WiSec 2016

This paper proposes a scheme for verifying motion claims (i.e., speed and direction) of mobile devices. The problem is motivated by the criticality of spatial information in vehicular and aerial transportation systems (e.g., unauthenticated ADS-B air transportation system). The proposed scheme relies on measuring the doppler shift by multiple receivers to detect fabricated claims. The authors show that with two verifiers (receivers) the adversary is significantly limited (to a hyperbola) and with three verifiers the adversary is further restricted and will be detected unless if at least two verifiers are on a straight line from the perspective of the claimed location.

The reviewers appreciated that the paper is grounded in theoretically sound techniques, and that the authors built a complete system using the Ettus USRP X300 and experimented with real data from the ADS-B air transportation system. They put significant effort in adapting the proposed scheme and optimizing it for real wireless communication constraints (extending existing libraries such as the OpenSky one), as well as identifying adequate thresholds for low false-positives and negatives. The reviewers expressed some doubts about the approach, e.g., the lack of resiliency to an adversary equipped with multiple antennas.

This paper introduces a software defined radio stack for 802.11 (up to 54Mbps) that is used to fingerprint Wi-Fi devices. This approach is much cheaper than using a signal analyzer to perform physical layer fingerprinting. It also allows control over the entire radio and protocol stack. In contrast, a signal analyzer is limited fingerprinting physical layer features, and hardware WiFi device does not expose information from the lower layers. The evaluation is performed on 93 different devices from 13 different models.

The reviewers felt that the paper reflects a significant amount of work to build and evaluate this low cost and very flexible fingerprinting tool and the introduced techniques. They thought that the fingerprinting success rate for model identification is quite good, but that the identification of similar devices is currently rather low. The reviewers hope that this could be improved in the future. Also future work should provide a comparison with similar work using standard fingerprinting metrics such as the ROC (Receiver Operating Characteristic) curves and EER (Equal Error Rates).

"This paper describes the design, implementation, and evaluation of WASp, a tool that automatically reverse-engineers proprietary wireless protocols that run over IEEE 802.15.4. Such protocols are commonly used in IoT systems, and two are studied in this paper: a smart plug system and a platform screen door system. Such protocols often depend on obscurity, because they 1) use proprietary binary encoding for the packets, 2) lack documentation, and 3) do not necessarily use encryption. Therefore it is a worthwhile goal to assess their security through reverse-engineering. The proposed WASp tool takes as input captured packets (pcap files) and some context information (e.g. number of nodes involved in the communication) and aims to first, reconstruct the packet format of the unknown protocol and then second, generate ("spoof") packets that conform to the protocol.

The reviewers appreciated practical problem studied by the paper, noting that it will become more important as IoT deployments increase. The implementation and practical testing of the tool adds value and shows that the protocols studied can be reverse-engineered to some extent. The paper is in general well-written and easy to understand.

The main concern shared by all reviewers is that the evaluation of WASp does not use ground truth (i.e. reverse-engineer known protocols and evaluate WASp by comparing its output to the known packet format). The evaluation metrics used are the "spoofing success rate" (a spoofed packet is considered a success, if any response packet is detected for it) and the "entropy reduction" (this is not formally defined in the paper, but it relates to the reduction in the number of packets that the attacker - before and after running WASp - considers as potentially valid protocol packets). However, due to legal and safety reasons, the "spoofing success rate" cannot always be used, which was the case in the second system studied (platform screen door). Also, the entropy reduction method would benefit from a clearer definition and more justification as an evaluation metric. Overall, wider and more rigorous evaluation is needed, and this is a useful direction for future research.

This paper studies an attack that fingerprints web sites on Android. The attack takes advantage of Android's data-usage statistics, which are available to all apps without requiring permissions. The authors evaluated the classification rate for a standard browser application as well as when the traffic is routed through Tor.

The reviewers appreciated the practical attack demonstration using an unprivileged application, as well as the quantitative measurement of impact. However, the reviewers had several reservations. First, recent literature has proposed many attacks that exploit information extracted from Android without permissions (e.g., citations [22] and [44] discuss fingerprinting app network traffic based on data usage statistics). Second, the evaluation is based on an ideal scenario where the user is browsing a single web site at a time. The reviewers wondered if it would it still be possible to fingerprint web sites when traffic from different tabs mixes. Finally, while the paper proposes countermeasures, the reviewers would have liked to see an evaluation of the countermeasures.

