Biblio

Given a history of detected malware attacks, can we predict the number of malware infections in a country? Can we do this for different malware and countries? This is an important question which has numerous implications for cyber security, right from designing better anti-virus software, to designing and implementing targeted patches to more accurately measuring the economic impact of breaches. This problem is compounded by the fact that, as externals, we can only detect a fraction of actual malware infections. In this paper we address this problem using data from Symantec covering more than 1.4 million hosts and 50 malware spread across 2 years and multiple countries. We first carefully design domain-based features from both malware and machine-hosts perspectives. Secondly, inspired by epidemiological and information diffusion models, we design a novel temporal non-linear model for malware spread and detection. Finally we present ESM, an ensemble-based approach which combines both these methods to construct a more accurate algorithm. Using extensive experiments spanning multiple malware and countries, we show that ESM can effectively predict malware infection ratios over time (both the actual number and trend) upto 4 times better compared to several baselines on various metrics. Furthermore, ESM's performance is stable and robust even when the number of detected infections is low.

There has been growing interest in using convolutional neural networks (CNNs) in the fields of image forensics and steganalysis, and some promising results have been reported recently. These works mainly focus on the architectural design of CNNs, usually, a single CNN model is trained and then tested in experiments. It is known that, neural networks, including CNNs, are suitable to form ensembles. From this perspective, in this paper, we employ CNNs as base learners and test several different ensemble strategies. In our study, at first, a recently proposed CNN architecture is adopted to build a group of CNNs, each of them is trained on a random subsample of the training dataset. The output probabilities, or some intermediate feature representations, of each CNN, are then extracted from the original data and pooled together to form new features ready for the second level of classification. To make best use of the trained CNN models, we manage to partially recover the lost information due to spatial subsampling in the pooling layers when forming feature vectors. Performance of the ensemble methods are evaluated on BOSSbase by detecting S-UNIWARD at 0.4 bpp embedding rate. Results have indicated that both the recovery of the lost information, and learning from intermediate representation in CNNs instead of output probabilities, have led to performance improvement.

Heterogeneous face recognition aims to identify or verify person identity by matching facial images of different modalities. In practice, it is known that its performance is highly influenced by modality inconsistency, appearance occlusions, illumination variations and expressions. In this paper, a new method named as ensemble of sparse cross-modal metrics is proposed for tackling these challenging issues. In particular, a weak sparse cross-modal metric learning method is firstly developed to measure distances between samples of two modalities. It learns to adjust rank-one cross-modal metrics to satisfy two sets of triplet based cross-modal distance constraints in a compact form. Meanwhile, a group based feature selection is performed to enforce that features in the same position of two modalities are selected simultaneously. By neglecting features that attribute to "noise" in the face regions (eye glasses, expressions and so on), the performance of learned weak metrics can be markedly improved. Finally, an ensemble framework is incorporated to combine the results of differently learned sparse metrics into a strong one. Extensive experiments on various face datasets demonstrate the benefit of such feature selection especially when heavy occlusions exist. The proposed ensemble metric learning has been shown superiority over several state-of-the-art methods in heterogeneous face recognition.

WebRTC is one of the latest additions to the ever growing repository of Web browser technologies, which push the envelope of native Web application capabilities. WebRTC allows real-time peer-to-peer audio and video chat, that runs purely in the browser. Unlike existing video chat solutions, such as Skype, that operate in a closed identity ecosystem, WebRTC was designed to be highly flexible, especially in the domains of signaling and identity federation. This flexibility, however, opens avenues for identity fraud. In this paper, we explore the technical underpinnings of WebRTC's identity management architecture. Based on this analysis, we identify three novel attacks against endpoint authenticity. To answer the identified threats, we propose and discuss defensive strategies, including security improvements for the WebRTC specifications and mitigation techniques for the identity and service providers.

Cloud computing has been a great enabler for both the Internet of Things and Big Data. However, as with all new computing developments, development of the technology is usually much faster than consideration for, and development of, solutions for security and privacy. In a previous paper, we proposed that a unikernel solution could be used to improve security and privacy in a cloud scenario. In this paper, we outline how we might apply this approach to the Internet of Things, which can demonstrate an improvement over existing approaches.

