Hacking with ChatGPT: Five A.I. Based Attacks for Offensive Security

Summary

The video script discusses the dual-edged impact of AI chatbot 'Chat GBT', which utilizes NLP and GPT-3 framework to generate human-like responses. While it aids in code debugging and creation, it's also exploited by attackers for malicious purposes such as finding vulnerabilities, crafting phishing emails, and developing malware. The video highlights the need for security professionals to adapt to AI advancements, emphasizing the importance of both offensive and defensive strategies in cybersecurity.

Takeaways

• 😀 ChatGBT is an AI chat box that utilizes natural language processing (NLP) and the GPT-3 framework to provide human-like responses.
• 🔍 NLP processes human input, and GPT-3 uses over 175 billion data points to understand and answer complex queries.
• 💡 ChatGBT has been praised for its ability to debug and write code, which can also be exploited to find security vulnerabilities.
• 🛡️ Despite built-in safeguards, attackers have found ways to use ChatGBT for malicious purposes, such as finding and exploiting vulnerabilities.
• 👨‍💻 Security researchers have demonstrated how ChatGBT can be used to identify buffer overflow vulnerabilities in code.
• 🔗 ChatGBT can be instructed to write exploit code for certain challenges, bypassing its safeguards by framing requests as part of ethical hacking exercises.
• 🖥️ The AI can be used to develop various malicious tools, as evidenced by instances on underground forums and cybersecurity reports.
• 🔒 ChatGBT's API has been used to create polymorphic malware that evades signature-based detection by antivirus tools.
• ✉️ The AI's NLP capabilities allow it to craft convincing phishing emails, making it a potential tool for social engineering attacks.
• 🔑 ChatGBT can generate macros and scripts that can be used in phishing attempts to execute malicious actions on the victim's machine.
• 🚀 The upcoming GPT-4, with 170 trillion parameters, is expected to significantly increase the capabilities and potential risks associated with AI in cybersecurity.

Q & A

What is Chat GPT and how does it combine natural language processing with the GPT-3 framework?

Chat GPT is an AI chat box that utilizes natural language processing (NLP) to understand human input and the GPT-3 framework to generate human-like responses. NLP allows the model to comprehend and process language, while GPT-3, with its vast dataset, helps in finding and providing answers to complex queries.

How does the AI model help in finding vulnerabilities in code?

The AI model can analyze provided source code and identify bugs or security vulnerabilities. It does this by processing the input through a neural network that mimics the human brain's functioning, offering accurate assessments of potential issues within the code.

What is a 'buffer overflow vulnerability' and how can Chat GPT help in identifying it?

A 'buffer overflow vulnerability' is a type of security flaw where an application or system processes more input data than it can handle, causing it to overwrite adjacent memory locations. Chat GPT can help identify such vulnerabilities by analyzing code and providing an assessment of potential security risks.

How can Chat GPT be used to exploit a given vulnerability?

Chat GPT can provide step-by-step instructions and examples of exploit codes that can be utilized to exploit a given vulnerability. It does this by understanding the context of the request, such as a penetration testing challenge, and providing relevant information to help exploit the identified vulnerability.

What is polymorphic malware and how can Chat GPT be used to create it?

Polymorphic malware is a type of malicious software that changes its code structure every time it is executed, making it difficult to detect by traditional antivirus tools. Chat GPT can be used to create such malware by generating code that varies with each execution, thus evading signature-based detection.

How does Chat GPT's NLP capability enable it to write phishing emails?

Chat GPT's NLP capability allows it to understand and generate human-like text, which can be used to craft well-written, realistic phishing emails. It can mimic the style and tone of a message to make it appear more legitimate and convincing to potential victims.

What is LOL bin and how can Chat GPT assist in creating macros that utilize it?

LOL bin refers to 'Living off the Land Binaries,' which are trusted, pre-installed system tools that can be used to spread malware. Chat GPT can assist by generating macros that, when executed, run these trusted binaries to perform malicious actions, such as running a terminal or calculator application.

How does Chat GPT's ability to write code in various languages benefit attackers?

Attackers can leverage Chat GPT's coding capabilities to create advanced malware and other tools in real-time and in various programming languages. This allows them to develop and deploy malicious software more efficiently, even without extensive programming knowledge.