“Wormholes” in the wireless networks literature have been synonymous to attacks: two (or more) adversarial devices establish one (or more) long-haul link(s) (different than those established by the wireless transceivers in use) and mislead those legitimate devices that they are communication neighbours. Countermeasures have been considered, naturally, with one of those defeated by the so-called “adaptive wormhole” technique. The same one is now considered in a positive manner in this paper: as a building block towards establishing shared secrets. Simply put, the adaptive wormhole allows the system inject correlated signals on both ends of the wormhole. The other main building block is a long known approach to extract common randomness (and thus secrecy) based on the wireless channel characteristics (observable on both ends). The paper proposes this approach to generate new session keys shared by two remote wireless devices assisted by two gateway nodes, leveraging an already available initially shared secret key. The motivation is to use the scheme when the wireless devices, for example, sensors in an industrial setting, cannot execute protools such as the ones widely used based on the Diffie-Hellman key establishment.

The review process appreciated the interesting view point and seeing a significant effort put in the system and its analysis. At the same time, a number of interesting questions were raised. For example, the assumed availability of a pre-shared key is it a limitation? Or can it not enable alternative, less involved approaches? What is a real-world scenario that could best fit the protocol elaborated in this paper? How close is the 802.15.4-based instantiation and analysis to the not yet formed 5G standards? The consensus within the TPC was that this combination of technically solid work and interesting questions made the paper a good contribution to the program.

This paper attempts to combat attacks on hardware sensors on mobile devices (e.g., accelerometers) via the injection of noise. Specifically, the Slogger system injects noise via the Android Debug Bridge (ADB), and then assess its ability to thwart attacks from malware implementing touch stroke detection (TapLogger) and touch stroke inference (ACCessory) techniques. The reviewers like this paper because it is one of the few to deal with defenses against this class of attack, which is receiving significant attention in the literature. Reviewers also appreciated the thoroughness of the evaluation. The two main concerns shared by the researchers were that the addition of noise had been previously proposed in other works, especially Marquardt et al. (but not implemented) and that it was unclear how difficult it would be for an adversary to remove such noise. Overall, the paper was reviewed positively.

The paper presents a system (priVy) to provide fine-grained access control to the Android native content providers, such as contacts and calendar. Access to these providers can potentially leak important data or cause other attacks such as denial of service. Therefore the provided solution should help prevent these issues. The technical solution is implemented using aspect-oriented programming and bytecode instrumentation. The experimental results show that runtime overhead is small and the approach works well.

The PC generally liked the paper. The authors presented a solid system with a considerable amount of engineering efforts. The approach, without OS/framework changes, seems to be more easily deployable than some previous solutions. However, the PC felt that the paper could be much stronger if move novel techniques are used, and stronger evaluations are conducted (e.g., using a much larger testing dataset).

This paper provides a deeper understanding of how machine learning techniques fail and succeed at classifying Android malware. Previously developed mobile malware classifiers seem to rely on features that vary widely over time, making them impractical. The main goals of this paper include studying how malware changes over time, identifying invariant features that allow a machine learning classifier to retain value, and allowing for detection of malware that was not seen during the training phase. To support the problem motivation, the paper demonstrates that while syntax-based classifiers work well on training and validation malware collected around the same time, they perform badly when the malware being classified is collected at a later time than the training set.

The paper suggests the use of semantic features of mobile apps to retain classifier value over time, building on the intuition that certain semantic attributes of mobile malware are invariant. Experiments are provided to verify that the incorporation of semantic features can significantly improve the performance of Android malware classification.

The reviewers particularly appreciated the intuitive explanations of an important aspect of how machine learning algorithms succeed and fail in malware classification. The experimental approach is very detailed and lends itself to repeatability and follow-on studies. The authors also provide strong insight into semantic feature selection and its associated issues. By revisiting previous measurement studies, the paper has provided an important challenge to the research community in regard to understanding what machine learning algorithms are actually doing and how they can be improved.

