Behavioral detectionLocal entrepreneur starts a new software venture
Behavioral Recognition Systems was founded by Ray Davis, a Houston-based serial entrepreneur who raised $47 million in funding for a start-up which develops behavioral-detection software; the software does not have to be “taught” what is normal and abnormal behavior; the company claims that the software learns what is normal within a few weeks, so there is no need to program rule
Where there is a security need, there is a business opportunity. This insight has not escaped Houston-based Ray Davis, who started several start-ups during his business career. He founded Cynet, which sold fax broadcasting technology and services, and later SimDesk, which hawked virtual office software.
Both companies shuttered after his departure, but Houston Chronicle’s By Purva Patel writes that now the serial entrepreneur is trying his hand at a new venture — one whose claims are raising eyebrows. He says the startup, Behavioral Recognition Systems Labs, can potentially change the security industry with its software, AISight. The software analyzes security video and learns what is normal behavior, from hotel lobbies to ports, he said.
Most security software used today alerts users only to movements and breaches they are programmed to notice. For example, they may send alerts when a door opens. The trouble is that software that can not tell if a regular delivery person or a burglar opened the door would send a false alarm.
Patel writes that AISight learns what is normal within a few weeks, so there is no need to program rules, Davis said. “We don’t have to teach it,” he said. “The computer is able to understand the actions it sees on camera and understand what is a threat or anomaly and report that to security personnel.”
It is a claim no security company has made before. “Most people will say their stuff performs better or has better optics or solves a specific problem, but most other guys don’t say they’ve done something completely different,” said Samuel Pfeifle, editor of Security Systems News. “It would be like a camera manufacturer saying they can see through walls.” The video analytics industry has long suffered a reputation of overpromising and underdelivering, so skepticism is not unusual. “But what they’re saying they’re doing is very difficult because what is normal and what would you get alerts on?” he said.
Youngchoon Park, head of security systems research at Milwaukee-based Johnson Controls, tested an early version of the software. The software sent out up to 4 percent fewer false alerts than other systems, he said. “That’s not really significant. It’s not a jawbreaker,” he said, adding that he’s not a fan of video analytics because false alerts cost clients too much money to investigate. He also worries AISight could overlearn, meaning it could eventually assume everything it sees is normal and send no
