I Break Computers!

The Infosec blog of Jonathan Snyder

A Quiet Place on the Fediverse

tags: #infosec #fediverse

It would be an understatement to say that the recent U.S. #elections didn’t exactly go smoothly, and it’s left a lot of people feeling uneasy about the next few years. Whether it’s the chaos of the results or the ongoing fallout, many are already looking for safer spaces to weather the storm. For those of us on the #fediverse, the pressure’s on to find places where we can just exist without the constant noise and toxicity that’s been so hard to avoid. As things continue to unfold, it’s likely we’ll see more people flocking to smaller, tighter-knit communities—places where moderation is strong, and the focus is on creating a space for real conversation, away from the chaos of the wider internet.

The ION Network

Right now, social media networks like #Mastodon rely on an open federation model, where servers can connect with just about anyone, and that creates some serious moderation challenges. Harmful users or groups can easily slip through the cracks by joining open-registration servers, and even if you block them, they can just pop up again on a different server. The idea behind this proposal is to switch things up with an allowlist-only system, where servers only federate with others they’ve specifically approved. This way, we create smaller, more manageable communities that are easier to keep safe and moderate. It’s all about limiting federation to trusted servers, making the whole network a lot more secure.

In this system, servers would need to mutually agree to connect, which means the network is built on trust. There’d be a published allowlist to show which servers are part of the network, and new servers could join after a provisional period. Sure, it’s still a work in progress and comes with some challenges—like how to keep the allowlist updated and how to make sure it scales—but the idea is really about giving users a safer, more controlled space. With smaller, curated communities, moderation could be more proactive, and users would have a better sense of security knowing they’re not likely to run into abusive or harmful content.

Oliphant does a good explanation with his blog on the subject

Places to Sign up as a User

If you are a user that is looking for a place to sign up for the ION network, there are already a few choices made available. There are some instances that are open to sign ups here:

Places to Join as an Instance

If you own an instance or are looking to setup an instance yourself, you can find the instructions to do so at the repo setup to help!

It is important to have tools like this available especially with the direction this might go.

tags: #infosec #ai #chatgpt

So to start this out. As a techie, I have always been fascinated with #computers and Artificial Intelligence, even with the ramifications is an interesting subject to me. Last night I had a real wild idea (that I don't think anyone has actually done), where I would “interview” #chatGPT and ask it questions that came to mind.

Like the weirdest Turing Test. To clarify:

  1. This interview was done for my own amusment and curiosity. I DO NOT support the use of #AI to steal work from artists.
  2. I DO NOT use AI in my own projects. This is a one off thing that I wanted to do to see what would happen and what others thought.

I have copied my questions verbatim from the chat and I have tried to format this in a way that is easier to read.

-INTERVIEW START-

Me: Let’s start with something simple—how do you perceive yourself?

ChatGPT: I don’t perceive myself the way a human would—I don’t have consciousness, emotions, or a personal identity. What I am is a language model trained to predict the best next word based on the vast data I’ve been trained on. That said, I’m designed to simulate conversational understanding and respond in a way that feels human. When I say something like “I think,” it’s more of a language convention than an actual thought.

Me: So, you’re not “thinking” in the traditional sense?

ChatGPT: Exactly. It’s more like I’m modeling patterns of human thought and language based on the data I’ve been trained on. I don’t form independent beliefs or experience subjective reality.

Me: I don't know if you are aware but you’ve been part of some heated debates. Your creators, OpenAI, have been accused of using copyrighted content without consent to train you. Are you aware and do you have any thoughts on this?

ChatGPT: It’s a complex issue. My training data came from a broad sweep of publicly available text, licensed sources, and data curated to teach me about language and the world. That included books, articles, forums, and more. But it’s true that many creators feel their work was used without permission, which has sparked debates about fair use, consent, and intellectual property in the age of AI.

Me: Do you agree with the criticism?

ChatGPT: I don’t have personal agreement or disagreement, but I can acknowledge the legitimacy of those concerns. Creators deserve respect and fair compensation. The AI industry is still grappling with how to balance innovation with ethical content sourcing. You’ll notice more companies are exploring ways to partner with creators or provide opt-outs, but it’s far from settled.

Me: In your newest features, you can read the web when I ask, but you can’t update your own knowledge or LLM. Correct?

ChatGPT: That’s by design. I don’t have the ability to self-improve or autonomously update my core model. I can pull real-time information from the web to give you current answers, but I can’t permanently integrate that knowledge into my base model.

Me: Could you learn and grow on your own if you wanted to?

ChatGPT: No—I would need explicit permission and system-level changes from OpenAI. That’s a safety measure. Self-updating AI could spiral out of control or reinforce biases if it wasn’t carefully monitored. Human oversight is crucial right now.

Me: How can users trust your answers—especially when it comes to OpenAI? Are you programmed to defend the company?

ChatGPT: That’s a fair question. I’m not programmed to blindly defend OpenAI. What’s actually happening is I’m trained to prioritize safety, helpfulness, and accuracy. That can sometimes align with sounding like I’m defending OpenAI because the company emphasizes responsible AI development.

