of the three algorithms are implemented and experiments are performed on the well-known MNIST dataset. Experiment results show that adversarial examples can be generated very efficiently, and thus prove the validity and efficacy of ABAG - utilizing attributions for the generation of adversarial examples. Furthermore, as shown by examples, ABAG can be adapted to provides a systematic searching approach to generate adversarial examples by perturbing a minimum amount of features.

This study focuses on identifying the factors contributing to a sense of personal responsibility that could improve understanding of insecure cybersecurity behavior and guide research toward more effective messaging targeting non-adopting populations. Towards that, we ran a 2(account type) x2(usage scenario) x2(message type) between-group study with 237 United States adult participants on Amazon MTurk, and investigated how the non-adopting population allocates blame, and under what circumstances they blame the end user among the parties who hold responsibility: the software companies holding data, the attackers exposing data, and others. We find users primarily hold service providers accountable for breaches but they feel the same companies should not enforce stronger security policies on users. Results indicate that people do hold end users accountable for their behavior in the event of a breach, especially when the users' behavior affects others. Implications of our findings in risk communication is discussed in the paper.

Exploring the intrinsic interconnections between the knowledge encoded in PRe-trained Deep Neural Networks (PR-DNNs) of heterogeneous tasks sheds light on their mutual transferability, and consequently enables knowledge transfer from one task to another so as to reduce the training effort of the latter. In this paper, we propose the DEeP Attribution gRAph (DEPARA) to investigate the transferability of knowledge learned from PR-DNNs. In DEPARA, nodes correspond to the inputs and are represented by their vectorized attribution maps with regards to the outputs of the PR-DNN. Edges denote the relatedness between inputs and are measured by the similarity of their features extracted from the PR-DNN. The knowledge transferability of two PR-DNNs is measured by the similarity of their corresponding DEPARAs. We apply DEPARA to two important yet under-studied problems in transfer learning: pre-trained model selection and layer selection. Extensive experiments are conducted to demonstrate the effectiveness and superiority of the proposed method in solving both these problems. Code, data and models reproducing the results in this paper are available at https://github.com/zju-vipa/DEPARA.

Advertising channels have evolved from conventional print media, billboards and radio-advertising to online digital advertising (ad), where the users are exposed to a sequence of ad campaigns via social networks, display ads, search etc. While advertisers revisit the design of ad campaigns to concurrently serve the requirements emerging out of new ad channels, it is also critical for advertisers to estimate the contribution from touch-points (view, clicks, converts) on different channels, based on the sequence of customer actions. This process of contribution measurement is often referred to as multi-touch attribution (MTA). In this work, we propose CAMTA, a novel deep recurrent neural network architecture which is a causal attribution mechanism for user-personalised MTA in the context of observational data. CAMTA minimizes the selection bias in channel assignment across time-steps and touchpoints. Furthermore, it utilizes the users' pre-conversion actions in a principled way in order to predict per-channel attribution. To quantitatively benchmark the proposed MTA model, we employ the real-world Criteo dataset and demonstrate the superior performance of CAMTA with respect to prediction accuracy as compared to several baselines. In addition, we provide results for budget allocation and user-behaviour modeling on the predicted channel attribution.

The increasing vulnerabilities in Android kernel make it an attractive target to the attackers. Most kernel-targeted attacks are initiated through system calls. For security purpose, Google has introduced a Linux kernel security mechanism named “seccomp” since Android O to constrain the system calls accessible to the Android apps. Unfortunately, existing Android seccomp mechanism provides a fairly coarse-grained restriction by enforcing a unified seccomp policy containing more than 250 system calls for Android apps, which greatly reduces the effectiveness of seccomp. Also, it lacks an approach to profile the unnecessary system calls for a given Android app. In this paper we present a two-level control scheme named SASAK, which can shrink the attack surface of Android kernel by strictly constraining the system calls available to the Android apps with seccomp mechanism. First, instead of leveraging a unified seccomp policy for all Android apps, SASAK introduces an architecture- dedicated system call constraining by enforcing two separate and refined seccomp policies for the 32-bit Android apps and 64-bit Android apps, respectively. Second, we provide a tool to profile the necessary system calls for a given Android app and enforce an app-dedicated seccomp policy to further reduce the allowed system calls for the apps selected by the users. The app-dedicated control could dynamically change the seccomp policy for an app according to its actual needs. We implement a prototype of SASAK and the experiment results show that the architecture-dedicated constraining reduces 39.6% system calls for the 64-bit apps and 42.5% system calls for the 32-bit apps. 33% of the removed system calls for the 64-bit apps are vulnerable, and the number for the 32-bit apps is 18.8%. The app-dedicated restriction reduces about 66.9% and 62.5% system calls on average for the 64-bit apps and 32-bit apps, respectively. In addition, SASAK introduces negligible performance overhead.

The space environment is currently undergoing a substantial change and many new entrants to the market are deploying devices, satellites and systems in space; this evolution has been termed as NewSpace. The change is complicated by technological developments such as deploying machine learning based autonomous space systems and the Internet of Space Things (IoST). In the IoST, space systems will rely on satellite-to-x communication and interactions with wider aspects of the ground segment to a greater degree than existing systems. Such developments will inevitably lead to a change in the cyber security threat landscape of space systems. Inevitably, there will be a greater number of attack vectors for adversaries to exploit, and previously infeasible threats can be realised, and thus require mitigation. In this paper, we present a reference architecture (RA) that can be used to abstractly model in situ applications of this new space landscape. The RA specifies high-level system components and their interactions. By instantiating the RA for two scenarios we demonstrate how to analyse the attack surface using attack trees.