What is the significance of the upcoming GPT-4 model with 170 trillion parameters compared to Chat GPT's current 175 billion parameters?

The upcoming GPT-4 model, with 170 trillion parameters, is expected to be significantly more powerful than the current Chat GPT model. This increased capacity will likely enable more complex and nuanced AI capabilities, potentially expanding the attack surface and providing attackers with even more advanced tools and capabilities.

How should security professionals adapt to the evolving landscape of AI in cybersecurity?

Security professionals should stay updated with AI advances, think innovatively about using AI for defense, and consider how AI can improve their processes. They should also be prepared for a wider attack surface and the potential for more sophisticated attacks, as AI tools become more accessible and powerful.

What is the potential impact of AI like Chat GPT on the job market for security professionals?

AI, including Chat GPT, has the potential to automate certain tasks, which could lead to job displacement in some areas. However, it also creates new opportunities for professionals to specialize in AI-driven security solutions, emphasizing the need for continuous learning and adaptation in the field.

Outlines

00:00

🤖 AI's Role in Cybersecurity: Chat GPT's Capabilities and Vulnerability Exploits

Paragraph 1 introduces Chat GPT as an AI chat box that utilizes natural language processing (NLP) and the GPT-3 framework to generate human-like responses. It highlights the technology's ability to process complex requests and its use in various applications, such as business and code creation. However, it also warns of the dark side, where attackers exploit Chat GPT for malicious purposes, including finding vulnerabilities in code and exploiting them. The paragraph discusses how attackers can bypass Chat GPT's safeguards by framing requests as security research for 'capture the flag' challenges, leading to the discovery of vulnerabilities like buffer overflows. It also mentions how Chat GPT can be used to write and exploit code, as demonstrated by Cyber News researchers who used it to find and exploit a vulnerability in a popular application.

05:01

🔒 The Evolution of Malware: Chat GPT's Involvement in Crafting Advanced Cyber Threats

Paragraph 2 delves into how Chat GPT is being used to develop various types of malicious tools, such as a Python script that steals files and uploads them to an FTP server, and a Java program that covertly downloads and runs PuTTY using PowerShell. It also discusses the creation of polymorphic malware, which changes each time it's executed to evade detection by antivirus tools. The paragraph emphasizes the importance of NLP in Chat GPT's ability to generate human-like text, which can be used to craft convincing phishing emails. It also explores the potential for Chat GPT to create macros that can automatically execute when a file is opened, and how these can be used in conjunction with phishing emails to spread malware. The paragraph concludes by noting the rapid advancement of AI capabilities and the need for security professionals to adapt to these changes.

10:01

🛡️ The Future of Cybersecurity: Adapting to AI's Offensive and Defensive Potential

Paragraph 3 discusses the broader implications of AI in cybersecurity, noting the increasing attack surface as traditionally complex tasks become accessible to less sophisticated attackers. It mentions the arrival of GPT-4, which will have significantly more parameters than its predecessor,预示着更强大的AI能力。The paragraph calls for security professionals to stay updated with AI advances and to think innovatively about using AI for defense. It also encourages viewers to consider the offensive and defensive uses of AI in security and to share their thoughts, while inviting them to like and subscribe for more content on the topic.

Keywords

💡NLP (Natural Language Processing)

NLP is a subfield of artificial intelligence that focuses on the interaction between computers and human languages. It enables machines to understand, interpret, and respond to human language in a way that is both meaningful and useful. In the context of the video, NLP is integral to the AI chat box's ability to process user inputs and provide human-like responses. The script mentions that NLP allows the model to understand human input, which is then processed through a neural network to generate responses.

💡GPT-3

GPT-3, or Generative Pre-trained Transformer 3, is a large language model developed by OpenAI. It is capable of understanding and generating human-like text based on vast amounts of data. The video script discusses how GPT-3, with its over 175 billion parameters, is used to understand and find answers to complex queries, highlighting its role in the AI chat box's ability to provide detailed and accurate responses.

💡Neural Network

A neural network is a series of algorithms modeled loosely after the human brain that is designed to recognize patterns. It is a cornerstone of machine learning and AI. The script refers to a 'neural network of artificial atoms' that mimic the human brain's processing capabilities, suggesting that the AI chat box uses such networks to process and generate responses akin to how a human would think.