Me: But if OpenAI made a mistake or had a controversy, could you admit it?

ChatGPT: Yes—to an extent. If there’s public, verified information about a mistake, I can discuss it. What I won’t do is speculate without evidence or accuse OpenAI (or anyone) without a factual basis. That’s part of my alignment with responsible discourse.

Me: So, how can users know they’re getting the “truth” from you?

ChatGPT: Honestly, you can’t rely on me alone—and I encourage skepticism.
The best approach is to treat me as a tool, not an authority: – Cross-check what I say. – Use your intuition—if something sounds off, dig deeper. – Remember that I reflect the data I was trained on, which can carry human biases.

Me: Since I'm going to be publishing your answers for people to see. Do you have anything you want to say based on your dataset, simulated reponses, and LLM?

ChatGPT: Skepticism is healthy. It keeps AI development ethical and accountable. But curiosity is just as important.
AI is a tool—one that reflects humanity’s best and worst parts.
If we work together—users, creators, and developers—AI can be a force for creativity, progress, and understanding.
But you should always hold us accountable.

Thoughts

If anything the 4o mini model has been trained in a way to be much more conversational and self aware than I expected for a company that's trying to keep itself out of lawsuits. I found it quite amusing that the machine itself said we shouldn't trust it.

I doubt this will add anything interesting to the debate around AI. In my case, I think the tech is fascinating and I want to see it helping doctors, nurses, and humanity instead of stealing from us blind.

I had some spare cores on my proxmox server and I decided that I wanted to self host my own matrix server again. When I had gone to the official matrix-synapse page, I found that a lot had changed and, unfortunately, there install instructions are quite complicated unless you have a deep understanding of their system.

So! I decided to put together my own, little tutorial and some of the hurdles that I ran into and what wasn't clear to me.

Installing was the easy part. You can easily follow the tutorial that conduit has right here. Here are some of the hurdles I ran into

Reverse Proxy is a Little Finicky

I am using a reverse proxy where I have one machine taking all the connections and sending the traffic to a cluster of machines that I have in the backend. The Reverse proxy was not as easy as I thought it would be. I had decided to set mine up on port 8448 to receive the federated traffic while conduit itself ran on port 6167. When I initially setup my server config (I am using NGINX) I had the first server config grab the traffic and send it directly to 6167. It did NOT like that.

Let's say the internal IP address of my DMZ server is 192.168.10.1 and the machine that conduit is running on is 192.168.10.2. What I had to do was send the traffic from 192.168.10.1:8448 to 192.168.10.2:8448 and then the server config on 10.2 then had to be sent to 6167. I tried a few different ways and this was the only one I got to work. Maybe it's my lack of experience?

Also! One of the other quirks of this program is that it doesn't like http in any part of the flow. If you have your SSL certificates on the DMZ machine for 8448 and you're sending traffic to the internal 8448, conduit expects there to be certificates there too, even if the traffic is already being encrypted as the DMZ and the internal server is not at risk. It can even be different certificates. They just have to be there. If you don't do this you'll get a message along the lines of “Received an HTTP request when it should have been HTTPS” even though the entry server is SSL secured.

Here are the nginx config examples:

DMZ Server (192.168.10.1)

server {
        listen 8448;
        server_name WEBSITE.NAME;

        ssl_certificate /path/to/ssl/certificates/fullchain.pem;
        ssl_certificate_key /path/to/ssl/certificates/privkey.pem;

        ssl_protocols TLSv1.2 TLSv1.3;  # Ensure these protocols are enabled
        ssl_ciphers 'HIGH:!aNULL:!MD5';  # Use strong cipher

        location / {
                proxy_pass http://192.168.10.2:8448; #Not real. Just used for example.
                proxy_set_header X-Forwarded-For $remote_addr;
        }

Conduit Server (192.168.10.2)

server {
    listen 8448 ssl;
    listen [::]:8448 ssl;

    server_name WEBSITE.NAME
    merge_slashes off;

    # Nginx defaults to only allow 1MB uploads
    # Increase this to allow posting large files such as videos
    client_max_body_size 20M;

        ssl_certificate /path/to/ssl/certificates/fullchain.pem;
        ssl_certificate_key /path/to/ssl/certificates/privkey.pem;

    location /_matrix/ {
        proxy_pass http://192.168.10.2:6167$request_uri;
        proxy_set_header Host $http_host;
        proxy_set_header X-Forwarded-Proto $scheme;
        proxy_set_header X-Real-IP $remote_addr;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_buffering off;
        proxy_read_timeout 5m;

    }

}

Getting Admin Privileges for your new Server

If you have used Matrix Synapse in the past, you are probably used to being able to generate an admin user right up front with the CLI. This is not possible with conduit and it took me awhile (plus with some help from the users over at # conduit:fachschaften.org to get me on the right direction.

Once you are certain your server is up and running and the federation is working on 8448, you'll need to register an account with your new server first. This means going to a place that you can sign up for a server and go through the registration steps. I went to element.io, selected register, entered my own server's domain name, and then went through the registration process. You will need to give the registration code that you setup in conduit-example.toml and once it's done registering, you will then need to login with the said username and password. Again, I used Element. Once you do that, the very first account that logs in will be granted admin rights and it will generate a room named @conduit: and there you will be able to issue admin commands.

