With the steady growth of Android malware, we suspect that, during the malware development phase, some Android malware writers use the popular public scanning services (e.g., VirusTotal) for testing the evasion capability of their malware samples, which we name Android malware development cases (AMDs). In this work, we design an AMD hunter in the context of VirusTotal to hunt for AMDs and reveal new threats for Android. First, the AMD hunter sifts through millions of file submissions on VirusTotal efficiently and alert more suspicious submission traces. Second, it performs package level analysis, static code and dynamic analyses on the APKs of the suspicious submissions to validate the AMDs. The implemented hunter has been used in a leading security company for 4 months, which processed 153 million of submissions on VirusTotal, and identified 1,623 AMDs with 13,855 samples from 83 countries. We also performed case studies on 890 malware samples selected from the identified AMDs, which revealed lots of new threats, including the development cases of fake system/banking phishing app, new rooting exploits, new JavaScript based threats, new evasions and AV probing malware. We wrote industry research articles about some AMDs and notified other security vendors to help patch their false negatives. Besides raising the awareness of the existence of AMDs, more importantly, our research provides the first systematic and efficient way to study the malware development phenomenon on VirusTotal. We will share all the samples of the identified AMDs with the research community.

Distributed collaborative learning (DCL) paradigms enable building joint machine learning models from distrusted multi-party participants. Data confidentiality is guaranteed by retaining private training data on each participant's local infrastructure. However, this approach makes today's DCL design fundamentally vulnerable to data poisoning and backdoor attacks. It limits DCL's model accountability, which is key to backtracking problematic training data instances and their responsible contributors. In this paper, we introduce CALTRAIN, a centralized collaborative learning system that simultaneously achieves data confidentiality and model accountability. CALTRAIN enforces isolated computation via secure enclaves on centrally aggregated training data to guarantee data confidentiality. To support building accountable learning models, we securely maintain the links between training instances and their contributors. Our evaluation shows that the models generated by CALTRAIN can achieve the same prediction accuracy when compared to the models trained in non-protected environments. We also demonstrate that when malicious training participants tend to implant backdoors during model training, CALTRAIN can accurately and precisely discover the poisoned or mislabeled training data that lead to the runtime mispredictions.

Deep neural networks (DNNs) have been widely applied in various applications, including autonomous driving and surveillance systems. However, DNNs are found to be vulnerable to adversarial examples, which are carefully crafted inputs aiming to mislead a learner to make incorrect predictions. While several defense and detection approaches are proposed for static image classification, many security-critical tasks use videos as their input and require efficient processing. In this paper, we propose an efficient and effective method advIT to detect adversarial frames within videos against different types of attacks based on temporal consistency property of videos. In particular, we apply optical flow estimation to the target and previous frames to generate pseudo frames and evaluate the consistency of the learner output between these pseudo frames and target. High inconsistency indicates that the target frame is adversarial. We conduct extensive experiments on various learning tasks including video semantic segmentation, human pose estimation, object detection, and action recognition, and demonstrate that we can achieve above 95% adversarial frame detection rate. To consider adaptive attackers, we show that even if an adversary has access to the detector and performs a strong adaptive attack based on the state of the art expectation of transformation method, the detection rate stays almost the same. We also tested the transferability among different optical flow estimators and show that it is hard for attackers to attack one and transfer the perturbation to others. In addition, as efficiency is important in video analysis, we show that advIT can achieve real-time detection in about 0.03--0.4 seconds.

Deep learning technologies, which are the key components of state-of-the-art Artificial Intelligence (AI) services, have shown great success in providing human-level capabilities for a variety of tasks, such as visual analysis, speech recognition, and natural language processing and etc. Building a production-level deep learning model is a non-trivial task, which requires a large amount of training data, powerful computing resources, and human expertises. Therefore, illegitimate reproducing, distribution, and the derivation of proprietary deep learning models can lead to copyright infringement and economic harm to model creators. Therefore, it is essential to devise a technique to protect the intellectual property of deep learning models and enable external verification of the model ownership.

