Previous Page  61 / 88 Next Page
Information
Show Menu
Previous Page 61 / 88 Next Page
Page Background

54

Bloomberg Businessweek

October 8, 2018

found inmade-to-order server setups at banks, hedge funds,

cloud computing providers, and web-hosting services, among

other places. Supermicro has assembly facilities in California,

the Netherlands, and Taiwan, but its motherboards—its core

product—are nearly all manufactured by contractors in China.

The company’s pitch to customers hinges on unmatched

customization, made possible by hundreds of full-time engi-

neers and a catalog encompassing more than 600 designs. The

majority of its workforce in San Jose is Taiwanese or Chinese,

and Mandarin is the preferred language, with

hanzi

illing the

whiteboards, according to six former employees. Chinese pas-

tries are delivered every week, and many routine calls are done

twice, once for English-only workers and again in Mandarin.

The latter are more productive, according to people who’ve

been on both. These overseas ties, especially the widespread

use of Mandarin, would have made it easier for China to gain

an understanding of Supermicro’s operations and potentially

to iniltrate the company. (A U.S. oicial says the government’s

probe is still examining whether spies were planted inside

Supermicro or other American companies to aid the attack.)

With more than 900 customers in 100 countries by 2015,

Supermicro ofered inroads to a bountiful collection of sen-

sitive targets. “Think of Supermicro as the Microsoft of the

hardware world,” says a former U.S. intelligence oicial

who’s studied Supermicro and its business model. “Attacking

Supermicro motherboards is like attacking Windows. It’s like

attacking the whole world.”

W

ell before evidence of the attack surfaced inside the

networks of U.S. companies, American intelligence

sources were reporting that China’s spies had plans

to introduce malicious microchips into the supply chain. The

sources weren’t speciic, according to a person familiar with

the information they provided, and millions of motherboards

are shipped into the U.S. annually. But in the irst half of 2014,

a diferent person briefed on high-level discussions says, intel-

ligence oicials went to the White House with something more

concrete: China’s military was preparing to insert the chips into

Supermicro motherboards bound for U.S. companies.

The speciicity of the information was remarkable, but

so were the challenges it posed. Issuing a broad warning to

Supermicro’s customers could have crippled the company, a

major American hardware maker, and it wasn’t clear from the

intelligence whom the operation was targeting or what its ulti-

mate aims were. Plus, without conirmation that anyone had

been attacked, the FBI was limited in how it could respond.

The White House requested periodic updates as information

came in, the person familiar with the discussions says.

Apple made its discovery of suspicious chips inside

Supermicro servers around May 2015, after detecting odd net-

work activity and irmware problems, according to a person

familiar with the timeline. Two of the senior Apple insiders

say the company reported the incident to the FBI but kept

details about what it had detected tightly held, even internally.

Government investigators were still chasing clues on their own

when Amazon made its discovery and gave them access to sab-

otaged hardware, according to one U.S. oicial. This created

an invaluable opportunity for intelligence agencies and the

FBI—by then running a full investigation led by its cyber- and

counterintelligence teams—to see what the chips looked like

and how they worked.

The chips on Elemental servers were designed to be as

inconspicuous as possible, according to one person who saw

a detailed report prepared for Amazon by its third-party secu-

rity contractor, as well as a second person who saw digital

photos and X-ray images of the chips incorporated into a later

report prepared by Amazon’s security team. Gray or of-white

in color, they looked more like signal conditioning couplers,

another common motherboard component, than microchips,

and so they were unlikely to be detectable without specialized

equipment. Depending on the board model, the chips varied

slightly in size, suggesting that the attackers had supplied dif-

ferent factories with diferent batches.

Oicials familiar with the investigation say the primary role

of implants such as these is to open doors that other attackers

can go through. “Hardware attacks are about access,” as one

former senior oicial puts it. In simpliied terms, the implants

on Supermicro hardware manipulated the core operating

instructions that tell the server what to do as data move across

a motherboard, two people familiar with the chips’ operation

say. This happened at a crucial moment, as small bits of the

operating system were being stored in the board’s temporary

memory en route to the server’s central processor, the CPU.

The implant was placed on the board in a way that allowed it to

efectively edit this information queue, injecting its own code

or altering the order of the instructions the CPU was meant to

follow. Deviously small changes could create disastrous efects.

Since the implants were small, the amount of code they

contained was small as well. But they were capable of doing

two very important things: telling the device to communicate

with one of several anonymous computers elsewhere on the

internet that were loaded with more complex code; and pre-

paring the device’s operating system to accept this new code.

The illicit chips could do all this because they were connected

to the baseboard management controller, a kind of superchip

that administrators use to remotely log in to problematic serv-

ers, giving them access to the most sensitive code even on

machines that have crashed or are turned of.

This system could let the attackers alter how the device

functioned, line by line, however they wanted, leaving no one

the wiser. To understand the power that would give them, take

this hypothetical example: Somewhere in the Linux operating

system, which runs in many servers, is code that authorizes a

user by verifying a typed password against a stored encrypted

one. An implanted chip can alter part of that code so the server

won’t check for a password—and presto! A secure machine is

open to any and all users. A chip can also steal encryption keys

for secure communications, block security updates that would

neutralize the attack, and open up new pathways to the inter-

net. Should some anomaly be noticed, it would likely be