The paper presents Shatter, an open-source framework for Android and desktop devices that uses threshold cryptography for sharing private keys across different devices. The idea is that the private key can then be securely used for identifying a single user across different services that are used on all devices of the user. The solution provides further convenience when the user loses one of the devices, as she does not need to revoke the key pair and create a new private/public key-pair, but rather simply revoke the share used in the lost device. The authors have not only implemented Shatter, but they have also demonstrated how it can be introduced in backwards compatible manner for two real-life applications. The solution is carefully analyzed from the performance point of view.

The reviewers appreciated that Shatter is a significant step forward in bringing threshold cryptography from theory to practice. However, concerns were raised about merits of the paper in terms of purely scientific contributions. Shatter enables further scientific studies (e.g., usability) of threshold cryptography-based mechanisms, but leaves these studies for future work. Milder concerns were raised on several issues. First, delays in communication are relatively long and performance would need to be improved significantly before wider adoption could be expected. Second, reviewers questioned whether automatic acknowledgements of signing requests would provide a sensible trade-off between security and usability in this context.

This paper looks at the orthogonal blinding technique: a physical layer security approach in which the transmitter sends information directed to the receiver (using beamforming) and sends noise toward the eavesdropper (in the case of known CSI) or in the null space of the receiver (in the case of unknown CSI). Authors discuss the security of orthogonal blinding in a multi-antenna eavesdropper scenario with security evaluations based on the number of the antennas. The paper categorizes the security based on the availability of the eavesdropper CSI (i.e., Honest/Dishonest Eve) and the number of antennas (i.e., Inferior/Superior Eve). Next, the paper focuses on the ciphertext-only attacks and proposes an algorithm, which uses the low-entropy parts of the packets to attack the aforementioned method. The algorithm is implemented in MATLAB.

The reviewers appreciated the topic, noting the importance of making physical layer approaches practical. They also found the paper well-structured and interesting to read, making use of intuitive discussions in many parts. Furthermore, the reveiwers appreciated the MATLAB implementation of the attack algorithm. However, the reviewers felt that the technical contribution of the paper could be strengthened. For example, the security level categorization uses known techniques and the success of the proposed attack algorithm does not have a proof.

This paper proposes and implements a reactive jamming attack against popular wireless systems such as Wi-Fi, which employ OFDM communications at the physical layer. The attack specifically targets the interleaving/convolutional coding function to reduce the energy necessary for denying wireless transmissions and also minimize the adversary’s presence on the wireless channel. The proposed jamming technique takes advantage of the publicly known interleaver structure to “surgically” insert shorts bursts of energy across subcarriers such that errors appear in long bursts after de-interleaving. These long bursts overwhelm the error correction capability of the decoder. In testbed experiments on low-cost SDRs and commercial Wi-Fi cards, interleaver jamming was shown to completely block Wi-Fi transmissions with less than 1% of average jamming power relative to the received signal power at the targeted receiver.

The reviewers appreciated the thoroughness in the treatise of the problem. The authors presented a detailed analysis of the interleaver/convolutional encoder structure in OFDM systems and justified the design of their attack. Moreover, the attack extended beyond previous theoretical approaches that focused on the temporal aspect of interleaving and exploited interleaving both in frequency and time. The extensive testbed experimentation on an SDR platform and on commercial Wi-Fi cards was highly valued. The reviewers agreed that the experimental results clearly demonstrate the performance advantages of the proposed jamming method. Some concerns were raised about the omission of secondary costs related to launching intelligent attacks such as the cost of acquiring frame synchronization. Moreover, the reviewers pointed out that the power advantage over random jamming could be diminished if cryptographic interleaving is applied. Overall, the PC was confident that jamming designed to target the interleaving function has a detrimental impact in OFDM communications.

The proliferation of embedded devices is a double-edged sword: it enables a multitude of applications but, at the same, it provides ‘opportunities’ for compromising such unattended (and resource-limited) devices. The remedy has been to leverage "remote attestation” techniques: a so-call "verifier” (the trusted entity) executes a protocol with a “prover” (the potentially compromised device), to assess the integrity of the software running on the prover. The challenge is how to execute such remote attestation protocols efficiently for a large system (network) of possibly heterogeneous devices; simply put, with overall cost lower than running a series of individual remote attestations. The paper considers most notably the case of physical compromise of the embedded devices. The intuition of the related defence proposed here lies in that any physical compromise will result in a sizeable, perceptible disconnection of the device. Then, software and software-physical compromise (“hybrid”) attacks are also considered. The review process appreciated the simplicity and the extent of the problem treatment offered by the paper, while, at the same time, it identified the limits of the proposed schemes. All in all, the feeling was that this was a contribution that can generate interest within the WiSec community.

