As though Brazil's devastating 7-1 loss to Germany on Tuesday wasn't horrific enough for Brazilian World Cup soccer fans, they may have been hit by more bank fraud attacks while they were watching the match (and weeping).

This week, Trusteer, an IBM company, discovered two new variants of the Boleto Bancario banking malware. Trusteer says that "approximately one in every 900 machines in Brazil is infected with some form of Boleto malware at any given point."

Worse, is that the percentage of Brazilian bank transactions that were deemed suspicious increased dramatically during the Brazil-Germany match Tuesday, according to BioCatch, an authentication and fraud detection startup.

In comparison to a regular Tuesday evening in Brazil, the amount of online banking activity this Tuesday was basically the same, until the match began, at which point the activity dropped dramatically.

"Brazilians did a lot of work Tuesday, but they got everything done early, before the game," says Oren Kedem, BioCatch's vice president of product management.

However, the amount of suspicious transactions during that same window did not drop.

"Fraudsters are apparently not that interested in soccer or they're not from Brazil," says Kedem. "Or these could be regular people behaving erratically."

Legitimate users did actually behave somewhat differently in the first half of the match (which ended with Brazil down by five goals) and the second half of the match. According to Kedem, customers took twice as long to complete transactions during the first half than they did in the second half.

"They were more engaged in [watching] the first half," he says. "In the second half, they basically gave up."

How does BioCatch know all this? By using "passive biometrics" -- very closely watching and analyzing things that users don't even know they're doing.

BioCatch collects user behavioral data from endpoint devices' keyboards, mouses, touchscreens, accelerometers, and gyroscopes.

They capture physiological behaviors like whether the user is left-handed or right-handed, the duration of their hand tremor, the size of their finger press, their hand-eye coordination, and their muscle structure. They capture cognitive indicators like how a user scrolls through a screen -- do they click the mouse, click and drag the mouse, use the arrow keys, use page up and page down, etc. -- how they interact with certain applications, and how they move the cursor -- quick and direct, slow and circuitous, curving up, curving down.

Then the BioCatch application issues "invisible challenges." The application may speed up or slow down how fast a selection wheel moves, or nudge a cursor in one direction, or create a "force field" that requires a user to press a touchscreen more firmly, and then see how the user responds.

All of those factors are combined into a "cognitive signature," which can then be used for "passive biometric" authentication or fraud detection.

For example, when a bank in Utah had a local customer suddenly log in from the Czech Republic and conduct a transaction, the bank's initial reaction would normally be to immediately flag that as suspicious activity by an attacker. However, the BioCatch information showed that it was indeed the same user, even though they were logging in from a different place with a different device.

The software can also tell the difference between a user and a bot, without the need for a CAPTCHA.

"Robots never respond to our [invisible] challenges," says Kedem.

That's because malware generate transactions automatically, without a user interacting with a device. So, if for example, a bot transfers money from one account to another, the fraudulent transaction may show up in the bank's records but not on BioCatch's sensors.

The next test: see how erratically Brazilian online banking customers behave on Sunday if the Cup is won by Argentina, their greatest rival.