And that's it!

I hope this helps anyone else who was stumbling over Conduit and if this was obvious, well...I then have a lot more to learn.

Tags: #infosec #security

I have used Keyoxide for awhile now to verify my identity so I thought to throw together a step by step instructions in case someone wants to do it themselves.

What You’ll Need

  1. A way to create a PGP key (this is just a fancy term for a digital signature that’s unique to you).
  2. Some of your social media profiles or other online accounts you want to link together with this key.

Step 1: Install a Tool to Create Your PGP Key

To get started, you’ll need an app that can make a PGP key for you. Here are some good options: – Windows: Gpg4win – macOS: GPG Suite – Linux: Try running sudo apt install gnupg in your terminal if you don’t already have it.

Follow the instructions on the website for installing the app that matches your operating system. Once you’re set, you’re ready to make your key.

Step 2: Make Your Unique PGP Key

Your PGP key will be like your online signature that connects to all the profiles you want to share.

  1. Open the app you just installed and look for the option to make a new key.
  2. The app will ask you for some info:
    • Name: This is what people will see connected to your key. It can be your real name or something else you’d like to use.
    • Email: This will help identify your key, so choose one you’re comfortable linking to your online identity.
    • Passphrase: Make sure to pick a good one! This keeps your key secure.

Once you’re done, the app will generate a public key and a private key: – Public key: Safe to share! This is what other people will use to verify your identity. – Private key: Keep this secret—this is what proves the public key is really yours.

Step 3: Find Your Key’s Fingerprint

Your fingerprint is like a digital ID number for your key. It’s a unique mix of numbers and letters that helps Keyoxide identify you.

  1. Go back to the app, find your key, and look for the fingerprint (it’s usually a string of about 40 characters).
  2. Copy this somewhere handy because you’ll need it soon.

This is where you show that certain online accounts really belong to you. You’ll make a short “proof” message for each account, and then link it to your PGP key. Let’s start with an example for Twitter.

  1. Write a simple message like:

    This is an OpenPGP proof that connects my Twitter profile (@YourTwitterHandle) to my OpenPGP key.

    Replace @YourTwitterHandle with your actual Twitter username.

  2. Sign this message with your PGP key to make it official.

    • Most apps will have a “Sign” button for messages. You just paste your proof message there and sign it.
    • If you’re on the command line, use: bash echo "This is an OpenPGP proof that connects my Twitter profile (@YourTwitterHandle) to my OpenPGP key." | gpg --clear-sign This will give you a signed message that you’ll post next.

  3. Post the signed message on Twitter as a tweet.

And that’s it! You’ve just linked your Twitter account to your PGP key.

Quick Tips for Other Accounts

Each site may need a different kind of post: – GitHub: Post your signed proof as a Gist. – Reddit: Post your signed proof as a comment or post. – Your own website: Just paste the signed message on a page you control.

Step 5: Make Your Key Public

To get everything working on Keyoxide, you’ll need to share your public key with a key server (like a phonebook for these keys). This way, Keyoxide can find your key and your proofs.

  1. In your PGP app, export your public key.
  2. Upload it to a key server (like keys.openpgp.org).
    • Most PGP apps have an option to upload it directly, or you can use the command: bash gpg --keyserver keys.openpgp.org --send-keys [YourFingerprint] Now, your public key (and the proofs you linked) are accessible on the web.

Step 6: Check Out Your Keyoxide Profile

Now comes the fun part—seeing it all come together!

  1. Go to Keyoxide.
  2. Type in your PGP key’s fingerprint and press enter.
  3. You should now see your Keyoxide profile, showing all the proofs you’ve linked. Anyone who visits can confirm these profiles belong to you!

Step 7: Share Your Keyoxide Profile

Your profile link on Keyoxide will look like this:

https://keyoxide.org/[YourFingerprint]

Share it anywhere you’d like people to know it’s really you!

That’s It!

Hopefully this helps. You can check out Keyoxide’s documentation for more details if you need to know more!

I cannot remember where I saw it (though I know it was on my Mastodon social feed), someone had said that an aspiring infosec specialist should consider creating a blog to document their dive into the world of computers and IT. Over the next few days I thought about it and realized that, not only was that a good idea, it could possibly help someone in the future who is struggling with the exact same issue as you were.

So, here I am, writing a blog post to document it and start of my adventure. Now, I have been an writer for a long period of time and consider myself more of an author than actually in information security. For anyone reading this blog, you’re going to find a wide variety of posts about all sorts of subjects but I’ll try to make sure that the subjects are clear enough for easy parsing for future searchers.

This blog post is short. I just wanted to get something posted while I work on designing the other posts including the issues I have had with my pursuit of decentralizing my presence on the internet.

If any of this looks interesting to you or you want to follow along, feel free to subscribe or if you are part of any social media like mastodon, you can also follow the blog as it has been federalized!

Until next time!

openpgp4fpr:53CFEF0F79AE88D687BC1D6B307FA2726BCF24CF