Drop toolsA common hacking/pen-testing technique is to drop a box physically on the local network. On this blog, there are articles going back 10 years discussing this. In the old days, this was done with $200 "netbook" (cheap notebook computers). These days, it can be done with $50 "Raspberry Pi" computers, or even $25 consumer devices reflashed with Linux.
A "Raspberry Pi" is a $35 single board computer, for which you'll need to add about another $15 worth of stuff to get it running (power supply, flash drive, and cables). These are extremely popular hobbyist computers that are used everywhere from home servers, robotics, and hacking. They have spawned a large number of clones, like the ODROID, Orange Pi, NanoPi, and so on. With a quad-core, 1.4 GHz, single-issue processor, 2 gigs of RAM, and typically at least 8 gigs of flash, these are pretty powerful computers.
Typically what you'd do is install Kali Linux. This is a Linux "distro" that contains all the tools hackers want to use.
You then drop this box physically on the victim's network. We often called these "dropboxes" in the past, but now that there's a cloud service called "Dropbox", this becomes confusing, so I guess we can call them "drop tools". The advantage of using something like a Raspberry Pi is that it's cheap: once dropped on a victim's network, you probably won't ever get it back again.
Gaining physical access to even secure banks isn't that hard. Sure, getting to the money is tightly controlled, but other parts of the bank aren't not nearly as secure. One good trick is to pretend to be a banking inspector. At least in the United States, they'll quickly bend over an spread them if they think you are a regulator. Or, you can pretend to be maintenance worker there to fix the plumbing. All it takes is a uniform with a logo and what appears to be a valid work order. If questioned, whip out the clipboard and ask them to sign off on the work. Or, if all else fails, just walk in brazenly as if you belong.
Once inside the physical network, you need to find a place to plug something in. Ethernet and power plugs are often underneath/behind furniture, so that's not hard. You might find access to a wiring closet somewhere, as Aaron Swartz famously did. You'll usually have to connect via Ethernet, as it requires no authentication/authorization. If you could connect via WiFi, you could probably do it outside the building using directional antennas without going through all this.
Now that you've got your evil box installed, there is the question of how you remotely access it. It's almost certainly firewalled, preventing any inbound connection.
One choice is to configure it for outbound connections. When doing pentests, I configure reverse SSH command-prompts to a command-and-control server. Another alternative is to create a SSH Tor hidden service. There are a myriad of other ways you might do this. They all suffer the problem that anybody looking at the organization's outbound traffic can notice these connections.
Another alternative is to use the WiFi. This allows you to physically sit outside in the parking lot and connect to the box. This can sometimes be detected using WiFi intrusion prevention systems, though it's not hard to get around that. The downside is that it puts you in some physical jeopardy, because you have to be physically near the building. However, you can mitigate this in some cases, such as sticking a second Raspberry Pi in a nearby bar that is close enough to connection, and then use the bar's Internet connection to hop-scotch on in.
The third alternative, which appears to be the one used in the article above, is to use a 3G/4G modem. You can get such modems for another $15 to $30. You can get "data only" plans, especially through MVNOs, for around $1 to $5 a month, especially prepaid plans that require no identification. These are "low bandwidth" plans designed for IoT command-and-control where only a few megabytes are transferred per month, which is perfect for command-line access to these drop tools.
With all this, you are looking at around $75 for the hardware, software, and 3G/4G plan for a year to remotely connect to a box on the target network.
As an alternative, you might instead use a cheap consumer router reflashed with the OpenWRT Linux distro. A good example would be a Gl.INET device for $19. This a cheap Chinese manufacturer that makes cheap consumer routers designed specifically for us hackers who want to do creative things with them.
The benefit of such devices is that they look like the sorts of consumer devices that one might find on a local network. Raspberry Pi devices stand out as something suspicious, should they ever be discovered, but a reflashed consumer device looks trustworthy.
The problem with these devices is that they are significantly less powerful than a Raspberry Pi. The typical processor is usually single core around 500 MHz, and the typical memory is only around 32 to 128 megabytes. Moreover, while many hacker tools come precompiled for OpenWRT, you'll end up having to build most of the tools yourself, which can be difficult and frustrating.
Hacking techniquesOnce you've got your drop tool plugged into the network, then what do you do?
One question is how noisy you want to be, and how good you think the defenders are. The classic thing to do is run a port scanner like nmap or masscan to map out the network. This is extremely noisy and even clueless companies will investigate.
