Deep Learning has been shown to be particularly vulnerable to adversarial samples. To combat adversarial strategies, numerous defensive techniques have been proposed. Among these, a promising approach is to use randomness in order to make the classification process unpredictable and presumably harder for the adversary to control. In this paper, we study the effectiveness of randomized defenses against adversarial samples. To this end, we categorize existing state-of-the-art adversarial strategies into three attacker models of increasing strength, namely blackbox, graybox, and whitebox (a.k.a. adaptive) attackers. We also devise a lightweight randomization strategy for image classification based on feature squeezing, that consists of pre-processing the classifier input by embedding randomness within each feature, before applying feature squeezing. We evaluate the proposed defense and compare it to other randomized techniques in the literature via thorough experiments. Our results indeed show that careful integration of randomness can be effective against both graybox and blackbox attacks without significantly degrading the accuracy of the underlying classifier. However, our experimental results offer strong evidence that in the present form such randomization techniques cannot deter a whitebox adversary that has access to all classifier parameters and has full knowledge of the defense. Our work thoroughly and empirically analyzes the impact of randomization techniques against all classes of adversarial strategies.

Copresence verification based on context can improve usability and strengthen security of many authentication and access control systems. By sensing and comparing their surroundings, two or more devices can tell whether they are copresent and use this information to make access control decisions. To the best of our knowledge, all context-based copresence verification mechanisms to date are susceptible to context-manipulation attacks. In such attacks, a distributed adversary replicates the same context at the (different) locations of the victim devices, and induces them to believe that they are copresent. In this paper we propose DoubleEcho, a context-based copresence verification technique that leverages acoustic Room Impulse Response (RIR) to mitigate context-manipulation attacks. In DoubleEcho, one device emits a wide-band audible chirp and all participating devices record reflections of the chirp from the surrounding environment. Since RIR is, by its very nature, dependent on the physical surroundings, it constitutes a unique location signature that is hard for an adversary to replicate. We evaluate DoubleEcho by collecting RIR data with various mobile devices and in a range of different locations. We show that DoubleEcho mitigates context-manipulation attacks whereas all other approaches to date are entirely vulnerable to such attacks. DoubleEcho detects copresence (or lack thereof) in roughly 2 seconds and works on commodity devices.

Context-centric sensor-based proximity detection (or, contextual co-presence detection) is a promising approach to defend against relay attacks in many mobile authentication systems, especially against unattended terminals (such as cars parked in unmonitored parking lots, remote gas station pumps, or stolen laptops). Prior work demonstrated the effectiveness of a variety of contextual sensor modalities for this purpose, including audio-radio environment (ambient audio, Wi-Fi, Bluetooth, and GPS, and combinations thereof) and physical environment (temperature, humidity, gas, and altitude, and combinations thereof). In this paper, we present a systematic assessment of such co-presence detection in the presence of a strong, context-manipulating attacker against unattended terminals. First, we show that it is feasible to manipulate, consistently control, and stabilize the readings of different acoustic and physical environment sensors (and even multiple sensors simultaneously) using low-cost, off-the-shelf equipment. Specifically, we show that it is possible to control the temperature using a home-grade hair dryer, affect the gas readings using a smoking cigarette, impact the altitude/pressure with a simple air compressor, or relay audio signals recorded at one end to the other thereby causing both sides to perceive a very similar acoustic environment. Second, based on these capabilities and the strengthened threat model, we show that an attacker who can manipulate the context gains a significant advantage in defeating contextual co-presence detection. For systems that use multiple sensors, we investigate two sensor fusion approaches based on machine learning classification techniques-features-fusion and decisions-fusion, and show that both are vulnerable to context manipulation attacks but the latter approach can be more resistant in some cases. We further consider other defensive approaches that may be used to reduce the impact of even such a strong context-manipulating attacker. Our work represents the first concrete step towards analyzing, extending, and systematizing prior work on contextual co-presence detection under a stronger, but realistic adversarial model.

Zero-Interaction Authentication (ZIA) refers to approaches that authenticate a user to a verifier (terminal) without any user interaction. Currently deployed ZIA solutions are predominantly based on the terminal detecting the proximity of the user's personal device, or a security token, by running an authentication protocol over a short-range wireless communication channel. Unfortunately, this simple approach is highly vulnerable to low-cost and practical relay attacks which completely offset the usability benefits of ZIA. The use of contextual information, gathered via on-board sensors, to detect the co-presence of the user and the verifier is a recently proposed mechanism to resist relay attacks. In this paper, we systematically investigate the performance of different sensor modalities for co-presence detection with respect to a standard Dolev-Yao adversary. First, using a common data collection framework run in realistic everyday settings, we compare the performance of four commonly available sensor modalities (WiFi, Bluetooth, GPS, and Audio) in resisting ZIA relay attacks, and find that WiFi is better than the rest. Second, we show that, compared to any single modality, fusing multiple modalities improves resilience against ZIA relay attacks while retaining a high level of usability. Third, we motivate the need for a stronger adversarial model to characterize an attacker who can compromise the integrity of context sensing itself. We show that in the presence of such a powerful attacker, each individual sensor modality offers very low security. Positively, the use of multiple sensor modalities improves security against such an attacker if the attacker cannot compromise multiple modalities simultaneously.

There is little information from independent sources in the public domain about mobile malware infection rates. The only previous independent estimate (0.0009%) [11], was based on indirect measurements obtained from domain-name resolution traces. In this paper, we present the first independent study of malware infection rates and associated risk factors using data collected directly from over 55,000 Android devices. We find that the malware infection rates in Android devices estimated using two malware datasets (0.28% and 0.26%), though small, are significantly higher than the previous independent estimate. Based on the hypothesis that some application stores have a greater density of malicious applications and that advertising within applications and cross-promotional deals may act as infection vectors, we investigate whether the set of applications used on a device can serve as an indicator for infection of that device. Our analysis indicates that, while not an accurate indicator of infection by itself, the application set does serve as an inexpensive method for identifying the pool of devices on which more expensive monitoring and analysis mechanisms should be deployed. Using our two malware datasets we show that this indicator performs up to about five times better at identifying infected devices than the baseline of random checks. Such indicators can be used, for example, in the search for new or previously undetected malware. It is therefore a technique that can complement standard malware scanning. Our analysis also demonstrates a marginally significant difference in battery use between infected and clean devices.

Within last years multi-synchronous collaborative editing systems became widely used. Multi-synchronous collaboration maintains multiple, simultaneous streams of activity which continually diverge and synchronized. These streams of activity are represented by means of logs of operations, i.e. user modifications. A malicious user might tamper his log of operations. At the moment of synchronization with other streams, the tampered log might generate wrong results. In this paper, we propose a solution relying on hash-chain based authenticators for authenticating logs that ensure the authenticity, the integrity of logs, and the user accountability. We present algorithms to construct authenticators and verify logs. We prove their correctness and provide theoretical and practical evaluations.

In multi-synchronous collaboration users replicate shared data, modify it and redistribute modified versions of this data without the need of a central authority. However, in this model, no usage restriction mechanism was proposed to control what users can do with the data after it has been released to them. In this paper, we extend the multi-synchronous collaboration model with contracts that express usage restrictions and that are checked a posteriori by users when they receive the modified data. We propose a merging algorithm that deals not only with changes on data but also with contracts. A log auditing protocol is used to detect users who do not respect contracts and to adjust user trust levels. Our contract-based model was implemented and evaluated by using PeerSim simulator.

In the push-pull-clone collaborative editing model widely used in distributed version control systems users replicate shared data, modify it and redistribute modified versions of this data without the need of a central authority. However, in this model no usage restriction mechanism is proposed to c

In order to overcome the disadvantages of a central authority, a tendency is to move towards a peer-to-peer collaboration where control over data is given to users who can decide with whom they want to share their private data. In this peer-to-peer collaboration it is very difficult to ensure that after data is shared with other peers, these peers will not misbehave and violate data privacy. In this paper, we propose a mechanism that addresses the issue of data privacy violation by auditing the collaboration logs. In our approach, trust values between users are adjusted according to their previous activities on the shared data. Users share their private data by specifying some obligations the receivers are expected to follow. We log modifications done by users as well as the obligations and use a log-auditing mechanism to detect users who misbehaved. We adjust their associated trust values by using any existing decentralized trust model.