We investigate the problem of constructing exponentially converging estimates of the state of a continuous-time system from state measurements transmitted via a limited-data-rate communication channel, so that only quantized and sampled measurements of continuous signals are available to the estimator. Following prior work on topological entropy of dynamical systems, we introduce a notion of estimation entropy which captures this data rate in terms of the number of system trajectories that approximate all other trajectories with desired accuracy. We also propose a novel alternative definition of estimation entropy which uses approximating functions that are not necessarily trajectories of the system. We show that the two entropy notions are actually equivalent. We establish an upper bound for the estimation entropy in terms of the sum of the system's Lipschitz constant and the desired convergence rate, multiplied by the system dimension. We propose an iterative procedure that uses quantized and sampled state measurements to generate state estimates that converge to the true state at the desired exponential rate. The average bit rate utilized by this procedure matches the derived upper bound on the estimation entropy. We also show that no other estimator (based on iterative quantized measurements) can perform the same estimation task with bit rates lower than the estimation entropy. Finally, we develop an application of the estimation procedure in determining, from the quantized state measurements, which of two competing models of a dynamical system is the true model. We show that under a mild assumption of exponential separation of the candidate models, detection is always possible in finite time. Our numerical experiments with randomly generated affine dynamical systems suggest that in practice the algorithm always works.

In this paper, we introduce Entropy/IP: a system that discovers Internet address structure based on analyses of a subset of IPv6 addresses known to be active, i.e., training data, gleaned by readily available passive and active means. The system is completely automated and employs a combination of information-theoretic and machine learning techniques to probabilistically model IPv6 addresses. We present results showing that our system is effective in exposing structural characteristics of portions of the active IPv6 Internet address space, populated by clients, services, and routers. In addition to visualizing the address structure for exploration, the system uses its models to generate candidate addresses for scanning. For each of 15 evaluated datasets, we train on 1K addresses and generate 1M candidates for scanning. We achieve some success in 14 datasets, finding up to 40% of the generated addresses to be active. In 11 of these datasets, we find active network identifiers (e.g., /64 prefixes or "subnets") not seen in training. Thus, we provide the first evidence that it is practical to discover subnets and hosts by scanning probabilistically selected areas of the IPv6 address space not known to contain active hosts a priori.

Contemporary vehicles are getting equipped with an increasing number of Electronic Control Units (ECUs) and wireless connectivities. Although these have enhanced vehicle safety and efficiency, they are accompanied with new vulnerabilities. In this paper, we unveil a new important vulnerability applicable to several in-vehicle networks including Control Area Network (CAN), the de facto standard in-vehicle network protocol. Specifically, we propose a new type of Denial-of-Service (DoS), called the bus-off attack, which exploits the error-handling scheme of in-vehicle networks to disconnect or shut down good/uncompromised ECUs. This is an important attack that must be thwarted, since the attack, once an ECU is compromised, is easy to be mounted on safety-critical ECUs while its prevention is very difficult. In addition to the discovery of this new vulnerability, we analyze its feasibility using actual in-vehicle network traffic, and demonstrate the attack on a CAN bus prototype as well as on two real vehicles. Based on our analysis and experimental results, we also propose and evaluate a mechanism to detect and prevent the bus-off attack.

Quality assurance for highly-configurable systems is challenging due to the exponentially growing configuration space. Interactions among multiple options can lead to surprising behaviors, bugs, and security vulnerabilities. Analyzing all configurations systematically might be possible though if most options do not interact or interactions follow specific patterns that can be exploited by analysis tools. To better understand interactions in practice, we analyze program traces to characterize and identify where interactions occur on control flow and data. To this end, we developed a dynamic analysis for Java based on variability-aware execution and monitor executions of multiple small to medium-sized programs. We find that the essential configuration complexity of these programs is indeed much lower than the combinatorial explosion of the configuration space indicates. However, we also discover that the interaction characteristics that allow scalable and complete analyses are more nuanced than what is exploited by existing state-of-the-art quality assurance strategies.

