c0d3 :: j0rg3

A collection of tips, tricks and snips. A proud Blosxom weblog. All code. No cruft.

Thu, 13 Jul 2017

Improved Anonymity on Kali Linux

I’m not entirely certain when BackTrack/Kali began behaving more like a regular desktop distro but I seem to recall that originally, networking subsystems were down when you booted up into Run Level 3. It was up to you to turn on the interfaces and fire up a GUI if such was desired. IMO, that’s precisely how it should be. I get it. Most of us aren’t ever won’t ever find ourselves in a clandestine lot, inside of a snack and caffeine filled, non-descript, conversion van with a Yagi pointed at the bubble-window, ready to pilfer innocent datums just trying to get by in this lossy-protocoled, collision-rife, world.

Rather, very many of us just want the stinking box online so we can run through our tutorials and hack our own intentionally vulnerable VMs. A thorough taste of hacking’s un-glamorous underbelly is quite enough for many.

I’m confident that the BT fora were inundated with fledgling hackers complaining that their fresh install couldn’t find WiFi or didn’t load the desktop. However, I feel that distros dedicated to the Red Team should try to instill good habits. Having your machine boot and activate an interface announcing your presence and spewing out MAC and hostname is bad for business. Booting into a (comparatively) heavy GUI is also not where I want to begin.

Let’s imagine that we’re trying to crack into a thing. Don’t we want to apply maximal CPU resources, rather than having GUI elements bringing little beyond cost? If you notice, very many of the related tools still live on the CLI. The typical course of development (e.g.: Nmap, Metasploit) is that the CLI version is thoroughly developed before someone drops a GUI atop (respectively: Zenmap, Armitage).


So let’s take our Kali and make a few quick changes. We want to boot up in text/CLI mode and we want networking left off until we choose to make noise. Further, we want to randomize our MAC address and hostname at every boot.

We’ll use iwconfig to enumerate our wireless interfaces.
lo        no wireless extensions.

wlan1     IEEE 802.11 ESSID:"ESSID"
          Mode:Managed Frequency:2.412 GHz Access Point: 17:23:53:96:BE:67
          Bit Rate=72.2 Mb/s Tx-Power=20 dBm
          Retry short limit:7 RTS thr:off Fragment thr:off
          Encryption key:off
          Power Management:off
          Link Quality=70/70 Signal level=-21 dBm
          Rx invalid nwid:0 Rx invalid crypt:0 Rx invalid frag:0
          Tx excessive retries:253 Invalid misc:400 Missed beacon:0

eth0      no wireless extensions.

wlan0     IEEE 802.11 ESSID:off/any
          Mode:Managed Access Point: Not-Associated Tx-Power=0 dBm
          Retry short limit:7 RTS thr:off Fragment thr:off
          Encryption key:off
          Power Management:on

We have two wireless interfaces: wlan0, wlan1

Okay, first let’s configure to start up in text mode:
> systemctl set-default multi-user.target
Created symlink /etc/systemd/system/default.target → /lib/systemd/system/multi-user.target.

Traditionally from text mode, we bring up the GUI desktop with the command startx. Since we don’t yet have that command, let’s create it:
> echo "systemctl start gdm3.service" > /usr/sbin/startx && chmod +x /usr/sbin/startx

Disable network-manager autostart:
> systemctl disable network-manager.service
> sed -i 's/5min/30sec/' /etc/systemd/system/network-online.target.wants/networking.service

Now, let’s randomize our hostname and MAC addresses at every boot by adding some cronjobs:
> crontab -e

We’ll add two jobs to randomize MAC address and one for our host name:
@reboot macchanger -r wlan0
@reboot macchanger -r wlan1
@reboot hostname `strings /dev/urandom | grep -o '[[:alnum:]]' | head -n 30 | tr -d '\n'`

We ‘re good! We’ve improved efficiency by staving off the GUI for when we genuinely want it and improved anonymity by randomizing some common ways of identifying the rig.


Tags: , , , , , , , ,
Permalink: 2017-07-10.improved.anonymity.on.kali.linux

Fri, 17 Feb 2017

The making of a Docker: Part I - Bitmessage GUI with SSH X forwarding

Lately, I’ve been doing a lot of work from a laptop running Kali. Engaged in pursuit of a new job, I’m brushing up on some old tools and skills, exploring some bits that have changed.

