Behavioral Analytics Risks Demonstrated
BRS Labs marketing relentlessly pushes 'behavioral analytics.' May 2010 events demonstrate the inherent risk in using such an approach.
On May 3rd, a video was released of a 'suspicious' person taking off his jacket/shirt and putting into his backpack. BRS labs immediately jumped on this, citing 'video surfacing of suspicious and abnormal behavior'.
The irony is that 'abnormal behavior' turned out not to be the terror suspect, simply a guy who was warm who stopped to remove an outer layer of clothing.
On May 7th, more 'suspicious' behavior was reported - twice in a single day. Both shut down a major metropolitan center - neither turned out to be incidents.
Despite this, BRS Labs issued a press release the same week, declaring [link no longer available], "Video Surveillance Based Behavioral Analytics Proven to be The Missing Key to Identifying Suspicious Behavior."
Ultimately, the flaw here is not any individual vendor's technology but a fundamental inability to determine what is genuinely suspicious, especially in crowded areas where lots of people are engaging in various activities.
This is the reason why rules are the predominant tool in video analytics. While rules limit where you can apply them, in specific scenarios (e.g., fencelines border with clear zones), high degree of accuracy is possible as what is 'right' or 'wrong' can be clearly measured.