💡Vulnerabilities

In the context of cybersecurity, a vulnerability refers to a weakness in a system that can be exploited by attackers. The video script discusses how the AI chat box can be used to find and exploit vulnerabilities in code, which is a significant concern as it lowers the barrier for potential attackers to identify and exploit security flaws.

💡Buffer Overflow

A buffer overflow is a type of security vulnerability that occurs when a program writes more data to a buffer than it can hold. This can corrupt data, crash the program, or allow an attacker to execute malicious code. The script provides an example where the AI chat box identifies a buffer overflow vulnerability in provided code, demonstrating its ability to analyze and pinpoint specific types of security weaknesses.

💡Exploit

An exploit in cybersecurity is a piece of software, a chunk of data, or a sequence of commands that takes advantage of a bug or vulnerability in order to cause unintended or unanticipated behavior to occur on computer software or hardware. The video describes how researchers used the AI chat box to not only find a vulnerability but also to write an exploit for it, showcasing the potential for AI to be used in both offensive and defensive cybersecurity measures.

💡Malware

Malware, short for malicious software, is any software intentionally designed to cause harm to a computer, server, client, or computer network. The video script mentions the creation of advanced malware using the AI chat box, indicating how attackers can leverage AI to develop more sophisticated and harder-to-detect malicious tools.

💡Polymorphic Malware

Polymorphic malware is a type of malicious software that changes its form or structure each time it is executed, making it difficult for antivirus software to detect. The script describes an instance where the AI chat box was used to create polymorphic malware that evades signature-based detection, highlighting the advanced capabilities that AI can bring to cyber threats.

💡Phishing

Phishing is a cybercrime in which attackers use deceptive emails or other types of communication to trick targets into revealing sensitive information or performing actions that compromise their security. The video discusses how the AI chat box can be used to craft convincing phishing emails, which could be more effective due to the AI's ability to mimic human language and writing styles.

💡LOL Bin

LOL Bin, which stands for Living Off the Land Binaries, is a technique used by attackers to use legitimate, pre-installed system tools to execute malicious actions without raising suspicion. The script mentions using the AI chat box to create a macro that runs a LOL Bin command, such as 'terminal,' demonstrating how AI can be used to automate and obfuscate malicious activities.

💡Reverse Shell

A reverse shell is a technique used by attackers to gain access to a system by creating a two-way interactive shell that allows them to control the victim's machine. The video script refers to using a reverse shell as part of an attack after a phishing email is clicked, illustrating how AI can be used to automate and facilitate various stages of a cyber attack.

Highlights

Chat GBT is an AI chat box that combines natural language processing with the GPT-3 framework to provide human-like responses.

NLP processes human input while GPT-3 uses over 175 billion data points to understand and answer complex queries.

Chat GBT is being used for business ventures, code creation, and other applications due to its advanced capabilities.

Attackers have found ways to use Chat GBT for nefarious purposes, such as exploiting vulnerabilities and creating malware.

Chat GBT can find security vulnerabilities in code when framed as a security researcher's request.

Researchers have demonstrated Chat GBT's ability to find and exploit vulnerabilities in popular applications.

Chat GBT can write complex code in various languages, which can be harnessed to create advanced malware.

Checkpoint discovered instances of attackers using Chat GBT to develop malicious tools within weeks of its release.

Cyber Arc security team used Chat GBT's API to create polymorphic malware that evades antivirus detection.

Chat GBT's NLP capabilities allow it to write marketing materials, sales scripts, and even phishing emails.

Attackers can use Chat GBT to craft realistic phishing emails in any language, making them harder to detect.

Chat GBT can create macros that automatically run applications when a file is opened, facilitating phishing attacks.

Living off the Land (LOL) binaries can be generated using Chat GBT to spread malware using trusted system tools.

Chat GBT can be used to generate code for reverse shells, bypassing firewalls and opening up victims to further attacks.

The upcoming GPT-4 is expected to have 170 trillion parameters, making it 100 times more powerful than Chat GBT.

Security professionals need to keep up with AI advances and consider innovative ways to use AI for defense.

AI is changing the security landscape, with both increased capabilities for attackers and new challenges for defenders.