To help establish a more scientific basis for security science, which will enable the development of fundamental theories and move the field from being primarily reactive to primarily proactive, it is important for research results to be reported in a scientifically rigorous manner. Such reporting will allow for the standard pillars of science, namely replication, meta-analysis, and theory building. In this paper we aim to establish a baseline of the state of scientific work in security through the analysis of indicators of scientific research as reported in the papers from the 2015 IEEE Symposium on Security and Privacy. To conduct this analysis, we developed a series of rubrics to determine the completeness of the papers relative to the type of evaluation used (e.g. case study, experiment, proof). Our findings showed that while papers are generally easy to read, they often do not explicitly document some key information like the research objectives, the process for choosing the cases to include in the studies, and the threats to validity. We hope that this initial analysis will serve as a baseline against which we can measure the advancement of the science of security.

To help establish a more scientific basis for security science, which will enable the development of fundamental theories and move the field from being primarily reactive to primarily proactive, it is important for research results to be reported in a scientifically rigorous manner. Such reporting will allow for the standard pillars of science, namely replication, meta-analysis, and theory building. In this paper we aim to establish a baseline of the state of scientific work in security through the analysis of indicators of scientific research as reported in the papers from the 2015 IEEE Symposium on Security and Privacy. To conduct this analysis, we developed a series of rubrics to determine the completeness of the papers relative to the type of evaluation used (e.g. case study, experiment, proof). Our findings showed that while papers are generally easy to read, they often do not explicitly document some key information like the research objectives, the process for choosing the cases to include in the studies, and the threats to validity. We hope that this initial analysis will serve as a baseline against which we can measure the advancement of the science of security.

Software security is an important concern in the world moving towards Information Technology. Detecting software vulnerabilities is a difficult and resource consuming task. Therefore, automatic vulnerability prediction would help development teams to predict vulnerability-prone components and prioritize security inspection efforts. Software source code metrics and data mining techniques have been recently used to predict vulnerability-prone components. Some of previous studies used a set of unit complexity and coupling metrics to predict vulnerabilities. In this study, first, we compare the predictability power of these two groups of metrics in cross-project vulnerability prediction. In cross-project vulnerability prediction we create the prediction model based on datasets of completely different projects and try to detect vulnerabilities in another project. The experimental results show that unit complexity metrics are stronger vulnerability predictors than coupling metrics. Then, we propose a new set of coupling metrics which are called Included Vulnerable Header (IVH) metrics. These new coupling metrics, which consider interaction of application modules with outside of the application, predict vulnerabilities highly better than regular coupling metrics. Furthermore, adding IVH metrics to the set of complexity metrics improves Recall of the best predictor from 60.9% to 87.4% and shows the best set of metrics for cross-project vulnerability prediction.

Virtual reality allows users to experience unusual immersive environments. There are still several aspect of design for virtual reality that need more investigation, such as transitioning between environments. Multiple studies have shown that physical movement in a virtual environment supports immersion and presence. Our setup will allow the comparative study of the coupling of virtual camera movements with simultaneous physical movements of the user in terms of user preference and comfort. This work-in-progress uses a within-subject experimental design for evaluating interaction prototypes based on the Oculus Rift DK2 where participants will be tasked with transitioning between different environments; once using physical motion to merely trigger the transition and once with the virtual camera movement being coupled to the physical motion. Qualitative and quantitative data will be collected utilizing questionnaires and in-game metrics. Pretests of a similar setup were used to establish minimal levels of comfort.

