Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.740511
Title: On the security of mobile sensors
Author: Mehrnezhad, Maryam
ISNI:       0000 0004 7227 1301
Awarding Body: Newcastle University
Current Institution: University of Newcastle upon Tyne
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
The age of sensor technology is upon us. Sensor-rich mobile devices are ubiquitous. Smart-phones, tablets, and wearables are increasingly equipped with sensors such as GPS, accelerometer, Near Field Communication (NFC), and ambient sensors. Data provided by such sensors, combined with the fast-growing computational capabilities on mobile platforms, offer richer and more personalised apps. However, these sensors introduce new security challenges to the users, and make sensor management more complicated. In this PhD thesis, we contribute to the field of mobile sensor security by investigating a wide spectrum of open problems in this field covering attacks and defences, standardisation and industrial approaches, and human dimensions. We study the problems in detail and propose solutions. First, we propose “Tap-Tap and Pay” (TTP), a sensor-based protocol to prevent the Mafia attack in NFC payment. The Mafia attack is a special type of Man-In-The-Middle attack which charges the user for something more expensive than what she intends to pay by relaying transactions to a remote payment terminal. In TTP, a user initiates the payment by physically tapping her mobile phone against the reader. We observe that this tapping causes transient vibrations at both devices which are measurable by the embedded accelerometers. Our observations indicate that these sensor measurements are closely correlated within the same tapping, and different if obtained from different tapping events. By comparing the similarity between the two measurements, the bank can distinguish the Mafia fraud apart from a legitimate NFC transaction. The experimental results and the user feedback suggest the practical feasibility of TTP. As compared with previous sensor-based solutions, ours is the only one that works even when the attacker and the user are in nearby locations or share similar ambient environments. Second, we demonstrate an in-app attack based on a real world problem in contactless payment known as the card collision or card clash. A card collision happens when more than one card (or NFC-enabled device) are presented to the payment terminal’s field, and the terminal does not know which card to choose. By performing experiments, we observe that the implementation of contactless terminals in practice matches neither EMV nor ISO standards (the two primary standards for smart card payment) on card collision. Based on this inconsistency, we propose “NFC Payment Spy”, a malicious app that tracks the user’s contactless payment transactions. This app, running on a smart phone, simulates a card which requests the payment information (amount, time, etc.) from the terminal. When the phone and the card are both presented to a contactless terminal (given that many people use mobile case wallets to travel light and keep wallet essentials close to hand), our app can effectively win the race condition over the card. This attack is the first privacy attack on contactless payments based on the problem of card collision. By showing the feasibility of this attack, we raise awareness of privacy and security issues in contactless payment protocols and implementation, specifically in the presence of new technologies for payment such as mobile platforms. Third, we show that, apart from attacking mobile devices by having access to the sensors through native apps, we can also perform sensor-based attacks via mobile browsers. We examine multiple browsers on Android and iOS platforms and study their policies in granting permissions to JavaScript code with respect to access to motion and orientation sensor data. Based on our observations, we identify multiple vulnerabilities, and propose “TouchSignatures” and “PINLogger.js”, two novel attacks in which malicious JavaScript code listens to such sensor data measurements. We demonstrate that, despite the much lower sampling rate (comparing to a native app), a remote attacker is able to learn sensitive user information such as physical activities, phone call timing, touch actions (tap, scroll, hold, zoom), and PINs based on these sensor data. This is the first report of such a JavaScript-based attack. We disclosed the above vulnerability to the community and major mobile browser vendors classified the problem as high-risk and fixed it accordingly. Finally, we investigate human dimensions in the problem of sensor management. Although different types of attacks via sensors have been known for many years, the problem of data leakage caused by sensors has remained unsolved. While working with W3C and browser vendors to fix the identified problem, we came to appreciate the complexity of this problem in practice and the challenge of balancing security, usability, and functionality. We believe a major reason for this is that users are not fully aware of these sensors and the associated risks to their privacy and security. Therefore, we study user understanding of mobile sensors, specifically their risk perceptions. This is the only research to date that studies risk perceptions for a comprehensive list of mobile sensors (25 in total). We interview multiple participants from a range of backgrounds by providing them with multiple self-declared questionnaires. The results indicate that people in general do not have a good understanding of the complexities of these sensors; hence making security judgements about these sensors is not easy for them. We discuss how this observation, along with other factors, renders many academic and industry solutions ineffective. This makes the security and privacy issues of mobile sensors and other sensorenabled technologies an important topic to be investigated further.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.740511  DOI: Not available
Share: