"AI & Machine Learning" Cybersecurity Conversation
HTML-код
- Опубликовано: 31 июл 2024
- To help support me, check out Kite! Kite is a coding assistant that helps you faster, on any IDE offer smart completions and documentation. www.kite.com/get-kite/?... (disclaimer, affiliate link) Hang with our community on Discord! johnhammond.org/discord
If you would like to support me, please like, comment & subscribe, and check me out on Patreon: / johnhammond010
E-mail: johnhammond010@gmail.com
PayPal: paypal.me/johnhammond010
GitHub: github.com/JohnHammond
Site: www.johnhammond.org
Twitter: / _johnhammond
![The moment we stopped understanding AI [AlexNet]](http://i.ytimg.com/vi/UZDiGooFs54/mqdefault.jpg)
I’ve enjoyed watching John progress throughout the years. He has learned a lot.
We, aye
This is such a great channel!
Can't wait for this ❤️❤️
thank you
A lot of interesting information in this video.
This is gonna be super interesting 🤠
Taught more about AI than professors can at universities
Waiting, man!
As someone who had this stuff as specialization, way back at university, I can not only attest to the accuracy of what he said, but also think this is one of the best broad stroke overviews out there in terms of clarity, compactness and general information about this field. Either he's really well prepared or just a natural born explainer. Complements and kudos to your buddy, John!
P.S.: I just stumbled over the thread with the guy claiming BS. Maybe my knowledge is a little dated (13-18 years old), but in my opinion your buddy explains complex topics quite well in laymans terms. If he cuts corners for clarity, it serves the purpose. The other guy claims to have made comments in chat - if I find the time, I'll look into them and check if he just nitpicks or has real points I overlooked when watching this...
Very flattered and grateful to see this, thank you! I passed the message along to Pat! :)
Don't worry about that PS. I'm 1 year out of an AI/ML master. That is one of the sad things of the burgeoning sphere of AI/ML, everyone uses different words for different things as it better applies to them and everyone thinks that their use of the words is gospel and others are wrong. There are a lot of people who use AI/ML to sell snake oil as Pat himself noted, so healthy cynicism is always needed, but I wouldn't go so far as to call BS. Everything that John wrote down is correct, it's just the examples of what the MLs are which were offbase to some definitions.
A few examples:
Support Vector Machines, at their base, use matrix math to draw conclusions.
• academically this takes the form of a classifier that tries to find a line that best bisects 2 classifiers at which the distance tangent to the line to the closest point of each class is maximized, take that X and O example at the very start - an SVM that is taught in academics would draw a line parallel to both of them, between them, and then new data would be tested by looking at which side the line is to determine the class, taking into account the relative distance to the line for some concept of confidence
• I have also seen the term SVM applied to like Principal Component Analysis which uses the eigenvectors to summarize the data and reduce the ratio of N vs M or I've even seen it used in place of the word regression when talking about the data in a matrix
Pat repeatedly called Adversarial networks "GANs", but this is likely a slip of the tongue as GAN is a known industry term on it's own
• by pure technicality GANs are Generative Adversarial Networks, in that the Adversary is tasked with *generating* fake records that the Primary will falsely classify, standard example of this is a Horse vs Zebra discriminator, with some amount of initial data tagged, train the Primary to discriminate, then have the Adversary take a horse picture and convolve it to look like a zebra (or generate one completely from scratch) and then have the two argue for a while allowing the Primary to train on actual data for a while from time to time so that it keeps a steady grasp of what a zebra and horse actually look like, once done, split them up and you have a Primary that can discriminate between different Equine and an Advisary that can either turn pictures of a horse into a zebra and (if run backwards) zebra into horses. Or more generally a picture horse-ifying and Zebra-ifying function.
• Adverserial networks is a more broad term of two systems working as checks to the other. What Pat described was more of a IAN, an Inverse Adversarial Network, that (depending on the loss function used) is usually exactly the same neural network but run with the other class as the positive case, if the loss function is bias towards reducing false positives over false negative rates, this inverse training allows for equal detection of either case. other IANs are really just the same neural network but with different starting parameters, or trained on separate data, the idea being to create a counter check where the primary says no, normal behaviour, but the Inverse says yes, hacker.
Time to learn
Egarly waiting ✌️
I'm starting my MSc Thesis, and I want to do research on AI concepts applied to Cybersecurity.
So this video was an absolute joy to find! Thank you! 😀
Great conversation, awesome work!
I'll share a YT channels 'Lex Fridman' that holds great talks about AI with other great people. I don't know, it just came to my mind (but there are also a lot of other great ones).
#& I am still a complete noob, just started with learning python .. your videos also help a lot, all of you guys are just amazing. Keep moving forward & stay ethical !
Start a try hack me series where you start from the bottom, literally with no dependencies installed, no word list pre installed, absolutely from scratch !!! Please !!
well then the video will go for hours
moreover one should try to install his own toolset by himself so that he can easily access it
and will also understand what to do if the tool misbehaves
and when john uses some tool that u dont have try installing it by urself u may get some errors
but believe me those errors are worth teaching u more about ur linux system
@@abisrug4898 I done the basic pentest right, I literally sat installing everything, error after error finally came right only to get stuck at sending linpeas over to a target and getting a weird error which the target couldn't find linpeas.sl file... but yes learning the hard way is well worth it.
Could you point me in the right direction on how to install a custom linux with the basic pen testing tools ?
@@PathFinder_ZA007 MATE U MUST TRY THE KALI LINUX LATEST DISTRO OR PARROT IS ALSO REAL GOOD
COZ BOTH OF THEM COMES WITH GOOD AMOUNT OF TOOLS WHICH CAN EASILY BE INSTALLED USING SOME COMMANDS
BTW,WHICH LINUX DISTRO R U USING?
@@abisrug4898 thanks for the quick reply bro, I just started studying ceh, I'm using debian 2020 v3
@@PathFinder_ZA007 GRT,TRY A KALI OR PAROOT IT WILL MAKE UR LIFE MUCH EASIER.....ALL THE BEST
I love AI and am learning it😀
John. What Linux OS do you favor?
what are the links and social of the guy speaking ?
Hi! I'd like your opinion on the platform INE Training, I don't know if it's worth it. Have you used it? Have you known anybody who has? They're quite expensive. Cheers mate!
Yes yes yes
😍
💞
This video seems helpful, but it will be a lot helpful if it is timestamped.
Dollarboysushil is here
john you are not to quit cybersec yea?
FLOOR GANG
ceiling gang
Hello
build one agent rl, keras.
Anyone believe robots with hacking skills..🙃
Hey John, Try - carpe diem 1 tryhackme, it is very good box and worth a walkthrough :)
attachment... attachment......LOL...