Date of Award

5-2011

Embargo Period

5-17-2012

Degree Name

Doctor of Philosophy (PhD)

Department

Electrical and Computer Engineering

Advisor(s)

Adrian Perrig

Second Advisor

Virgil D. Gligor

Third Advisor

Ragunathan Rajkumar

Fourth Advisor

Gene Tsudik

Abstract

The proliferation of electronic devices supporting sensing, actuation, and wireless communication enables the monitoring and/or control of a variety of physical systems with digital communication. Such “cyber physical systems” blur the boundaries of the digital and physical worlds, where correct information about the physical world is needed for the correct operation of the digital system. Often in these systems the physical source or destination of information is as important as the information itself. However, the omni-directional and invisible nature of wireless communication makes it difficult to determine communication endpoints. This allows a malicious party to intercept wireless messages or pose as other entities in the system. As such, these systems require new protocols to associate the endpoints of digital communication with physical entities.

Traditional security approaches that associate cryptographic keys with names can help verify endpoints in static systems where a string accurately describes the role of a device. In other systems, the role of a device depends on its physical properties, such as location, which change over time. This dynamic nature implies that identification of an endpoint based on a static name is insufficient. Instead, we can leverage devices’ sensing and actuation capabilities to verify the physical properties and determine the physical endpoints of communication. We investigate three different scenarios where the physical source and/or destination is important and propose endpoint verification techniques: verifying the physical endpoints during an exchange between two smartphones, verifying the receiver of information is in a physical space to enable location-based access control, and verifying the source of information to protect Vehicle-to-Vehicle (V2V) applications. We evaluate our proposals in these systems and show that our solutions fulfill the security requirements while utilizing existing hardware.

Exchanging Information Between Smartphones Shake on it (SHOT) allows users to verify the endpoints during an exchange of information between two smartphones. In our protocol, the phones use their vibrators and accelerometers to establish a human-observable communication channel. The users hold the phones together while the phones use this channel to bootstrap and verify the authenticity of an exchange that occurs over the higher-bandwidth wireless channel. Users can detect the injection of information from other devices as additional vibrations, and prevent such attacks. Our implementation of SHOT for the DROID smartphone is able to support sender and receiver verification during an exchange between two smartphones in 15 seconds on average.

Location-Based Access Control We propose using location-based access control to protect sensitive files on laptops, without requiring any effort from the user to provide security. With a purely wireless electronic system, verifying that a given device is in a physical space is a challenge; either the definition of the physical space is vague (radio waves can travel beyond walls) or the solution requires expensive hardware to measure a message’s time of flight. Instead, we use infrared as a signal that walls can contain. We develop key derivation protocols that ensure only a receiver in the physical room with access to the signal can derive the key. We implement a system that uses the laptop’s webcam to record the infrared signal, derive a key, and decrypt sensitive files in less than 5 seconds.

Source Verification for V2V Networks A number of V2V applications use information about nearby vehicles to prevent accidents or reduce fuel consumption. However, false information about the positioning of vehicles can cause erroneous behavior, including accidents that would not occur in the absence of V2V. As such, we need a way to verify which vehicle sent a message and that the message accurately describes the physical state of that vehicle. We propose using LED lights on vehicles to broadcast the certificate a vehicle is currently using. Receivers can use onboard cameras to film the encoding of the certificate and estimate the relative location of the vehicle. This visual channel allows a receiver to associate a physical vehicle at a known location with the cryptographic credentials used to sign a location claim. Our simulations indicate that even with a pessimistic visual channel, visual verification of V2V senders provides sufficient verification capabilities to support the relevant applications.

Share

COinS