A North Korean hacking group has tried to steal at least $1.1 billion in a series of attacks on global banks over the past four years, according to cybersecurity firm FireEye Inc.

The hackers, which FireEye identified as APT38, have infiltrated more than 16 organizations in 11 countries including the U.S., and stolen more than $100 million. The group has hacked heavily defended servers at banks and spent time scouring their networks. Security officials should be alarmed, FireEye said last week in a report.

“They understand banking networks pretty well,” Charles Carmakal, vice president of consulting at FireEye, said in an interview. “And they probably have geopolitical considerations behind the timing, location of their attacks.”

The most prominent attack by APT38 was the theft of funds from the Bangladeshi central bank’s accounts at the U.S. Federal Reserve in 2016. In that case, the hackers got the Fed to transfer some $100 million by sending fake wiring orders. About $40 million was recovered when the hack was discovered and transfers reversed before they could be withdrawn.

In January, Mexico’s state-owned trade bank thwarted the attempted theft of $110 million using similar methods. In May, a Chilean bank lost $10 million. Both attacks were carried out by APT38, FireEye said in the report.

North Korean diplomats and official media have denied that the country plays any role in cyberattacks.

In recent attacks, the hacking group “burns the house down,” wiping out computer hard drives to erase its tracks, Carmakal said. While the attacks continue, APT38 hasn’t targeted American banks amid North Korea’s peace talks with the U.S., he said.

Banks and other financial institutions are targeted by the most sophisticated cyber criminals, who are attracted to the lure of big-money paydays, FireEye and other groups have said. That has prompted banks to outspend other industries to protect themselves, with the biggest U.S. firms’ annual cybersecurity budgets reaching $1 billion.

Financial firms face the highest number of attempted breaches from computer addresses that have already been blocked because of misbehavior, according to a report set for release Tuesday by cybersecurity firm eSentire. That points to targeted campaigns and persistent efforts by sophisticated attackers, according to Eldon Sprickerhoff, founder and chief innovation officer of eSentire.