Fabrication process introduces some inherent variability to the attributes of transistors (in particular length, widths, oxide thickness). As a result, every chip is physically unique. Physical uniqueness of microelectronics components can be used for multiple security applications. Physically Unclonable Functions (PUFs) are built to extract the physical uniqueness of microelectronics components and make it usable for secure applications. However, the microelectronics components used by PUFs designs suffer from external, environmental variations that impact the PUF behavior. Variations of temperature gradients during manufacturing can bias the PUF responses. Variations of temperature or thermal noise during PUF operation change the behavior of the circuit, and can introduce errors in PUF responses. Detailed knowledge of the behavior of PUFs operating over various environmental factors is needed to reliably extract and demonstrate uniqueness of the chips. In this work, we present a detailed and exhaustive analysis of the behavior of two PUF designs, a ring oscillator PUF and a timing path violation PUF. We have implemented both PUFs using FPGA fabricated by Xilinx, and analyzed their behavior while varying temperature and supply voltage. Our experiments quantify the robustness of each design, demonstrate their sensitivity to temperature and show the impact which supply voltage has on the uniqueness of the analyzed PUFs.

Software systems are increasingly called upon to autonomously manage their goals in changing contexts and environments, and under evolving requirements. In some circumstances, autonomous systems cannot be fully-automated but instead cooperate with human operators to maintain and adapt themselves. Furthermore, there are times when a choice should be made between doing a manual or automated repair. Involving operators in self-adaptation should itself be adaptive, and consider aspects such as the training, attention, and ability of operators. Not only do these aspects change from person to person, but they may change with the same person. These aspects make the choice of whether to involve humans non-obvious. Self-adaptive systems should trade-off whether to involve operators, taking these aspects into consideration along with other business qualities it is attempting to achieve. In this chapter, we identify the various roles that operators can perform in cooperating with self-adapting systems. We focus on humans as effectors-doing tasks which are difficult or infeasible to automate. We describe how we modified our self-adaptive framework, Rainbow, to involve operators in this way, which involved choosing suitable human models and integrating them into the existing utility trade-off decision models of Rainbow. We use probabilistic modeling and quantitative verification to analyze the trade-offs of involving humans in adaptation, and complement our study with experiments to show how different business preferences and modalities of human involvement may result in different outcomes.

A mobile ad hoc network (MANET) is an infrastructure-less network of various mobile devices and generally known for its self configuring behavior. MANET can communicate over relatively bandwidth constrained wireless links. Due to limited bandwidth battery power and dynamic network, topology routing in MANET is a challenging issue. Collaborative attacks are particularly serious issues in MANET. Attacks are liable to occur if routing algorithms fail to detect prone threats and to find as well as remove malicious nodes. Our objective is to examine and improve the performance of network diminished by variety of attacks. The performance of MANET network is examined under Black hole, Wormhole and Sybil attacks using Performance matrices and then major issues which are related to these attacks are addressed.

This paper evaluates a new video surveillance platform presented in a previous study, through an abandoned object detection task. The proposed platform has a function of automated detection and alerting, which is still a big challenge for a machine algorithm due to its recall-precision tradeoff problem. To achieve both high recall and high precision simultaneously, a hybrid approach using crowdsourcing after image analysis is proposed. This approach, however, is still not clear about what extent it can improve detection accuracy and raise quicker alerts. In this paper, the experiment is conducted for abandoned object detection, as one of the most common surveillance tasks. The results show that detection accuracy was improved from 50% (without crowdsourcing) to stable 95-100% (with crowdsourcing) by majority vote of 7 crowdworkers for each task. In contrast, alert time issue still remains open to further discussion since at least 7+ minutes are required to get the best performance.

In the age of the IoT (Internet of Things), a random number generator plays an important role of generating encryption keys and authenticating a piece of an embedded equipment. The random numbers are required to be uniformly distributed statistically and unpredictable. To satisfy the requirements, a physical true random number generator (TR-NG) is used. In this paper, we implement a TRNG using an SR latch on 40 nm CMOS ASIC. This TRNG generates the random number by exclusive ORing (XORing) the outputs of 256 SR latches. We evaluate the random number generated using statistical tests in accordance with BSI AIS 20/31 and using an IID (Independent and Identically Distributed) test, and the entropy estimation in accordance with NIST SP800-90B changing the supply voltage and environmental temperature within its rated values. As a result, the TRNG passed all the tests except in a few cases. From this experiment, we found that the TRNG has a robustness against environmental change. The power consumption is 18.8 micro Watt at 2.5 MHz. This TRNG is suitable for embedded systems to improve security in IoT systems.