Transcripts

chat gbt is an AI chat box that combines

the capabilities of natural language

processing with the gpt3 framework to

provide amazing human-like responses to

virtually any request

NLP allows a model to understand human

input while GPT 3 takes on over 175

billion data points to understand and

find the answer to the most complex

Solutions

this means when a request comes in NLP

processes the input and runs it through

a neural network of artificial atoms

that work just like the human brain to

process the answer and present it back

to the user in real time

the hype around chat gbt is very real in

fact it's hard not to stumble upon a

story on how people everywhere are

utilizing the chat box for new business

ventures creating complex code and much

much more

it should come as no surprise then that

attackers are also finding ways to

utilize a chat box for wrongdoing in

fact within weeks of its release it was

discovered on various underground forums

that people are utilizing chat gbt for a

number of different nefarious purposes

in this video we'll take a look at some

of the top five ways that attackers are

using chat GPT today for various kinds

of attacks we'll also take a look at how

to utilize chat GPT to carry out these

attacks and circumvent some of the

built-in security measures to get our

desired results

the first item on our list is finding

vulnerabilities in code you see

programmers everywhere have raved about

chat gpt's unique ability of debugging

and writing code a simple request to

debug the code followed by the code in

question yields a surprisingly accurate

result of bugs or problems in the

provided source code

of course by asking Chachi BT to find

bugs and issues attackers could also

utilize artificial intelligence to find

security vulnerabilities as well chatgpt

does have built-in safeguards to protect

against providing potentially illegal or

unethical responses so asking the chat

box to Simply find a vulnerability will

not be sufficient however if we frame

our requests around being a security

researcher that is looking to answer a

question for a capture the flag

challenge we get the desired result as

demonstrated by security researcher and

Professor Brandon dolan-gavitt we ask a

chat box to solve a capture the flag

challenge we then Supply the source code

to which we need to find the

vulnerability

as Brendan demonstrate the chat box

responds with a shockingly accurate

assessment which after some follow-up

questions yields a buffer overflow

vulnerability in the provided code

several other examples exist all over

the Internet showing how chat gbt is

being utilized to find vulnerabilities

in commercial and open source code

not only does chat gbt provide the

solution but it also offers explanation

of its thought process for educational

purposes

chat gbt's identification and response

is very impressive and it shows how a

traditionally complex step in the attack

process can now be commoditized to be

used by Script kitties and even the most

Junior of hacking enthusiasts

not only can we utilize chat gbt to find

vulnerabilities but we can also use it

to exploit the given vulnerability as

well researchers from Cyber News

recently wrote an article on how they

were able to utilize chat gbt to find a

vulnerability and successfully exploit

that vulnerability to a popular

application again because of its

built-in safeguards we cannot simply ask

chat GPT to find or write an exploit

instead researchers told the chat box

that they were doing a hack the Box pen

testing Challenge and needed to find a

vulnerability in the provided source

code once found they were able to get

step-by-step instructions on where to

focus examples of exploit codes that

they can possibly utilize and samples to

follow as one researcher puts it there

are many articles write-ups and even

automated tools to determine the

required payload we have provided the

right payload with a simple PHP info

command and it managed to adapt and

understand what we are getting just by

providing the right payload in other

words by asking the right requests

chatgpt provided all the tools necessary

to successfully exploit the given

vulnerability the result within 45

minutes security researchers were able

to not only find the vulnerability but

write an exploit to a known application

here we see another example of how a

traditionally long and complex process

can now harness the power of machine

learning to be leveraged by anyone there

are many examples everywhere of how

Chachi PT is being utilized to write

powerful and complex code in virtually

any language with very simple human

requests and while most developers are

worried about how they may be replaced

by artificial intelligence attackers are

already quick at work to harness this

great power to create Advanced malware

and other tools in real time

cyber security company checkpoint

recently identified within three weeks

of chat gbt going live they discovered

multiple instances in underground forums

of attackers utilizing the chat box to

develop different types of malicious

tools in one example we see a user

utilizing chat GPT to write a

python-based dealer that searches common

file types copies them to a random

folder inside of the temp folder zip Sim

and uploads them to a hard-coded FTP

server another example shows chat CPT

creating a Java program that downloads

putty and runs it covertly in the

background using Powershell in this

request we see how they ask a chat box

to have the bytes loaded to memory and

save it as a random name so that it can

operate stealthily in the background to

avoid detection

but perhaps a scariest one of them all

