Biblio

Cybersecurity research involves publishing papers about malicious exploits as much as publishing information on how to design tools to protect cyber-infrastructure. It is this information exchange between ethical hackers and security experts, which results in a well-balanced cyber-ecosystem. In the blooming domain of AI Safety Engineering, hundreds of papers have been published on different proposals geared at the creation of a safe machine, yet nothing, to our knowledge, has been published on how to design a malevolent machine. Availability of such information would be of great value particularly to computer scientists, mathematicians, and others who have an interest in AI safety, and who are attempting to avoid the spontaneous emergence or the deliberate creation of a dangerous AI, which can negatively affect human activities and in the worst case cause the complete obliteration of the human species. This paper provides some general guidelines for the creation of a Malevolent Artificial Intelligence (MAI).

Spectrum sensing (signal detection) under low signal to noise ratio is a fundamental problem in cognitive radio networks. In this paper, we have analyzed maximum eigenvalue detection (MED) and energy detection (ED) techniques known as semi-blind spectrum sensing techniques. Simulations are performed by using independent and identically distributed (iid) signals to verify the results. Maximum eigenvalue detection algorithm exploits correlation in received signal samples and hence, performs same as energy detection algorithm under high signal to noise ratio. Energy detection performs well under low signal to noise ratio for iid signals and its performance reaches maximum eigenvalue detection under high signal to noise ratio. Both algorithms don't need any prior knowledge of primary user signal for detection and hence can be used in various applications.

Operating system kernel is the de facto trusted computing base for most computer systems. To secure the OS kernel, many security mechanisms, e.g., kASLR and StackGuard, have been increasingly deployed to defend against attacks (e.g., code reuse attack). However, the effectiveness of these protections has been proven to be inadequate-there are many information leak vulnerabilities in the kernel to leak the randomized pointer or canary, thus bypassing kASLR and StackGuard. Other sensitive data in the kernel, such as cryptographic keys and file caches, can also be leaked. According to our study, most kernel information leaks are caused by uninitialized data reads. Unfortunately, existing techniques like memory safety enforcements and dynamic access tracking tools are not adequate or efficient enough to mitigate this threat. In this paper, we propose UniSan, a novel, compiler-based approach to eliminate all information leaks caused by uninitialized read in the OS kernel. UniSan achieves this goal using byte-level, flow-sensitive, context-sensitive, and field-sensitive initialization analysis and reachability analysis to check whether an allocation has been fully initialized when it leaves kernel space; if not, it automatically instruments the kernel to initialize this allocation. UniSan's analyses are conservative to avoid false negatives and are robust by preserving the semantics of the OS kernel. We have implemented UniSan as passes in LLVM and applied it to the latest Linux kernel (x86\_64) and Android kernel (AArch64). Our evaluation showed that UniSan can successfully prevent 43 known and many new uninitialized data leak vulnerabilities. Further, 19 new vulnerabilities in the latest kernels have been confirmed by Linux and Google. Our extensive performance evaluation with LMBench, ApacheBench, Android benchmarks, and the SPEC benchmarks also showed that UniSan imposes a negligible performance overhead.

UnlimitID is a method for enhancing the privacy of commodity OAuth and applications such as OpenID Connect, using anonymous attribute-based credentials based on algebraic Message Authentication Codes (aMACs). OAuth is one of the most widely used protocols on the Web, but it exposes each of the requests of a user for data by each relying party (RP) to the identity provider (IdP). Our approach allows for the creation of multiple persistent and unlinkable pseudo-identities and requires no change in the deployed code of relying parties, only in identity providers and the client.

Address clustering tries to construct the one-to-many mapping from entities to addresses in the Bitcoin system. Simple heuristics based on the micro-structure of transactions have proved very effective in practice. In this paper we describe the primary reasons behind this effectiveness: address reuse, avoidable merging, super-clusters with high centrality,, the incremental growth of address clusters. We quantify their impact during Bitcoin's first seven years of existence.

Maintaining and updating signature databases is a tedious task that normally requires a large amount of user effort. The problem becomes harder when features can be distorted by observation noise, which we call volatility. To address this issue, we propose algorithms and models to automatically generate signatures in the presence of noise, with a focus on stack fingerprinting, which is a research area that aims to discover the operating system (OS) of remote hosts using TCP/IP packets. Armed with this framework, we construct a database with 420 network stacks, label the signatures, develop a robust classifier for this database, and fingerprint 66M visible webservers on the Internet.