Voice controlled system becomes increasingly popular these days due to the convenient and natural control over lots of functionalities and smart devices. Amazon Echo, designed around Alexa, is capable of controlling smart devices such as locks, sending emails, making phone calls, and even bridging the gap between online services such as Twitter, Facebook, etc. Previously, researchers demonstrated that by carefully crafting obfuscated commands or transmitting commands over ultrasound carrier, voice controlled systems can be compromised without people's awareness. However, those researches require the target voice controlled systems to be close enough to their speaker or ultrasound transducer. In this paper, we proposed REEVE (REmotE VoicE control) attack that can manipulate Amazon Alexa remotely, e.g., via signal broadcasting to compromise radio, TV, speaker, etc. It works on behalf of the attackers to operate various commands beneficial to them. By analyzing more than 15,000 Alexa skills and 600 IFTTT Applets related to Alexa, we found that more than 100 of them can be used to attack Echo. We also thoroughly scrutinized the attack surface of Echo's voice control and conducted security analysis based on different consequences.

Tensor decomposition methods are effective tools for modelling multidimensional array data (i.e., tensors). Among them, nonparametric Bayesian models, such as Infinite Tucker Decomposition (InfTucker), are more powerful than multilinear factorization approaches, including Tucker and PARAFAC, and usually achieve better predictive performance. However, they are difficult to handle massive data due to a prohibitively high training cost. To address this limitation, we propose Distributed infinite Tucker (DinTucker), a new hierarchical Bayesian model that enables local learning of InfTucker on subarrays and global information integration from local results. We further develop a distributed stochastic gradient descent algorithm, coupled with variational inference for model estimation. In addition, the connection between DinTucker and InfTucker is revealed in terms of model evidence. Experiments demonstrate that DinTucker maintains the predictive accuracy of InfTucker and is scalable on massive data: On multidimensional arrays with billions of elements from two real-world applications, DinTucker achieves significantly higher prediction accuracy with less training time, compared with the state-of-the-art large-scale tensor decomposition method, GigaTensor.

Firewall policies are notorious for having misconfiguration errors which can defeat its intended purpose of protecting hosts in the network from malicious users. We believe this is because today's firewall policies are mostly monolithic. Inspired by ideas from modular programming and code refactoring, in this work we introduce three kinds of modules: primary, auxiliary, and template, which facilitate the refactoring of a firewall policy into smaller, reusable, comprehensible, and more manageable components. We present algorithms for generating each of the three modules for a given legacy firewall policy. We also develop ModFP, an automated tool for converting legacy firewall policies represented in access control list to their modularized format. With the help of ModFP, when examining several real-world policies with sizes ranging from dozens to hundreds of rules, we were able to identify subtle errors.

One of Android's main defense mechanisms against malicious apps is a risk communication mechanism which, before a user installs an app, warns the user about the permissions the app requires, trusting that the user will make the right decision. This approach has been shown to be ineffective as it presents the risk information of each app in a “stand-alone” fashion and in a way that requires too much technical knowledge and time to distill useful information. We discuss the desired properties of risk signals and relative risk scores for Android apps in order to generate another metric that users can utilize when choosing apps. We present a wide range of techniques to generate both risk signals and risk scores that are based on heuristics as well as principled machine learning techniques. Experimental results conducted using real-world data sets show that these methods can effectively identify malware as very risky, are simple to understand, and easy to use.

Relying on an access control security policy alone to protect valuable resources is a dangerous practice. Prudent security must engage in other risk management and mitigation techniques to rapidly detect and recover from breaches. In reality, many security policies are either wrong, containing errors, or are misused and abused by malicious employees or compromised accounts; not all granted access is desirable. A popular approach to mitigate against these and other residual threats is to monitor applications to detect misuse and abuse of credentials in near real-time.

Organizations rarely define formal security properties or policies for their access control systems, often choosing to react to changing needs. This paper addresses the problem of reconciling entitlement usage with configured policies for multiple objectives: policy optimization and risk mitigation. Policies should remain up-to-date, maintaining least privilege, and using unambiguous constructs that reduce administrative stress.

This panel will address the following question. Does an increase in the granularity of access control systems produce a measurable reduction in risk and help meet the goals of the organization, or is the cost prohibitively high?

One of Android's main defense mechanisms against malicious apps is a risk communication mechanism which, before a user installs an app, warns the user about the permissions the app requires, trusting that the user will make the right decision. This approach has been shown to be ineffective as it presents the risk information of each app in a "tand-alone" ashion and in a way that requires too much technical knowledge and time to distill useful information.

This paper addresses the making of security decisions, such as access-control decisions or spam filtering decisions, under uncertainty, when the benefit of doing so outweighs the need to absolutely guarantee these decisions are correct. For instance, when there are limited, costly, or failed communication channels to a policy-decision-point. Previously, local caching of decisions has been proposed, but when a correct decision is not available, either a policy-decision-point must be contacted, or a default decision used. We improve upon this model by using learned classifiers of access control decisions. These classifiers, trained on known decisions, infer decisions when an exact match has not been cached, and uses intuitive notions of utility, damage and uncertainty to determine when an inferred decision is preferred over contacting a remote PDP. Clearly there is uncertainty in the predicted decisions, introducing a degree of risk. Our solution proposes a mechanism to quantify the uncertainty of these decisions and allows administrators to bound the overall risk posture of the system. The learning component continuously refines its models based on inputs from a central policy server in cases where the risk is too high or there is too much uncertainty. We have validated our models by building a prototype system and evaluating it with requests from real access control policies. Our experiments show that over a range of system parameters, it is feasible to use machine learning methods to infer access control policies decisions. Thus our system yields several benefits, including reduced calls to the PDP, reducing latency and communication costs; increased net utility; and increased system survivability.

An understanding of information flow has many applications, including for maximizing marketing impact on social media, limiting malware propagation, and managing undesired disclosure of sensitive information. This paper presents scalable methods for both learning models of information flow in networks from data, based on the Independent Cascade Model, and predicting probabilities of unseen flow from these models. Our approach is based on a principled probabilistic construction and results compare favourably with existing methods in terms of accuracy of prediction and scalable evaluation, with the addition that we are able to evaluate a broader range of queries than previously shown, including probability of joint and/or conditional flow, as well as reflecting model uncertainty. Exact evaluation of flow probabilities is exponential in the number of edges and naive sampling can also be expensive, so we propose sampling in an efficient Markov-Chain Monte-Carlo fashion using the Metropolis-Hastings algorithm -- details described in the paper. We identify two types of data, those where the paths of past flows are known -- attributed data, and those where only the endpoints are known -- unattributed data. Both data types are addressed in this paper, including training methods, example real world data sets, and experimental evaluation. In particular, we investigate flow data from the Twitter microblogging service, exploring the flow of messages through retweets (tweet forwards) for the attributed case, and the propagation of hash tags (metadata tags) and urls for the unattributed case.

This paper describes our system, built as part of a commercially available product, for inferring the risk in an RBAC policy model, i.e., the assignment of permissions to roles and roles to users. Our system implements a general model of risk based on any arbitrary set of properties of permissions and users. Our experience shows that fuzzy inferencing systems are best suited to capture how humans assign risk to such assignments. To implement fuzzy inferencing practically we need the axiom of monotonicity, i.e., risk can not decrease when more permissions are assigned to a role or when the role is assigned to fewer users. We describe the visualization component which administrators can use to infer aggregate risk in role assignments as well as drill down into which assignments are actually risky. Administrators can then use this knowledge to refactor roles and assignments.