This can be partly mitigated by spoofing your MAC and IP addresses. However, a properly run network will still be able to track back the addresses to the proper port switch. Therefore, you might want to play with a bunch of layer 2 things. For example, passively watch for devices that get turned off a night, then spoof their MAC address during your night time scans, so that when they come back in the morning, they'll trace it back to the wrong device causing the problem.
An easier thing is to passively watch what's going on. In purely passive mode, they really can't detect that you exist at all on the network, other than the fact that the switch port reports something connected. By passively looking at ARP packets, you can get a list of all the devices on your local segment. By passively looking at Windows broadcasts, you can map out large parts of what's going on with Windows. You can also find MacBooks, NAT routers, SIP phones, and so on.
This allows you to then target individual machines rather than causing a lot of noise on the network, and therefore go undetected.
If you've got a target machine, the typical procedure is to port scan it with nmap, find the versions of software running that may have known vulnerabilities, then use metasploit to exploit those vulnerabilities. If it's a web server, then you might use something like burpsuite in order to find things like SQL injection. If it's a Windows desktop/server, then you'll start by looking for unauthenticated file shares, man-in-the-middle connections, or exploit it with something like EternalBlue.
The sorts of things you can do is endless, just read any guide on how to use Kali Linux, and follow those examples.
Note that your command-line connection may be a low-bandwidth 3G/4G connection, but when it's time to exfiltrate data, you'll probably use the corporate Internet connection to transfer gigabytes of data.
USB hacking toolsThe above paper described not only drop tools attached to the network, but also tools attached view USB. This is a wholly separate form of hacking.
According to the description, the hackers used BashBunny, a $100 USB device. It's a computer than can emulate things like a keyboard.
However, a cheaper alternative is the Raspberry Pi Zero W for $15, with Kali Linux installed, especially a Kali derivative like this one that has USB attack tools built in and configured.
One set of attacks is through a virtual keyboard and mouse. It can keep causing mouse/keyboard activity invisibly in the background to avoid the automatic lockout, then presumably at night, run commands that will download and run evil scripts. A good example is the "fileless PowerShell" scripts mentioned in the article above.
This may be combined with emulation of a flash drive. In the old days, hostile flash drives could directly infect a Windows computer once plugged in. These days, that won't happen without interaction by the user -- interaction using a keyboard/mouse, which the device can also emulate.
Another set of attacks is pretending to be a USB Ethernet connection. This allows network attacks, such as those mentioned above, to travel across the USB port, without being detectable on the real network. It also allows additional tricks. For example, it can configure itself to be the default route for Internet (rather than local) access, redirecting all web access to a hostile device on the Internet. In other words, the device will usually be limited in that it doesn't itself have access to the Internet, but it can confuse the network configuration of the Windows device to cause other bad effects.
Another creative use is to emulate a serial port. This works for a lot of consumer devices and things running Linux. This will get you a shell directly on the device, or a login that accepts a default or well-known backdoor password. This is a widespread vulnerability because it's so unexpected.
In theory, any USB device could be emulated. Today's Windows, Linux, and macOS machines have a lot of device drivers that are full of vulnerabilities that an be exploited. However, I don't see any easy to use hacking toolkits that'll make this easy for you, so this is still mostly just theoretical.
The purpose of this blogpost isn't "how to hack" by "how to defend". Understanding what attackers do is the first step in understanding how to stop them.
Companies need to understand the hardware on their network. They should be able to list all the hardware devices on all their switches and have a running log of any new device that connects. They need to be able to quickly find the physical location of any device, with well-documented cables and tracking which MAC address belongs to which switch port. Better yet, 802.11x should be used to require authentication on Ethernet just like you require authentication on WiFi.
The same should be done for USB. Whenever a new USB device is plugged into Windows, that should be logged somewhere. I would suggest policies banning USB devices, but they are so useful this can become very costly to do right.
Companies should have enough monitoring that they can be notified whenever somebody runs a scanner like nmap. Better yet, they should have honeypot devices and services spread throughout their network that will notify them if somebody is already inside their network.
Hacking a target like a bank consists of three main phrases: getting in from the outside, moving around inside the network to get to the juice bits, then stealing money/data (or causing harm). That first stage is usually the hardest, and can be bypassed with physical access, dropping some sort of computer on the network. A $50 device like a Raspberry Pi running Kali Linux is perfect for this.
Every security professional should have experience with this. Whether it's actually a Raspberry Pi or just a VM on a laptop running Kali, security professionals should have experience with this. They should run nmap on their network, they should run burpsuite on their intranet websites, and so on. Of course, this should only be done with knowledge and permission from their bosses, and ideally, boss's bosses.