In a wireless system, a signal map shows the signal strength at different locations termed reference points (RPs). As access points (APs) and their transmission power may change over time, keeping an updated signal map is important for applications such as Wi-Fi optimization and indoor localization. Traditionally, the signal map is obtained by a full site survey, which is time-consuming and costly. We address in this paper how to efficiently update a signal map given sparse samples randomly crowdsourced in the space (e.g., by signal monitors, explicit human input, or implicit user participation). We propose Compressive Signal Reconstruction (CSR), a novel learning system employing Bayesian compressive sensing (BCS) for online signal map update. CSR does not rely on any path loss model or line of sight, and is generic enough to serve as a plug-in of any wireless system. Besides signal map update, CSR also computes the estimation error of signals in terms of confidence interval. CSR models the signal correlation with a kernel function. Using it, CSR constructs a sensing matrix based on the newly sampled signals. The sensing matrix is then used to compute the signal change at all the RPs with any BCS algorithm. We have conducted extensive experiments on CSR in our university campus. Our results show that CSR outperforms other state-of-the-art algorithms by a wide margin (reducing signal error by about 30% and sampling points by 20%).

The ineffectiveness of phishing warnings has been attributed to users' poor comprehension of the warning. However, the effectiveness of a phishing warning is typically evaluated at the time when users interact with a suspected phishing webpage, which we call the effect with phishing warning. Nevertheless, users' improved phishing detection when the warning is absent—or the effect of the warning—is the ultimate goal to prevent users from falling for phishing scams. We conducted an online study to evaluate the effect with and of several phishing warning variations, varying the point at which the warning was presented and whether procedural knowledge instruction was included in the warning interface. The current Chrome phishing warning was also included as a control. 360 Amazon Mechanical-Turk workers made submission; 500¬ word maximum for symposia) decisions about 10 login webpages (8 authentic, 2 fraudulent) with the aid of warning (first phase). After a short distracting task, the workers made the same decisions about 10 different login webpages (8 authentic, 2 fraudulent) without warning. In phase one, the compliance rates with two proposed warning interfaces (98% and 94%) were similar to those of the Chrome warning (98%), regardless of when the warning was presented. In phase two (without warning), performance was better for the condition in which warning with procedural knowledge instruction was presented before the phishing webpage in phase one, suggesting a better of effect than for the other conditions. With the procedural knowledge of how to determine a webpage’s legitimacy, users identified phishing webpages more accurately even without the warning being presented.

The ineffectiveness of phishing warnings has been attributed to users' poor comprehension of the warning. However, the effectiveness of a phishing warning is typically evaluated at the time when users interact with a suspected phishing webpage, which we call the effect with phishing warning. Nevertheless, users' improved phishing detection when the warning is absent—or the effect of the warning—is the ultimate goal to prevent users from falling for phishing scams. We conducted an online study to evaluate the effect with and of several phishing warning variations, varying the point at which the warning was presented and whether procedural knowledge instruction was included in the warning interface. The current Chrome phishing warning was also included as a control. 360 Amazon Mechanical-Turk workers made submission; 500¬ word maximum for symposia) decisions about 10 login webpages (8 authentic, 2 fraudulent) with the aid of warning (first phase). After a short distracting task, the workers made the same decisions about 10 different login webpages (8 authentic, 2 fraudulent) without warning. In phase one, the compliance rates with two proposed warning interfaces (98% and 94%) were similar to those of the Chrome warning (98%), regardless of when the warning was presented. In phase two (without warning), performance was better for the condition in which warning with procedural knowledge instruction was presented before the phishing webpage in phase one, suggesting a better of effect than for the other conditions. With the procedural knowledge of how to determine a webpage’s legitimacy, users identified phishing webpages more accurately even without the warning being presented.

Event discovery from single pictures is a challenging problem that has raised significant interest in the last decade. During this time, a number of interesting solutions have been proposed to tackle event discovery in still images. However, a large scale benchmarking image dataset for the evaluation and comparison of event discovery algorithms from single images is still lagging behind. To this aim, in this paper we provide a large-scale properly annotated and balanced dataset of 490,000 images, covering every aspect of 14 different types of social events, selected among the most shared ones in the social network. Such a large scale collection of event-related images is intended to become a powerful support tool for the research community in multimedia analysis by providing a common benchmark for training, testing, validation and comparison of existing and novel algorithms. In this paper, we provide a detailed description of how the dataset is collected, organized and how it can be beneficial for the researchers in the multimedia analysis domain. Moreover, a deep learning based approach is introduced into event discovery from single images as one of the possible applications of this dataset with a belief that deep learning can prove to be a breakthrough also in this research area. By providing this dataset, we hope to gather research community in the multimedia and signal processing domains to advance this application.

