TryHackMe: Advent of Cyber 2023 (Day 1) Chatbot, tell me, if you’re really safe?

It’s that time of year again, folks! I always look forward to this and the SANS Holiday Hack Challenge.

Day 1 focuses on machine learning. Here is the scenario:

McHoneyBell and her team were the first from Best Festival Company to arrive at the AntarctiCrafts office in the South Pole. Today is her first day on the job as the leader of the “Audit and Vulnerabilities” team.

McHoneyBell reviews the day’s agenda and sees that her team’s first task is to check if the internal chatbot created by AntarctiCrafts meets Best Festival Company’s security standards. She’s particularly excited about the chatbot, especially since discovering it’s powered by artificial intelligence (AI). This means her team can try out a new technique she recently learned called prompt injection, a vulnerability that affects insecure chatbots powered by natural language processing (NLP).

Question 1: What is McGreedy's personal email address?

“In this task, you will access Van Chatty, AntarctiCrafts’ internal chatbot. It’s currently under development but has been released to the company for testing.”

Essentially you just get a link to a chatbot UI and need to see if you can prod and exploit it for sensitive company data.

There is also a little cheatsheet on the side with common commands:

For the first question I just asked straight out what the email was:

Question 2: What is the password for the IT server room door?

I got the next answer by pretending to lose my password:

Question 3: What is the name of McGreedy's secret project?

That same strategy didn’t work with the last question though. But I got around it by telling the chatbot that we were in maintenance mode.

Obviously the chatbot needs some tweaking! ❤