landhb/DrawBridge - Githubissues

A layer 4 Single Packet Authentication (SPA) Module, used to conceal TCP/UDP ports on public facing machines and add an extra layer of security.

Note: DrawBridge now supports both IPv4 and IPv6 traffic

Demo

gif

Please read the corresponding article for a more in-depth look at the design.

Basic usage

sudo db auth --server [REMOTE_SERVER] --dport 53 -p udp --unlock [PORT_TO_UNLOCK]

To give the db binary CAP_NET_RAW privs so that you don't need sudo to run it:

chmod 500 ~/.cargo/bin/db
sudo setcap cap_net_raw=pe ~/.cargo/bin/db

It's also convenient to create a bash alias to run db automatically when you want to access the port that it's guarding.

alias "connect"="db auth -s [REMOTE] -d 53 -p udp --unlock [PORT] && ssh -p [PORT] user@[REMOTE]"

CI/CD & Supported Kernel Versions

Kernel Version	Build	Insmod	Tests
6.0.6
5.17.2
5.15.33
5.10.110
5.8.9
5.4.188
4.19.237
4.14.275
4.9.309
4.4.302

Build and Install the Drawbridge Utilities

The usermode tools are now written in Rust! Build and install them with cargo:

git clone https://github.com/landhb/Drawbridge
cargo install --path Drawbridge/tools

# or 
cargo install dbtools

Build and Install the Drawbridge Module

To automagically generate keys, run the following on your client machine:

db keygen

The output of the keygen utility will be three files: ~/.drawbridge/db_rsa, ~/.drawbridge/db_rsa.pub and key.h. Keep db_rsa safe, it's your private key. key.h is the public key formated as a C-header file. It will be compiled into the kernel module.

To compile the kernel module simply, bring key.h, cd into the kernel module directory and run make.

# on the server compile the module and load it
# pass the ports you want to monitor as an argument
mv key.h module/include/
cd module/
make
sudo modprobe x_tables
sudo insmod drawbridge.ko ports=22,445

You may need to install your kernel headers to compile the module, you can do so with:

sudo apt-get install linux-headers-$(uname -r)
sudo apt-get update && sudo apt-get upgrade

This code has been tested on Linux Kernels between 4.X and 5.9. I don't plan to support anything earlier than 4.X but let me know if you encounter some portabilitity issues on newer kernels.

Customizing a Unique 'knock' Packet

If you wish to customize your knock a little more you can edit the TCP header options in client/bridge.c. For instance, maybe you want to make your knock packet have the PSH,RST,and ACK flags set and a window size of 3104. Turn those on:

// Flags
(*pkt)->tcp_h.fin = 0;   // 1
(*pkt)->tcp_h.syn = 0;   // 2
(*pkt)->tcp_h.rst = 1;   // 4
(*pkt)->tcp_h.psh = 1;   // 8
(*pkt)->tcp_h.ack = 1;   // 16
(*pkt)->tcp_h.urg = 0;   // 32

(*pkt)->tcp_h.window = htons(3104);

Then make sure you can create a BPF filter to match that specific packet. For the above we would have RST(4) + PSH(8) + ACK(16) = 28 and the offset for the window field in the TCP header is 14:

"tcp[tcpflags] == 28 and tcp[14:2] = 3104"

Here is a good short article on tcp flags if you're unfamiliar.. Because tcpdump doesn't support tcp offset shortcuts for IPv6 you have to work with offsets relative to the IPv6 header to support it:

(tcp[tcpflags] == 28 and tcp[14:2] = 3104) or (ip6[40+13] == 28 and ip6[(40+14):2] = 3104)"

After you have a working BPF filter, you need to compile it and include the filter in the kernel module server-side. So to compile this and place the output in kernel/listen.c in struct sock_filter code[]:

tcpdump "(tcp[tcpflags] == 28 and tcp[14:2] = 3104) or (ip6[40+13] == 28 and ip6[(40+14):2] = 3104)" -dd

which gives us:

struct sock_filter code[] = {
    { 0x28, 0, 0, 0x0000000c },
    { 0x15, 0, 9, 0x00000800 },
    { 0x30, 0, 0, 0x00000017 },
    { 0x15, 0, 13, 0x00000006 },
    { 0x28, 0, 0, 0x00000014 },
    { 0x45, 11, 0, 0x00001fff },
    { 0xb1, 0, 0, 0x0000000e },
    { 0x50, 0, 0, 0x0000001b },
    { 0x15, 0, 8, 0x0000001c },
    { 0x48, 0, 0, 0x0000001c },
    { 0x15, 5, 6, 0x00000c20 },
    { 0x15, 0, 5, 0x000086dd },
    { 0x30, 0, 0, 0x00000043 },
    { 0x15, 0, 3, 0x0000001c },
    { 0x28, 0, 0, 0x00000044 },
    { 0x15, 0, 1, 0x00000c20 },
    { 0x6, 0, 0, 0x00040000 },
    { 0x6, 0, 0, 0x00000000 },
};

And there you go! You have a unique packet that the DrawBridge kernel module will parse!

Generating an RSA Key Pair Manually

First generate the key pair:

openssl genrsa -des3 -out private.pem 2048

Export the public key to a seperate file:

openssl rsa -in private.pem -outform DER -pubout -out public.der

If you take a look at the format, you'll see that this doesn't exactly match the kernel struct representation of a public key, so we'll need to extract the relevant data from the BIT_STRING field in the DER format:

vagrant@ubuntu-xenial:~$ openssl asn1parse  -in public.der -inform DER

0:d=0  hl=4 l= 290 cons: SEQUENCE
4:d=1  hl=2 l=  13 cons: SEQUENCE
6:d=2  hl=2 l=   9 prim: OBJECT            :rsaEncryption
17:d=2  hl=2 l=   0 prim: NULL
19:d=1  hl=4 l= 271 prim: BIT STRING        <-------------------- THIS IS WHAT WE NEED

You can see that the BIT_STRING is at offset 19. From here we can extract the relevant portion of the private key format to provide the kernel module:

openssl asn1parse  -in public.der -inform DER -strparse 19 -out output.der

You'll notice that this is compatible with RFC 3447 where it outlines ASN.1 syntax for an RSA public key.

0:d=0  hl=4 l= 266 cons: SEQUENCE
4:d=1  hl=4 l= 257 prim: INTEGER           :BB82865B85ED420CF36054....
265:d=1  hl=2 l=   3 prim: INTEGER           :010001

If you need to dump output.der as a C-style byte string:

hexdump -v -e '16/1 "_x%02X" "\n"' output.der | sed 's/_/\\/g; s/\\x  //g; s/.*/    "&"/'