Estimating the distance (or, typically, a bound of the distance) of two nodes (devices) leveraging RF communication in the presence of adversaries has received extensive attention in the literature. A class of attacks, termed "terrorist fraud attacks", are perpetrated by a remote malicious “prover” and its accomplice, a device that is close(r) to the “verifier” (to be misled that the prover is not as far as it actually is). The approach to thwart a terrorist fraud relies on the assumption that the malicious prover would not be willing to leak private information to its accomplice (e.g., its private, signing key). Of course, if it did so willingly, the attack would be easily successful. Along these lines, this paper works on the same approach while adding anonymous authentication: a "session key" is signed with a “group signature” and an alternative mechanism (that could make the verifier provide anyone with 'enough information' the session key) is added to deter the malicious prover from leaking information.

The reviewing process appreciated the novelty of the proposed scheme and the solid technical treatment of the problem at hand. At the same time, it brought forth a number of discussion or, possibly, debate, points. For example: under what conditions can the protocol defend against an accomplice that obtained cryptographically transformed information (granted, not a private key)? What is the spectrum of protection one can consider and how meaningful is it in this context to do so based on probabilities? What is a use-case or real-world scenario that could best fit the protocol elaborated in this paper? In the end, the consensus within the TPC was that this combination of technically solid work and the stimulating questions it raises, would make the paper a good contribution to the program.

This short paper challenges the assumption that IoT devices within a home are protected from external attackers by NAT and/or firewalls. To do so, the authors create a proof-of-concept iOS application that discovers IoT devices and uses Universal Plug-n-Play to reconfigure routers to enable external access. The paper describes how this functionality could be bundled along with an otherwise benign iOS application. While the reviewers had concerns over the technical novelty of the described attacks, they felt that the message was important for the community and therefore was sufficient contribution for a short paper.

A variety of systems perform dynamic analysis of Android applications in an automated manner. However, malicious applications may disguise harmful behavior if they can detect when they are executed in an emulator. This paper proposes several techniques to perform the detection of dynamic analysis systems. This includes examining the parameters of API calls and comparing the patterns of interactions with what is expected from humans. To evaluate their techniques, the authors submit their test application to several application analysis services. The paper provides a comparison of the analysis systems based on how resilient they are to the described detection of programmed interactions.

The program committee liked the fact that the approach is effective for detection of emulators. In addition, the provided comparison of analysis services shows interesting insight into the state-of-the-art of academic and commercial dynamic analysis services. However, the PC would have liked to see a more detailed technical description of the techniques that were applied, such as the specific algorithms used to detect programmed interaction patterns. Also, some of the proposed techniques seem relatively easy to defeat. Finally, the authors could have provided a more detailed analysis of their evaluation and the implications of their findings. Nevertheless, the PC felt that the paper will be a nice addition to the program of WiSec and hopefully spark some interesting discussions.

It is often assumed that Ethernet cables are less prone to eavesdropping than wireless technologies when attackers do not have physical access to the network cable and devices. This paper challenges that assumption by performing a wireless probing attack on 10BASE-T Ethernet over twisted-pair cables, comparing success for various types of Ethernet cable shielding. The results show that near-field probes can capture data exchanged through the Ethernet wire without tampering with it or damaging it in any way.

The reviewers found the implementation and evaluation of the attack compelling and believe the paper is an interesting experiment to practically prove the scientific result discovered by Bell Labs in the 1940s about information leakage by electromagnetic radiation. It is valuable that the authors discuss the feasibility of the attack on 100BASE-TX and 1000BASE_T and suggest countermeasures to protect against the attack.

Although the authors argue the practicality of the approach against more advanced technologies, the reviewers feel the need for real-world experiments to demonstrate the validity of these arguments and feasibility of the attack in realistic scenarios. The countermeasures suggested by the paper sound reasonable (although not novel). However, such countermeasure might be costly and impractical in real-world settings.

The paper proposes a machine learning technique to identify mobile applications based on the network traffic they generate at launch. While prior work has performed similar experiments focusing application-layer data (using deep packet inspection), the experiments in this paper use only TCP/IP headers. Indeed, with the increasing use of end-to-end encryption, deep packet inspection may not always provide meaningful results.