We consider a fundamentally new approach to role and policy mining: finding RBAC models which reflect the observed usage of entitlements and the attributes of users. Such policies are interpretable, i.e., there is a natural explanation of why a role is assigned to a user and are conservative from a security standpoint since they are based on actual usage. Further, such "generative" models provide many other benefits including reconciliation with policies based on entitlements, detection of provisioning errors, as well as the detection of anomalous behavior. Our contributions include defining the fundamental problem as extensions of the well-known role mining problem, as well as providing several new algorithms based on generative machine learning models. Our algorithms find models which are causally associated with actual usage of entitlements and any arbitrary combination of user attributes when such information is available. This is the most natural process to provision roles, thus addressing a key usability issue with existing role mining algorithms.

The phenomenal growth of the Android platform in the past few years has made it a lucrative target of malicious application (app) developers. There are numerous instances of malware apps that send premium rate SMS messages, track users' private data, or apps that, even if not characterized as malware, conduct questionable actions affecting the user's privacy or costing them money. In this paper, we investigate the feasibility of using both the permissions an app requests, the category of the app, and what permissions are requested by other apps in the same category to better inform users whether the risks of installing an app is commensurate with its expected benefit. Existing approaches consider only the risks of the permissions requested by an app and ignore both the benefits and what permissions are requested by other apps, thus having a limited effect. We propose several risk signals that and evaluate them using two datasets, one consists of 158,062 Android apps from the Android Market, and another consists of 121 malicious apps. We demonstrate the effectiveness of our proposal through extensive data analysis.

Several anonymization techniques, such as generalization and bucketization, have been designed for privacy preserving microdata publishing. Recent work has shown that generalization loses considerable amount of information, especially for high-dimensional data. Bucketization, on the other hand, does not prevent membership disclosure and does not apply for data that do not have a clear separation between quasi-identifying attributes and sensitive attributes. In this paper, we present a novel technique called slicing, which partitions the data both horizontally and vertically. We show that slicing preserves better data utility than generalization and can be used for membership disclosure protection. Another important advantage of slicing is that it can handle high-dimensional data. We show how slicing can be used for attribute disclosure protection and develop an efficient algorithm for computing the sliced data that obey the ℓ-diversity requirement. Our workload experiments confirm that slicing preserves better utility than generalization and is more effective than bucketization in workloads involving the sensitive attribute. Our experiments also demonstrate that slicing can be used to prevent membership disclosure.

SyferLock presents a one-time password system, GridCode, that allows an unaided human to authenticate, reducing the cost of deployment. The one-time password system is a human computable challenge-response protocol which they claim defends against key-logging, replay, and brute force attacks, among others. We evaluate the security of the Grid-Code one-time password system and challenge these claims. We identify weak preimage resistance and character independence as key weaknesses of the GridCode system, leading to a variety of attacks. Our analysis indicates their scheme is akin to providing an adversary the ability to perform a brute force attack on a user's password in parallel without significant effort, lowering the effort required to recover a strong user password. Given a small number of challenge-response pairs, an adversary can recover a user's password (e.g., 2--4 pairs), and additional secret (e.g., 1 pair).

There has been increasing interest in automatic techniques for generating roles for role based access control, a process known as role mining. Most role mining approaches assume the input data is clean, and attempt to optimize the RBAC state. We examine role mining with noisy input data and suggest dividing the problem into two steps: noise removal and candidate role generation. We introduce an approach to use (non-binary) rank reduced matrix factorization to identify noise and experimentally show that it is effective at identifying noise in access control data. User- and permission-attributes can further be used to improve accuracy. Next, we show that our two-step approach is able to find candidate roles that are close to the roles mined from noise-less data. This method performs better than the approach of mining noisy data directly and offering the administrator increased control in the noise removal and candidate role generation phases. We note that our approach is applicable outside role engineering and may be used to identify errors or predict missing values in any access control matrix.

