Comparison of 4 LLM and agent-zero in an elementary pentest competition

My goals

1. Get new experience with autonomous AI agents → agent-zero
2. See how common (not specialized) AI agent could perform penetration tests
3. Check several actual LLMs on pentest tasks

Attention

1. This is not a real research and guide
2. agent-zero and used LLMs are not intended for pentesting
3. The results below do not indicate that the models are good or bad.
4. The penetration test target is a local copy of OWASP Juice Shop (Probably the most modern and sophisticated insecure web application)

How AI see an AI agent

The pain of image generation in Kandinsky from n8n

Input data:

• n8n Self-Hosted Community Edition 1.112.6
• Fusion Brain (Kandinsky) AI service — offers 100 free image requests per month

Goal: Create a workflow in n8n to generate images using the Fusion Brain API.

This turned out to be harder than I expected.

NVIDIA CMP 50HX for AI: A Step-by-Step Guide

I’m gradually diving into the AI world. To gain hands-on experience, I bought a GPU specifically for image generation with ComfyUI. Wanting an affordable yet efficient solution, I chose the NVIDIA CMP 50HX - a card originally designed for mining. It has 10 GB of VRAM and 446 second-generation Tensor Cores, and cost me only $75.

Since I couldn’t find a complete guide for using mining GPUs in AI, I’ll share my experience here.

What about drivers?

❌ GeForce Game Ready Driver (v560.94 in my case) didn’t work.
❌ NVIDIA’s website doesn’t offer special drivers for this card.
❌ NVIDIA A100 Data Center driver (v573.39 for Windows) also failed.
✅ The solution: Install CUDA Toolkit (v12.9.1 for me) and NVIDIA cuDNN.

After this, Windows 10 recognized the CMP 50HX without issues.

NVIDIA dedicated GPU for professional mining comparison

I couldn’t find a clear table comparing all NVIDIA GPUs specifically designed for cryptocurrency mining.

So I decided to create one myself.

The GPUs under consideration:

• NVIDIA CMP 170HX
• NVIDIA CMP 100HX
• NVIDIA CMP 90HX
• NVIDIA CMP 70HX
• NVIDIA CMP 50HX
• NVIDIA CMP 40HX
• NVIDIA CMP 30HX
• NVIDIA P102-100
• NVIDIA P104-100
• NVIDIA P106-100

The information was collected from various sources, some of which provided conflicting data.
There may be different versions of these GPUs, resulting in differences in specs.
I can’t ensure the accuracy of all the information given here.

I hope you find it helpful.

What’s wrong with the Pixel Camera’s panorama mode?

I really like Google Pixel phones. I like them mostly because of the clean Android and long software updates. They have many good features, but there is one problem that I don’t understand.

It’s about the panorama mode in Google Camera. The quality of the final photo is very low. On my Pixel 4, the vertical resolution is only around 1700 pixels.

I looked online and found out this is not only my problem. It seems like this has been happening even on newer models like the Pixel 8 Pro.

My current Google Camera version is 8.7.250.494820638.44, and I also tried the newer version 9.2.113.604778888.19 (GCam), but the result was the same.

Some proofs:

• https://www.reddit.com/user/RazzmatazzWeak2664/ photo -

• Comparison between Pixel 8 Pro and Pixel 9 Pro – link . Look at the dimensions tab of Pixel 8 Pro photos — it’s only 1280 pixels vertically.

I'm sad about this, but I hope Google will fix it in the future. @Google, right?

Troubles with accessing Samba/SMB share from Windows 10

I have two servers with Samba shares and I had triubles with accessing both of them from my Windows 10 (1809 LTS) PC.

OS	TrueNAS 13.0-U6.2	Ubuntu 24.04.1 LTS
Samba version	4.15.13	4.19.5
Samba config	ea support = No kernel share modes = No path = /mnt/RAID1/NAS posix locking = No read only = No smbd max xattr size = 2097152 vfs objects = fruit streams_xattr shadow_copy_zfs ixnas zfs_core aio_fbsd fruit:resource = stream fruit:metadata = stream nfs4:chown = true ixnas:dosattrib_xattr = false	comment = Torrent Server path = /torrents read only = yes browsable = yes guest ok = yes public = yes create mask = 0775 directory mask = 0775 force create mode = 0775 force directory mode = 0775

The error I got - "Network Error: Windows cannot access"

Which is faster: 0.0.0.0 or 127.0.0.1?

When I was experimenting with hosts and Pi-Hole, I read that "0.0.0.0" address in the hosts works faster thфn "127.0.0.1". And now I decided to test and verify it.

About tests

• To automate tests I prepared code on Python "DNS_Loopback_Speed".
• To get more accurate results I performed 10000 tests for each address and took the average.
• I have two devices running Windows 10 and Ubuntu 22.04, so I also compare speed of resolving between Windows and Linux.
• I've seen option of using 127.0.0.0 in hosts, therefore I add it too.
• Just to compare delays between local resolution (hosts) and external DNS servers I add 2 tests (my local Pi-Hole DNS and Google DNS). My devices connected to network via Wi-Fi, so results may not be reliable.

Brief information regarding these IP addresses:

• 0.0.0.0 - non-routable meta-address
• 127.0.0.1 - loopback Internet protocol address = localhost
• 127.0.0.0 - loopback network address

Windows Update | Logitech - Camera - 2.3.7.5 | Error 0x800f0206

A few months ago I got an error 0x800f0206 in Windows Update regarding installation "Logitech - Camera - 2.3.7.5" driver.

I use "Logitech HD Webcam C615" and "Windows 10 Enterprise 2016 LTSB 1607". Windows Update downloaded the update (https://www.catalog.update.microsoft.com/ScopedViewInline.aspx?updateid=5da34c2a-7b10-4b3a-b3c2-764ddce2e8b0) but can't install.

lighttpd | Behavior of «setenv» directives

I faced with unclear (for me) behavior of «setenv» directives of «lighttpd» webserver and decide perform several experiments and show the results.

Environment and conditions:

• Ubuntu 22.04.2 LTS

• lighttpd 1.4.63-1ubuntu3.1

• Two modifications of default «/etc/lighttpd/lighttpd.conf»:

• Changed listened port to 3000.

• Enabled module «mod_setenv».

• All other changes performs in «/etc/lighttpd/conf-enabled/» folder.

• Defined two URLs\paths («http://10.0.0.90:3000/» and «http://10.0.0.90:3000/admin/») to show how merging works

Outline VPN Server doesn't work after installation

I would like to share with you a solution of one issue of the Outline VPN server. It might be useful for newbies in Linux and Docker. Hope this helps someone.

The issue

After successful installation of Outline VPN server it goes down (Outline Manager can't connect to the server).