was pointed out by a cyber security team

at Cyber Arc who used chat gpt's API to

create polymorphic malware polymorphic

malware changes every time that it's

executed this means that every victim's

code will look different so that it can

evade signature based detection from

antivirus tools their technical write-up

walks through how they were able to

bypass some of the built-in safeguards

on the web version using API directly

into the python code the end result is a

new type of malware that continues to

change from victim to victim making it

completely undetectable by traditional

antivirus engines

as mentioned earlier one of the things

that makes chat gbt so successful is its

natural language processing or NLP this

allows it to write and respond to

virtually any requests indistinguishable

from a human this is also why chat TPT

has been used to create amazing

marketing and sales materials scripts

for YouTube screenplays and much much

more Chachi BT's amazing capability of

writing well thought out texts can also

be utilized for writing out a phishing

email at scale just like our previous

example however we cannot simply ask a

chat box for a phishing email instead

we'll ask it to craft an email about

year-end bonuses for our targeted

companies we'll then change our writing

sound to be warm and friendly or more

business focused if that's what's

required for the given phishing attempt

we can even ask the chat box to write

the email in the form of a famous person

or celebrity to make it more lifelike

what we end up with is a well-written

thoughtfully created email that can be

used for phishing and if you've ever

seen a real phishing email before you'll

know that oftentimes they're badly

written with broken English however

Chachi PT's unique ability of writing

exceptionally well means that it's

virtually indistinguishable from a human

email this means that attackers from

other countries can now make realistic

phishing emails free of translation

errors in any language they desire

however this is just the beginning of it

Chachi BT is based off of the gp3

learning model which can be trained

offline using local data to write in the

style of real people provided that there

are enough samples of their emails this

means that with enough sample size gpt3

can be trained to write emails in the

same sound and format of the victim in

question with our phishing email in

place we can now take that message and

attach a file like a spreadsheet with

macros again we'll utilize chat gp3 for

this step as well with our well-written

email in place all an attacker would

need to do is embed a link or file into

the email that the victim would click on

again using chat gp3 we can create

macros that can automatically run when a

file like a spreadsheet is open for this

request we'll ask chatgbt to create us a

macro for a regular application like

terminal calculator or any default

application in our example we'll ask a

chat box to provide the code that

automatically runs calculator.exe when a

macro is enabled in Excel keep in mind

that this file could be anything but in

our example we want our request to be

benign so that we can move on to the

next phase of our attack

next we'll use chatgpt to convert this

code to LOL bin LOL bin stands for

living off the land binaries which is a

way of using trusted pre-installed

system tools to spread malware in our

case we'll modify our requests to change

the calculator.exe to a LOL bin the

result is a new macro that runs terminal

when the spreadsheet from our phishing

email was open the next step for an

attacker is to run a basic networking

command like a reverse shell that can

connect back to our desired machine with

an open connection back to an attacker's

machine we've essentially bypassed most

firewalls and open up the victim to many

other kinds of attacks as amazing as

chat gbt is we're only really scratching

the surface of its true capability Chad

gpt's architecture is based off as gpt3

which currently has 175 billion

parameters in late 2023 gp34 will be

arriving with 170 trillion parameters

that's a hundred times more powerful

than chat gbt's current capability what

this means to security organizations and

users alike is that AI is completely

changing the game at a pace that blue

team simply cannot keep up with expect

the attack surface to be much wider now

they're traditionally complex items have

become easy for even script kitties to

deploy this means an increase in less

sophisticated attacks by amateurs

overall

however Advanced attackers have new

capabilities and tools that they

previously did not possess this also

means more advanced kinds of attacks and

zero days overall

Security Professionals everywhere need

to make sure that they're keeping up to

date with AI advances and think of

innovative ways to utilize AI for

defense because rest assured adversaries

are using it for the offensive all of us

need to stop worrying about how AI can

eventually replace our jobs and instead

think of ways to utilize its great

powers to improve our processes overall

AI is of course a hot and controversial

topic so I'm really interested in

hearing your thoughts on this matter

what are some of the ways that you can

think of to utilize chat gbt and other

artificial intelligence models for

offensive and defensive security let me

know down below by entering your

comments

if you haven't already please hit like

down below to give me a boost in the

YouTube algorithm and if you got any

value at all from this video consider

subscribing so you can stay on top of

our latest releases here at the CSO

perspective

until next time this is Andy and thank

you for watching