Researcher: Adrian Perrig
Research Area: Next Generation Secure and Available Networks
Sensors and sensor networks technology has seen tremendous advancements in the past years. Much ongoing research and development has been done on the core functions that enable sensor network deployments. These functions include adaptive routing algorithms, energy-efficient transports, secure broadcast protocols, specialized operating systems and even tiny databases targeted at the tiny devices. There has been little work done on providing monitoring and management support for sensor networks. Unlike tradition computer equipment, sensors are usually place in public and unguarded locations, making it easy for an attacker to tamper and compromise some or all of the sensors. It is thus essential to not only design secure sensor networks, but also to provide continuous monitoring of the network health for early signs of anomalous or malicious activities. This research proposed to architect a security monitoring and management system for sensor networks. The system will include micro-level diagnostics and macro-level analytic tools.
Micro-Level Diagnostic Tools
These tools collect fine-grained, short-term network information such as loss, latency and bandwidth, as well as sensor information, such as memory integrity, resource and aliveness. Analogous tools in the Internet, such as ping, tcpdump and traceroute, have also been used heavily as fundamental mechanisms to investigate attacks. However, the Internet tools are not applicable to sensor networks because of the drastically different constraints and requirements of the two environments. This part of our project will involve designing novel network measurement algorithms that are efficient in terms of bandwidth, computation and energy consumption. These algorithms should also be scalable to network size by amortizing data collection costs across the sensors.
Macro-level Analytics Tools
These tools leverage the results from the micro-level tools to detect anomalous or malicious activities. Instead of relying on a single type of information sources, we plan to correlate and analyze the various information collected to better pinpoint and understand incidents. The high volatility nature of sensor networks means unavoidable fluctuations in the surrounding environment, which can mistakenly be interpreted as attacks. For example, disappearance and subsequent reappearance of a region of sensors on the radar screen can mean either a temporary network partition because of high-loss, or that the sensors were taken off-line to be compromised by an attacker. The analytic tools will help in discerning the cause of an incident.
To design practical tools, we need to understand the common operational problems and best practices facing sensor network deployment. The proposed testbed will be the initial environment for our monitoring and measuring system, providing a realistic setting to test new security primitives and attacks on sensor networks. We will also collect and archive network and sensor data from the testbed and make the data available to the research community.