The reviewers particularly liked the large scale data collection that was done for this paper. For their analysis, the authors performed over 86,000 app launches on 4 distinct devices, which provided plenty of training data for their supervised learning method. The authors demonstrated that apps can be identified with relatively high accuracy based only on their launch traffic.

A common theme throughout the reviews was that despite being a short paper, the authors managed to pack a significant amount of insight beyond simply identifying applications. For instance, the authors also looked at the impact of application updates and different devices on detection accuracy.

This short paper presents a method to identify devices despite MAC address randomization through the observation of network scan timings. This method can be implemented with an off-the-shelf Wi-Fi interface and uses a device signature based on inter-frame arrival times of probe requests. This signature can be used to group together frames coming from the same device even though these use different MAC addresses.

The reviewers appreciated the problem, particularly given that MAC randomization is still a relatively new countermeasure. They also liked how the authors clearly outlined the proposed method while providing valuable insight. The initial results along with the discussion on the effectiveness of different distance metrics are promising. The reviewers encourage this preliminary work to be developed further, especially with regards to 1) a larger and better defined data set, 2) a better explanation of how signature are constructed and parameters are chosen for realistic network conditions, and 3) more detailed evaluation and further discussion of the results.

This paper looks at the security of authenticated and secure out-of-band (AS-OOB) channels that are based on the use of mechanical vibrations to communicate a pairing secret between two devices. Some earlier work has shown such vibration-based channels to be vulnerable against audio eavesdropping, as the typical generation of vibrations also causes audio leakage. As a remedy against such eavesdropping attacks, this paper presents “Vibreaker”, a system that protects the vibration- based pairing against eavesdropping adversaries by using masking signals emitted by the pairing devices. The purpose of the masking signal is to cloak the acoustic leakage emanating from the vibrations used in the pairing process. Two kinds of masking signals are analysed, white noise and fake vibration noise.

The paper first demonstrates the feasibility of eavesdropping attacks based on audio leakage by reimplementing the attack presented in literature. It then presents the design of the countermeasure that is based on emitting audio-based cloaking signals during the pairing process. The effectiveness of the countermeasure is evaluated by qualitative analysis of the resulting audio signal that an eavesdropper adversary observes at a relatively close distance to the pairing devices. The evaluation shows that the introduction of the cloaking signal effectively masks the acoustic signal emanating from the pairing vibrations. This makes it very difficult for the adversary to retrieve the correct pairing secret based on the audio signal it can observe.

While this paper does a good job in describing the countermeasure and reasoning for its efficiency, the evaluation remains at a qualitative level, as no detailed numeric data about the performed tests and their success rates are provided. Another point of criticism is that only a basic method for decoding the pairing signal by the adversary is assumed. As the authors acknowledge, further studies would be necessary to verify the efficacy of the proposed countermeasure also against adversaries that utilize more advanced methods based on, e.g., machine-learning algorithms to retrieve information about the pairing signal. Nevertheless the paper provides interesting information as a proof-of- concept for utilizing acoustic cloaking for securing vibration-based AS-OOB pairing channels.

This short paper tackles several interesting challenges stemming from the rise of bulk text messages, which are now also used for a variety of non-cellular services such as two-factor authentication. This change limits the usefulness of traditional approaches to characterizing SMS spam. The authors present a novel approach that remarkably improves precision and recall compared to previously proposed classifiers. To empirically evaluate the proposed approach, the authors crawl public SMS gateways and create a novel, larger dataset for SMS, which they have labeled and publicly released. In the process of creating this dataset, the authors also show that existing SMS spam corpora do not sufficiently reflect the prevalence of bulk messages in modern SMS communications.

The reviewers appreciated the experimental work showing that existing spam detection methods do not work well after rapid increase of legitimate bulk traffic, e.g. two factor authentication messages. The authors also managed to fix the problem by significantly increasing the accuracy of these methods. The reviewers were also happy to see that a useful dataset is released for future work. Finally, some of the reviewers had concerns about the ethical implications of collecting messages from public SMS gateways, which may inadvertently contain private information. However, the final consensus of the program committee is that the authors properly handled user privacy in the publicly released dataset.