SOLDIERX.COM Nobody Can Stop Information Insemination

Posted by the grugq via Dailydave on Aug 16

Cyber is Calvinball.

I gave a talk back in 2015 [1] which I think has held up rather well. My argument was that cyber is evolving in
unpredictable ways as we learn more about the domain. That the current state of the art has huge blind spots we aren’t
even thinking about. The next year was, of course, the 2016 disinformation campaign fed by cyber loot.

I feel that a great deal of cyber war literature is based on knowledge derived from...

Posted by Dave Aitel via Dailydave on Aug 15

Before there were words, calculated as the softmax of a list of possible
tokens, there were just vectors of nano-electrical potential in cells
soaked in a hormonal brew of electrolytes, operating on a clock cycle of
"slow, but fast enough". In this sense, as we now know
<https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10472538/>, we generate words
and we know, in our heads, what we are, in the same way as we generate
limbs, with each...

Posted by Dave Aitel via Dailydave on Aug 13

https://github.com/CloudCrowSec001/CVE-2024-38077-POC/blob/main/CVE-2024-38077.md
https://github.com/Wlibang/CVE-2024-38077/blob/main/One%20bug%20to%20Rule%20Them%20All%2C%20Exploiting%20a%20Preauth%20RCE%20vulnerability%20on%20Windows%20(2024_8_9%2010_59_06).html

But while you are at it, always good to watch a video for no reason:
https://www.youtube.com/watch?v=mVXrl4W1jOU

-dave

Posted by Don A. Bailey via Dailydave on Aug 13

Posted by Dave Aitel via Dailydave on Aug 12

DefCon is a study in cacophony, and like many of you I'm still digging
through my backlog of new research in multifarious browser tabs, the way a
dragonfly keeps track of the world through scintillated compound lenses. In
between AIxCC (which proved, if anything, the boundaries
<https://dashboard.aicyberchallenge.com/collectivesolvehealth> of automated
bug finding using current LLM tech?), James Kettle's timing attack research...

Posted by Dave Aitel via Dailydave on Aug 02

Cordyceps Analysis Report on PRANA Network Hack and Leak Operation:
https://docs.google.com/document/d/1oOJbBTUwyK85ZKYAAdwWqxk-sMvqrBqzJYX1oziTFu4/edit?usp=sharing

Lately I've been reading a lot of academic papers, mostly the Research
Handbook on Cyberwarfare
<https://www.elgaronline.com/edcollchap/book/9781803924854/book-part-9781803924854-6.xml>.
Some of them are good papers! JD Work has a paper in it! But also some of
them get...

Posted by David Manouchehri via Dailydave on Jul 28

Breaking down a prompt into multiple steps works pretty well for us. e.g.
first we get generic mean reasons:

[image: image.png]

Then I just shove the mean reasons into the system message (you can do this
with another LLM call instead in real life, I just cheated by copy pasting
since there's already too many screenshots in this email):

[image: image.png]

This is with gpt-4o-2024-05-13 above, but you can see below it works with
Llama 3.1...

Posted by Jason Ross via Dailydave on Jul 25

It's likely this is going to happen anyway, the new Mistral just dropped
and seems to perform roughly on par with llama3 and gpt4o, so the next wave
of fine tuned versions like dolphin are almost certainly coming soon.

OpenAI also has announced free fine tuning of gpt4o mini until late
September (up to 2m tokens/day) so it may be possible to fine tune around
some of its guardrails for a reasonable cost.

Posted by Robert Lee via Dailydave on Jul 24

Posted by Dave Aitel via Dailydave on Jul 24

[image: image.png]
Above: LLAMA3.1 8B-4Q test results via OLLAMA

So recently I've been doing a lot of work with LLMs handling arbitrary
unstructured data, and using them to generate structured data, which then
gets put into a graph database for graph algorithms to iterate on so you
can actually distill knowledge from a mass of nonsense.

But obviously this can get expensive via APIs, so like many of you, I set
up a server with a A6000 that...