Two firms that specialize in making E-commerce safer and less costly are becoming one. EFunds Corp. is buying ClearCommerce Corp. for $19.4 million in cash.

ClearCommerce makes software and hardware used to help prevent fraud during so-called "card not present" transactions. Such transactions happen in E-commerce, of course, but also in mail and phone transactions. The privately held company''s client list includes Adobe, Apple Computer, and Staples.

Publicly traded eFunds sells outsourced services including electronic transaction processing and risk management. Its clients generally are financial institutions, electronic-funds-transfer networks, government agencies, and retailers.

The card-not-present capability was attractive to eFunds, which lacks a product that can help prevent fraud in those kinds of transactions. EFunds is anxious to link ClearCommerce''s FraudDetect database with its DebitBureau database.

The deal is expected to close in the first quarter of this year and net eFunds 140 new retail customers in seven nations.

There are other companies with expertise in preventing online fraud, but they haven''t built their firms around anti-fraud products the way eFunds has, says Rahul Gupta, senior vice president for eFund''s risk and electronic-payments group.

"We have fraud expertise in the traditional world of retail, banking, ATMs, and point-of-service devices, but they have a focused and deep expertise in online fraud," Gupta says.

ClearCommerce sells software modules that plug into a retailer''s Web site. When a would-be buyer clicks on "pay," the transaction is routed to ClearCommerce''s software, which runs a series of tests to gauge the transaction''s legitimacy.

There are more than 100 routines, as ClearCommerce calls the tests. One looks to see how many times the person has purchased something at the site and in what span of time. Rapid shopping trips are fishy. So, too, would be buying five flat-panel TVs at Apple.com, says Gupta.

"There are cheaper places to buy a flat-panel display than Apple.com," he says. "And why would someone buy five of them?" The routines run in a matter of seconds.

A neural-network application, which can learn from experience, adds up a transaction''s score. If too many red flags have been raised, the neural net holds the transaction and refers it to human analysts who can approve the transaction or reject it in a letter to the cardholder.