A security startup is preparing to emerge from stealth mode with a new technology that detects in real-time whether an online user or member of a social network is legitimate -- and not a bot, automated tool, or criminal performing financial or other online fraud.

Pramana, which will officially launch in July, has developed what it calls HumanPresent, a technology spun off from research at Georgia Tech that catches online fraud in action, real-time, using a dynamic method of identifying human behavior anomalies while at the same time preventing the fraudsters from detecting that they're being watched. "We are looking at real-time behavior, but we never disclose our schemes or strategies...We are observing inputs and collecting data, and we are never fixed on one [approach]. We deploy different collection mechanism strategies on different pages to evade detection, as well as evolve our system with new strategies," says Sanjay Sehgal, CEO of Pramana, who is keeping the details of the inner workings of HumanPresent close to the vest so as not to tip off the bad guys.

"Those APIs [operate] in a dynamic way, so we keep changing the back end so scammers never know what we are doing," he says.

HumanPresent doesn't rely on blacklists of botnet IP addresses, nor does it address bot activity at the network level like other organizations that sell antibotnet products and services do, according to Sehgal. "We are in the abuse and fraud detection and prevention part of the security space, not network security," he says.

In addition, Pramana's technology doesn't use device fingerprinting to identify a bot or rogue activity like other online fraud firms, such as Iovation and 41st Parameter. It uses special APIs placed on customers' Web pages that then communicate back to the HumanPresent server. The tool monitors and validates an entire user session during a transaction, including attempts to fake the CAPTCHA process. HumanPresent alerts the online retailer, financial institution, or social network operator, for example, within 10 milliseconds of finding bot or other online fraud activity.

"We give [the customer] the flexibility on the back end -- we can help take the [fraudster] to a sandbox or honeypot" or block them, Sehgal says. "I give them a score and work with them to manage their traffic."

For an ISP, for example, HumanPresent could detect which of its subscribers' machines are bot-infected. "In a social network, I can tell them one of their users had their credentials stolen, their machine compromised, and that there is a lot of automated activity from that machine," he says.

Pramana offers both a Linux-based virtual appliance that handles the fraud detection on-site or a software-as-a-service model. Among Pramana's customers so far are financial services firms, social networking sites, online gaming sites, and Webmail sites. HumanPresent could also be used to prevent blog spam, Sehgal says.

Pricing for the HumanPresent SaaS ranges from 50 cents to $2 per user or per transaction; the company has not yet determined pricing for its appliance, which is based on a hardened version of Linux.

Among the investors in Pramana, which is based on Alpharetta, Ga., are Imlay Investments; Tom Noonan, former CEO of Internet Security Systems; Mitch Free, CEO of MFG.com; and Rich DeMillo, former CTO of HP & Dean.

Have a comment on this story? Please click "Discuss" below. If you'd like to contact Dark Reading's editors directly, send us a message.