Maintaining the security and privacy hygiene of mobile apps is a critical challenge. Unfortunately, no program analysis algorithm can determine that an application is “secure” or “malware-free.” For example, if an application records audio during a phone call, it may be malware. However, the user may want to use such an application to record phone calls for archival and benign purposes. A key challenge for automated program analysis tools is determining whether or not that behavior is actually desired by the user (i.e., user expectation). This talk presents recent research progress in exploring user expectations in mobile app security.

User identity linkage across social platforms is an important problem of great research challenge and practical value. In real applications, the task often assumes an extra degree of difficulty by requiring linkage across multiple platforms. While pair-wise user linkage between two platforms, which has been the focus of most existing solutions, provides reasonably convincing linkage, the result depends by nature on the order of platform pairs in execution with no theoretical guarantee on its stability. In this paper, we explore a new concept of ``Latent User Space'' to more naturally model the relationship between the underlying real users and their observed projections onto the varied social platforms, such that the more similar the real users, the closer their profiles in the latent user space. We propose two effective algorithms, a batch model(ULink) and an online model(ULink-On), based on latent user space modelling. Two simple yet effective optimization methods are used for optimizing objective function: the first one based on the constrained concave-convex procedure(CCCP) and the second on accelerated proximal gradient. To our best knowledge, this is the first work to propose a unified framework to address the following two important aspects of the multi-platform user identity linkage problem –- (I) the platform multiplicity and (II) online data generation. We present experimental evaluations on real-world data sets for not only traditional pairwise-platform linkage but also multi-platform linkage. The results demonstrate the superiority of our proposed method over the state-of-the-art ones.

User modeling of individual users on the Social Web platforms such as Twitter plays a significant role in providing personalized recommendations and filtering interesting information from social streams. Recently, researchers proposed the use of concepts (e.g., DBpedia entities) for representing user interests instead of word-based approaches, since Knowledge Bases such as DBpedia provide cross-domain background knowledge about concepts, and thus can be used for extending user interest profiles. Even so, not all concepts can be covered by a Knowledge Base, especially in the case of microblogging platforms such as Twitter where new concepts/topics emerge everyday. In this short paper, instead of using concepts alone, we propose using synsets from WordNet and concepts from DBpedia for representing user interests. We evaluate our proposed user modeling strategies by comparing them with other bag-of-concepts approaches. The results show that using synsets and concepts together for representing user interests improves the quality of user modeling significantly in the context of link recommendations on Twitter.

This paper presents a framework for privacy-preserving video delivery system to fulfill users' privacy demands. The proposed framework leverages the inference channels in sensitive behavior prediction and object tracking in a video surveillance system for the sequence privacy protection. For such a goal, we need to capture different pieces of evidence which are used to infer the identity. The temporal, spatial and context features are extracted from the surveillance video as the observations to perceive the privacy demands and their correlations. Taking advantage of quantifying various evidence and utility, we let users subscribe videos with a viewer-dependent pattern. We implement a prototype system for off-line and on-line requirements in two typical monitoring scenarios to construct extensive experiments. The evaluation results show that our system can efficiently satisfy users' privacy demands while saving over 25% more video information compared to traditional video privacy protection schemes.

The huge popularity of online social networks and the potential financial gain have led to the creation and proliferation of zombie accounts, i.e., fake user accounts. For considerable amount of payment, zombie accounts can be directed by their managers to provide pre-arranged biased reactions to different social events or the quality of a commercial product. It is thus critical to detect and screen these accounts. Prior arts are either inaccurate or relying heavily on complex posting/tweeting behaviors in the classification process of normal/zombie accounts. In this work, we propose to use a bi-level penalized logistic classifier, an efficient high-dimensional data analysis technique, to detect zombie accounts based on their publicly available profile information and the statistics of their followers' registration locations. Our approach, termed (B)i-level (P)enalized (LO)gistic (C)lassifier (BPLOC), is data adaptive and can be extended to mount more accurate detections. Our experimental results are based on a small number of SINA WeiBo accounts and have demonstrated that BPLOC can classify zombie accounts accurately.