My primary desktop rig is currently running Arch because I love the fine grain control and the aggressive releases. Over the years, I’ve Gentoo’d and Slacked, Crunchbanged, BSD’d, Solarised, et cet. And I’ve a fondness for all of them, especially the security-minded focus of OpenBSD. But, these days we’re usually on Arch or Kali. Initially, I went with Black Arch on the laptop but I felt the things and ways I was fixing things were too specific to my situation to be good material for posts.

Anyway, I wanted to get Bitmessage running, corresponding to another post I have in drafts. On Kali, it wasn’t going well so I put it on the Arch box and just ran it over the network. A reasonable solution if you’re in my house but also the sort of solution that will keep a hacker up at night.

If you’re lucky, there’s someone maintaining a package for the piece of software that you want to run. However, that’s often not the case.

If I correctly recall, to “fix” the problem with Bitmessage on Kali would’ve required the manual installation an older version of libraries that were already present. Those libraries should, in fact, be all ebony and ivory, living together in harmony. However, I just didn’t love the idea of that solution. I wanted to find an approach that would be useful on a broader scale.

Enter containerization/virtualization!

Wanting the lightest solution, I quickly went to Docker and realized something. I have not before built a Docker container for a GUI application. And Bitmessage’s CLI/daemon mode doesn’t provide the fluid UX that I wanted. Well, the easy way to get a GUI out of a Docker container is to forward DISPLAY as an evironment variable (i.e., docker run -e DISPLAY=$DISPLAY). Splendid!

Except that it doesn’t work on current Kali which is using QT4. There’s a when graphical apps are run as root and though it is fixed in QT5, we are using current Kali. And that means we are, by default, uid 0 and QT4.

I saw a bunch of workarounds that seemed to have spotty (at best) rates of success including seting QT’s graphics system to Native and giving Xorg over to root. They, mostly, seemed to be cargo cult solutions.

What made the most sense to my (generally questionable) mind was to use X forwarding. Since I had already been running Bitmessage over X forwarding from my Arch box, I knew it should work just the same.

To be completely truthful, the first pass I took at this was with Vagrant mostly because it’s SO easy. Bring up your Vagrant Box and then:
vagrant ssh -- -X
Viola!

Having proof of concept, I wanted a Docker container. The reason for this is practical. Vagrant, while completely awesome, has substantially more overhead than Docker by virtualizing the kernel. We don’t want a separate kernel running for each application. Therefore Docker is the better choice for this project.

Also, we want this whole thing to be seemless. We want to run the command bitmessage and it should fire up with minimal awkwardness and hopefully no extra steps. That is we do not want to run the Docker container then SSH into it and execute Bitmessage as individual steps. Even though that’s going to be how we begin.

The Bitmessage wiki accurately describes how to install the software so we’ll focus on the SSH setup. Though when we build the Dockerfile we will need to add SSH to the list from the wiki.

We’re going to want the container to start so that the SSH daemon is ready. Until then we can’t SSH (with X forwarding) into the container. Then we’ll want to use SSH to kick off the Bitmessage application, drawing the graphical interface using our host system’s X11.

We’re going to take advantage of Docker’s -v --volume option which allows us to specify a directory on our host system to be mounted inside our container. Using this feature, we’ll generate our SSH keys on the host and make them automatically available inside the container. We’ll tuck the keys inside the directory that Bitmessage uses for storing its configuration and data. That way Bitmessage’s configuration and stored messages can be persistent between runs — and all of your pieces are kept in a single place.

When we generate the container /etc/ssh/sshd_config is configured to allow root login without password only (i.e., using keys). So here’s how we’ll get this done:
mkdir -p ~/.config/PyBitmessage/keys #Ensure that our data directories exist
cd ~/.config/PyBitmessage/keys
ssh-keygen -b 4096 -P "" -C $"$(whoami)@$(hostname)-$(date -I)" -f docker-bitmessage-keys #Generate our SSH keys
ln -fs docker-bitmessage-keys.pub authorized_keys #for container to see pubkey

Build our container (sources available at Github and Docker) and we’ll make the script to handle Bitmessage to our preferences. #!/bin/bash
# filename: bitmessage
set -euxo pipefail

