Monday, October 12th, 2009

What’s So Ubiquitous About ONS?

There may be life on March, and there just may be people out there wondering why I wrote a post about ONS under the “ubiquitous” theme. What’s ubiquitous about an Object Name Service?

(Note in passing: If my use of ubiquitous seems suspect, please download the “U-Essence” report from this website.)

The reason is wireless sensor networks, also known as Ubiquitous Sensor Networks (USN) in Korea and Japan.

Wireless sensors are regarded as an extension of RFID technology even if they differ in several respects. Simply put, a sensor is active, transmitting a varying measurement value and depends on continuous power supply. A plain RFID tag is passive because it relies on radio frequencey energy beamed to it by the tag reader. It transmits the same identity every time.

The ONS (Object Name Service) was invented for passive RFID tags. By extrapolation the ONS is also found in some sensor network architectures. My previous post raised objections about the ONS for passive RFID tags, mainly based on security considerations. For sensor networks the ONS functionality is plain irrelevant in many cases. Let me sketch an alternative.

Why would anybody make sensors directly accessible from the Internet? The scenario seems extremely unlikely. Let’s find out why.

The price point is a major factor for passive RFID tags. Sensors will always be many times more costly than passive tags. A sensor network needs continuous supervision and calibration to be useful for non-trivial purposes. The initial investment and subsequent operation is far from free. A sensor network is a precious asset.

Any device setting up itself for Internet access must be carefully protected. It will be subject to countless hacking and other attacks. What’s more, raw sensor output is seldom the right thing to deliver to a consumer. Measurements must often be converted and sanitized to be useful to a consumer.

A producer may run a geographically scattered sensor network. Rather than making sensors directly accessible the natural solution is for the producer to run the network like a VPN (Virtual Private Network) over the Internet. Quality measurement data is provided to consumers over an ordinary web service.

The ubiquitous home (u-home) is an important sensor market. Running a sensor network is way beyond most home owners. The natural solution is for the home sensors to connect to a concentrator device in the home that also acts as a firewall. It connects to a ubiquitous service provider over the Internet using an encrypted protocol. The service provider displays measurements in a meaningful way in their web site. Home owners would have to log in to the web site to check the status of their home. The ubiquitous service provider would supervise many home networks.

This architecture is an example of cloud computing.

Assume you interrogate a device on the Internet that says the air temperature in the Pusan harbor is 15 degrees. Unless you know the organization setting up the sensor there is no way to assess the quality of this measurement. In fact, you cannot even tell if the remote device is a random number generator located in Romania. A trust relationship to the producer is necessary.

In summary there is very little space left for ONS in sensor networks. ONS is designed for RFID tags distributed ad hoc to unknown recipients. Why should sensor networks be hooked up to a fragile ONS? The Internet already has reliable mechanisms for authentication and security in the digital world. Such mechanisms are necessary to protect the relationship of trust between producers and consumers of sensor data.

Comments are closed.