Interactive systems are developed according to requirements, which may be, for instance, documentation, prototypes, diagrams, etc. The informal nature of system requirements may be a source of problems: it may be the case that a system does not implement the requirements as expected, thus, a way to validate whether an implementation follows the requirements is needed. We propose a novel approach to validating a system using formal models of the system. In this approach, a set of traces generated from the execution of the real interactive system is searched over the state space of the formal model. The scalability of the approach is demonstrated by an application to an industrial system in the nuclear plant domain. The combination of trace analysis and formal methods provides feedback that can bring improvements to both the real interactive system and the formal model.

Authorship attribution is a stylometric technique that associates text to authors based on the type of writing styles. Researchers have looked for ways to analyze the context of these texts, in some cases with limited results. Most of the approaches view information at the syntactic and physical levels and tend to ignore information from the semantic levels. In this paper, we present a technique that incorporates the use of semantic frames as a method for authorship attribution. We hypothesize that it provides a deeper view into the semantic level of texts, which is an influencing factor in a writer's style. We use a variety of online resources in a pipeline fashion to extract information about frames within the text. The results show that our “bag of frames” approach can be used successfully for stylometry.
 

Authorship attribution is a stylometric technique that associates text to authors based on the type of writing styles. Researchers have looked for ways to analyze the context of these texts, in some cases with limited results. Most of the approaches view information at the syntactic and physical levels and tend to ignore information from the semantic levels. In this paper, we present a technique that incorporates the use of semantic frames as a method for authorship attribution. We hypothesize that it provides a deeper view into the semantic level of texts, which is an influencing factor in a writer's style. We use a variety of online resources in a pipeline fashion to extract information about frames within the text. The results show that our “bag of frames” approach can be used successfully for stylometry.

The cold start problem in recommender systems refers to the inability of making reliable recommendations if a critical mass of items has not yet been rated. To bypass this problem existing research focused on developing more reliable prediction models for situations in which only few items ratings exist. However, most of these approaches depend on adjusting the algorithm that determines a recommendation. We present a complimentary approach that does not require any adjustments to the recommendation algorithm. We draw on motivation theory and reward users for rating items. In particular, we instantiate different gamification patterns and examine their effect on the average user’s number of provided report ratings. Our results confirm the positive effect of instantiating gamification patterns on the number of received report ratings.

Cyber-attacks and breaches are often detected too late to avoid damage. While “classical” reactive cyber defenses usually work only if we have some prior knowledge about the attack methods and “allowable” patterns, properly constructed redundancy-based anomaly detectors can be more robust and often able to detect even zero day attacks. They are a step toward an oracle that uses knowable behavior of a healthy system to identify abnormalities. In the world of Internet of Things (IoT), security, and anomalous behavior of sensors and other IoT components, will be orders of magnitude more difficult unless we make those elements security aware from the start. In this article we examine the ability of redundancy-based a nomaly detectors to recognize some high-risk and difficult to detect attacks on web servers—a likely management interface for many IoT stand-alone elements. In real life, it has taken long, a number of years in some cases, to identify some of the vulnerabilities and related attacks. We discuss practical relevance of the approach in the context of providing high-assurance Webservices that may belong to autonomous IoT applications and devices

We present a process for detection of IP theft in VLSI devices that exploits the internal test scan chains. The IP owner learns implementation details in the suspect device to find evidence of the theft, while the top level function is public. The scan chains supply direct access to the internal registers in the device, thus making it possible to learn the logic functions of the internal combinational logic chunks. Our work introduces an innovative way of applying Boolean function analysis techniques for learning digital circuits with the goal of IP theft detection. By using Boolean function learning methods, the learner creates a partial dependency graph of the internal flip-flops. The graph is further partitioned using the SNN graph clustering method, and individual blocks of combinational logic are isolated. These blocks can be matched with known building blocks that compose the original function. This enables reconstruction of the function implementation to the level of pipeline structure. The IP owner can compare the resulting structure with his own implementation to confirm or refute that an IP violation has occurred. We demonstrate the power of the presented approach with a test case of an open source Bitcoin SHA-256 accelerator, containing more than 80,000 registers. With the presented method we discover the microarchitecture of the module, locate all the main components of the SHA-256 algorithm, and learn the module's flow control.