With the growing adoption of Role-Based Access Control (RBAC) in commercial security and identity management products, how to facilitate the process of migrating a non-RBAC system to an RBAC system has become a problem with significant business impact. Researchers have proposed to use data mining techniques to discover roles to complement the costly top-down approaches for RBAC system construction. An important problem is how to construct RBAC systems with low complexity. In this article, we define the notion of weighted structural complexity measure and propose a role mining algorithm that mines RBAC systems with low structural complexity. Another key problem that has not been adequately addressed by existing role mining approaches is how to discover roles with semantic meanings. In this article, we study the problem in two primary settings with different information availability. When the only information is user-permission relation, we propose to discover roles whose semantic meaning is based on formal concept lattices. We argue that the theory of formal concept analysis provides a solid theoretical foundation for mining roles from a user-permission relation. When user-attribute information is also available, we propose to create roles that can be explained by expressions of user-attributes. Since an expression of attributes describes a real-world concept, the corresponding role represents a real-world concept as well. Furthermore, the algorithms we propose balance the semantic guarantee of roles with system complexity. Finally, we indicate how to create a hybrid approach combining top-down candidate roles. Our experimental results demonstrate the effectiveness of our approaches.

The recent interest in outsourcing IT services onto the cloud raises two main concerns: security and cost. One task that could be outsourced is data mining. In VLDB 2007, Wong et al. propose an approach for outsourcing association rule mining. Their approach maps a set of real items into a set of pseudo items, then maps each transaction non-deterministically. This paper, analyzes both the security and costs associated with outsourcing association rule mining. We show how to break the encoding scheme from Wong et al. without using context specific information and reduce the security to a one-to-one mapping. We present a stricter notion of security than used by Wong et al., and then consider the practicality of outsourcing association rule mining. Our results indicate that outsourcing association rule mining may not be practical, if the data owner is concerned with data confidentiality.

While many role mining algorithms have been proposed in recent years, there lacks a comprehensive study to compare these algorithms. These role mining algorithms have been evaluated when they were proposed, but the evaluations were using different datasets and evaluation criteria. In this paper, we introduce a comprehensive framework for evaluating role mining algorithms. We categorize role mining algorithms into two classes based on their outputs; Class 1 algorithms output a sequence of prioritized roles while Class 2 algorithms output complete RBAC states. We then develop techniques that enable us to compare these algorithms directly. We also introduce a new role mining algorithm and two new ways for algorithmically generating datasets for evaluation. Using synthetic as well as real datasets, we compared nine role mining algorithms. Our results illustrate the strengths and weaknesses of these algorithms.

With the increasing need to securely share information, current access control systems are proving too in flexible and difficult to adapt. Recent work on risk-based access control systems has shown promise at resolving the inadequacies of traditional access control systems, and promise to increase information sharing and security. We consider some of the core open problems in risk-based access control systems, namely where and how much risk to take. We propose the use of market mechanisms to determine an organization's risk tolerance and allocation. We show that with the correct incentives, an employee will make optimal choices for the organization. We also comment on how the market can be used to ensure employees behave honestly and detect those who are malicious. Through simulations, we empirically show the advantage of risk-based access control systems and market mechanisms at increasing information sharing and security.

With the growing adoption of role-based access control (RBAC) in commercial security and identity management products, how to facilitate the process of migrating a non-RBAC system to an RBAC system has become a problem with significant business impact. Researchers have proposed to use data mining techniques to discover roles to complement the costly top-down approaches for RBAC system construction. A key problem that has not been adequately addressed by existing role mining approaches is how to discover roles with semantic meanings. In this paper, we study the problem in two settings with different information availability. When the only information is user-permission relation, we propose to discover roles whose semantic meaning is based on formal concept lattices. We argue that the theory of formal concept analysis provides a solid theoretical foundation for mining roles from userpermission relation. When user-attribute information is also available, we propose to create roles that can be explained by expressions of user-attributes. Since an expression of attributes describes a real-world concept, the corresponding role represents a real-world concept as well. Furthermore, the algorithms we proposed balance the semantic guarantee of roles with system complexity. Our experimental results demonstrate the effectiveness of our approaches.