HackTheBox - Blurry
Source: YouTube · IppSec · published Oct 12, 2024 · 56:48
The video demonstrates a security assessment of a company using Clear ML, focusing on exploiting a pickle deserialization vulnerability to gain unauthorized access to their AI development artifacts.
Key Takeaways:
• The target environment involves a company experimenting with AI/ML tools, specifically Clear ML, which is described as an open-source platform for streamlining AI adoption 0:08.
• The attacker identifies a known vulnerability in Clear ML related to pickle deserialization when retrieving artifacts, allowing for code execution 0:23.
• By analyzing user behavior, the attacker discovers a script that retrieves artifacts with specific tags, providing a vector for injection 0:28.
• A malicious artifact containing a pickle gadget is crafted to trigger a reverse shell, granting the attacker control over the system 0:36.
This case highlights the critical risks associated with insecure handling of serialized objects in machine learning pipelines.
Sources:
Generate CPE Credits
Generate a professional CPE document from this video's transcript.
Estimated credit: 1 CPE hours
Estimate uses the video runtime (1 hour ≈ 1 CPE, rounded to the nearest 0.5, minimum 0.5, maximum 2.0). The final amount can be lower after review, never higher.
Topic: Penetration Testing. Commonly maps to: Security Assessment and Testing, Security Operations. Exact CISSP domains are assigned during generation.
Free account. One generation at a time, with a daily limit.
CPEBuddy is independent and not affiliated with or endorsed by ISC2, ISACA, or any certification body. Exports are formatted for common CPE submissions; acceptance is at your certification body's discretion.
Transcript Preview
First 800 characters of the transcript
what's going on YouTube this is ipag I'm doing blurry from hack the box which I don't know how the Box got its name but it's all about attacking a company that's just starting to dabble in Ai and machine learning they have a domain that runs clear ML and I don't know what this is really either but according to Google it's an open source end AI platform designed to streamline AI adoption and the entire development life cycle looking at Google for vulnerabilities there is one for pickle derealization on retrieving artifacts and if you look around you can see that one of the users is is retrieving every artifact that's running on blurry with a specific tag and in a project so you can create a malicious artifact using a pickle Gadget to send you a reverse shell and then within a minute the use…