# open Docker container:
# port 8444 available, sharing local directories for SSH and Bitmessage data
# detatched, interactive, pseudo-tty (-dit)
# record container ID in $DID (Docker ID)
DID=$(docker run -p 8444:8444 -v ~/.config/PyBitmessage/:/root/.config/PyBitmessage -v ~/.config/PyBitmessage/keys/:/root/.ssh/ -dit j0rg3/bitmessage-gui bash)

# find IP address of new container, record in $DIP (Docker IP)
DIP=$(docker inspect $DID | grep IPAddress | cut -d '"' -f 4)

# pause for one second to allow container's SSHD to come online
sleep 1

# SSH into container and execute Bitmessage
ssh -oStrictHostKeyChecking=no -oUserKnownHostsFile=/dev/null -oIdentityFile=~/.config/PyBitmessage/keys/docker-bitmessage-keys -X $DIP ./PyBitmessage/src/bitmessagemain.py

# close container if Bitmessage is closed
docker kill $DID

Okay, let’s make it executable: chmod +x bitmessage

Put a link to it where it can be picked up system-wide: ln -fs ~/docker-bitmessage/bitmessage /usr/local/bin/bitmessage

There we have it! We now have a functional Bitmessage inside a Docker container. \o/

In a future post we’ll look at using eCryptfs to further protect our Bitmessage data stores.

  Project files: Github and Docker


Tags: , , , , , , , , , , ,
Permalink: 20170217.making.a.docker.bitmessage

Mon, 02 Jan 2017

Securing a new server

Happy new year! New year means new servers, right?

That provides its own set of interesting circumstances!

The server we’re investigating in this scenario was chosen for being a dedicated box in a country that has quite tight privacy laws. And it was a great deal offered on LEB.

So herein is the fascinating bit. The rig took a few days for the provider to set up and, upon completion, the password for SSHing into the root account was emailed out. (o_0)

In very security-minded considerations, that means that there was a window of opportunity for bad guys to work on guessing the password before its owner even tuned in. That window remains open until the server is better secured. Luckily, there was a nice interface for reinstalling the OS permitting its purchaser to select a password.

My preferred approach was to script the basic lock-down so that we can reinstall the base OS and immediately start closing gaps.


In order:

  • Set up SSH keys (scripted)
  • Disable password usage for root (scripted)
  • Install and configure IPset (scripted. details in next post)
  • Install and configure fail2ban
  • Install and configure PortSentry

  • In this post, we’re focused on the first two steps.


    The tasks to be handled are:

  • Generate keys
  • Configure local SSH to use key
  • Transmit key to target server
  • Disable usage of password for ‘root’ account

  • We’ll use ssh-keygen to generate a key — and stick with RSA for ease. If you’d prefer ECC then you’re probably reading the wrong blog but feel encouraged to contact me privately.

    The code:

    #!/bin/bash
    #configure variables
    remote_host="myserver.com"
    remote_user="j0rg3"
    remote_pass="thisisaratheraquitecomplicatedpasswordbatterystaple" # https://xkcd.com/936/
    local_user=`whoami`
    local_host=`hostname`
    local_date=`date -I`
    local_filename=~/.ssh/id_rsa@$remote_host

    #generate key without passphrase
    ssh-keygen -b 4096 -P "" -C $local_user@local_host-$local_date -f $local_filename

    #add reference to generated key to local configuration
    printf '%s\n' "Host $remote_host" "IdentityFile $local_filename" >> ~/.ssh/config

    #copy key to remote host
    sshpass -p $remote_pass ssh-copy-id $remote_user@$remote_host

    #disable password for root on remote
    ssh $remote_user@$remote_host "cp /etc/ssh/sshd_config /etc/ssh/sshd_config.bak && sed -i '0,/RE/s/PermitRootLogin yes/PermitRootLogin no/' /etc/ssh/sshd_config"

    We just run this script soon as the OS is reinstalled and we’re substantially safer. As a Deb8 install, quickly pulling down fail2ban and PortSentry makes things quite a lot tighter.

    In another post, we’ll visit the 2017 version of making a DIY script to batten the hatches using a variety of publicly provided blocklists.

    Download here:
        ssh_quick_fix.sh


    Tags: , , , ,
    Permalink: 20170102.securing.a.new.server