From the course: Certified Ethical Hacker (CEH)

Unlock this course with a free trial

Join today to access over 24,700 courses taught by industry experts.

Exploring model theft attacks

Exploring model theft attacks

From the course: Certified Ethical Hacker (CEH)

Start my 1-month free trial Buy for my team

Exploring model theft attacks

“

- [Instructor] Now let's discuss the model theft attack or the unauthorized access and exfiltration or, you know, stealing LLM model architectural components, which definitely introduces significant threats from malicious actors or advanced persistent threats, you know, or APTs within, you know, these LLM model implementations, right, so. This is particularly applicable to proprietary LLM models, right, where they're closed, not open weight, not like the models you find in Hugging Face or any other open source platform. This is where an attacker can potentially manipulate, you know, the model through prompts and any other methods to then copy or extract the weights and the parameters of those closed source models, right? And this can have different consequences, right? Now, the impact of LLM model theft, you know, can include things like economic loss, you know, the damage to the brand of a company. Let's say to against OpenAI or Anthropic or Microsoft or, you know, whoever is…

Contents