Abstract:

Data mining is the process of posing queries and extracting patterns, often previously unknown from large quantities of data using pattern matching or other reasoning techniques. Data mining has many applications for national security, also referred to as homeland security. The threats to national security include attacking buildings, destroying critical infrastructures such as power grids and telecommunication systems. Data mining techniques are being investigated to find out who the suspicious people are and who is capable of carrying out terrorist activities. One particular security application that can benefit from data mining is surveillance. We need to build infrastructures to conduct surveillance so that we can determine who might be suspicious. However, we also need to protect the privacy of the individuals who are law abiding citizens.

This presentation will first discuss data mining for surveillance applications. We will survey various surveillance applications and discuss the developments on applying data mining. Suspicious event detection is an area that has been investigated in some detail. The idea here is to represent various events, some of which are suspicious such as entering a secure room. Later when that event occurs the system will flag this event as suspicious. One of the challenges here is to combine suspicious event detection with say facial recognition techniques to determine who the suspicious people are in addition to detecting the suspicious events. Another challenge is to conduct distributed surveillance where there are multiple video feeds and the system has to monitor and combine events which may be suspicious. The system should also be able to detect the movements of people as they travel from one place to another. Link analysis techniques could be utilized to follow such movements and determine the links that are suspicious. A third challenge is to associate people with unidentified luggage or bags. When the system detects an unaccompanied bag, it should then be able to carry out a trace back and determine who has left the bag. Finally a person by him or herself may not be suspicious, but seen together in a group he/she may be. That is, the system has to identify groups of suspicious individuals. Other challenges include conducting on-line analysis of surveillance data where the system should have the capability to analyze the surveillance data in real-time, make decisions and take appropriate actions.

The critical need for applying data mining for surveillance poses serious privacy threats. The challenge here is to carry out privacy preserving surveillance. There are some efforts on blanking the face of a person so that his/her privacy is maintained. However by doing this, the suspicious people’s identity is also not revealed. Some efforts have focused on individuals carrying tags so that the faces of those with the appropriate tags are not revealed. This approach has a problem as the suspicious person can steal tags from others. A solution we are investigating is to encrypt all the faces of people with some keys. Only trusted agents have the keys for decryption. If the surveillance data shows that an individual is carrying out suspicious activities, then the trusted agents can reveal the identity of these suspicious people.

In addition to mining surveillance data, data mining can also be applied for geospatial applications. For example, one could combine web services provided by Google Maps or Map quest and connect the maps with say “friend of a friend” ontologies and determine the locations of various individuals. Suspicious people can use this information to terrorize the individuals whose locations have been revealed. Geospatial data can be mined to detect changes as well as detect unusual objects. The presentation will also discuss mining geospatial data.

In summary, the presentation will provide an overview mining surveillance data as well as conducting privacy preserving surveillance. Applying data mining to geospatial data such as maps will also be discussed.

Biography: