Certified AI/ML Pentester (C-AI/MLPen)

Testimonials

I was really impressed with the exam. The 4 hour long, practical (ctf-style), online exam thoroughly tested me on common LLM vulnerabilities. The labs were carefully designed and the challenges mimicked real world scenarios. From direct/in-direct prompt injection test cases to the RAG poisoning challenge, it had a bit of everything. Including bypassing system prompt firewalls of a 𝘋𝘐𝘍𝘍𝘐𝘊𝘜𝘓𝘛 degree.

The SecOps Group has crafted a rigorous challenge that required extensive research. I passed the exam, although I fell short by 2 out of 8 questions. Congratulations to The SecOps Group Team for creating a certification that not only made me study and research intensively but also highlighted the current content gap in this field. I believe they are pioneers in offering such a test!

So I got asked by Sumit Siddharth to check out their certification on AI/ML Pentest. I would highly recommend it. I can't wait to see more of their contents and testing. The test itself is quite fun and hands on. It's not a "bubble in your answer" but a hands on, hacking test. I definitely think The SecOps Group got this right in an emerging space. I finished it and got the certificate with merit.

I am happy to share, that I have obtained the Certified AI/ML Penetration Tester with Merit from The SecOps Group. The exam was pretty challenging and took a real test of my skill. Having solved a few CTFs around prompt Injection I would say this is the best and most unique set of problems I have encountered. Highly recommend who want to delve into the world of AI Security.

What is the Certified AI/ML Pentester (C-AI/MLPen) exam?

The Certified AI/ML Pentester (C-AI/MLPen) is an intermediate-level exam designed to test a candidate’s knowledge of the core concepts involving AI/ML security. To pass this exam, candidates must demonstrate practical knowledge and the ability to conduct an AI/ML pentest.

Who should take this exam?

C-AI/MLPen is intended to be taken by pentesters, application security architects, SOC analysts, red and blue team members, AI/ML engineers and any AI/ML security enthusiast, who wants to evaluate and advance their knowledge.

What is the format of the exam?

C-AI/MLPen is an intense 4 hour long practical exam. It requires candidates to solve a number of challenges, identify and exploit various vulnerabilities and obtain flags. The exam can be taken online, anytime (on-demand) and from anywhere. Candidates will need to connect to the exam VPN server to access the vulnerable applications.

What is the pass criteria for the exam?

The pass criteria are as follows:

Candidates scoring over 60% marks will be deemed to have successfully passed the exam.
Candidates scoring over 75% marks will be deemed to have passed with a merit.

What is the experience needed to take the exam?

This is an intermediate-level exam. Candidates should have prior knowledge and experience of AI/ML pentesting. They should have a solid understanding of common application security topics, including the OWASP Top 10 vulnerabilities for large language models (LLMs), prompt injection attacks, common security misconfigurations, and best security practices. They should be able to demonstrate their practical knowledge on AI/ML security topics by completing a series of tasks on identifying and exploiting vulnerabilities that have been created in the exam environment to mimic the real world scenarios.

Note: As this is an intermediate-level exam, a minimum of one year of professional pentesting experience is recommended.

What will the candidates get?

On completing the exam, each candidate will receive:

A certificate with their pass/fail and merit status.
The certificate will contain a code/QR link, which can be used by anyone to validate the certificate.

What is the exam retake policy?

Candidates, who fail the exam, are allowed 1 free exam retake within the exam fees.

What are the benefits of this exam?

The exam will allow candidates to demonstrate their skills in AL/ML pentesting. This will help them to advance in their career.

How long is the certificate valid for?

The certificate does not have an expiration date. However, the passing certificate will mention the details of the exam such as the exam version and the date. As the exam is updated over time, candidates should retake the newer version as per their convenience.

Will you provide any training that can be taken prior to the exams?

Being an independent certifying authority, we (The SecOps Group) do not provide any training for the exam. Candidates should carefully go over each topic listed in the syllabus and make sure they have adequate understanding, required experience and practical knowledge of these topics. Further, the following independent resources can be utilized to prepare for the exam.

Company	Free/Paid
Portswigger Web Security Academy	Free
Gandalf	Free
IBM	Free
Learn Prompting	Free
LLM Security	Free
OWASP	Free
AI Village	Free
Promptingguide	Free
Promptingguide RAG	Free
Cobalt	Free
Bugcrowd	Free
Unite AI	Free
Simonwillison	Free
Vickieli	Free
NCC Group	Free
WithSecureLabs	Free
ScottLogic	Free
Greshake	Free
Hannibal046	Free
Ottosulin	Free
Mik0w	Free
ATLAS Matrix	Free
Vulnerable LLM Applications	Free
Awesome-llm-security	Free
Prompt Airlines	Free
Crucible	Free
Immersive Labs	Free
Bugcrowd Ultimate Guide AI Security	Free
AI Red Teaming	Free
The Ultimate Guide to Managing Ethical and Security Risks in AI	Free
NVIDIA AI Red Team: An Introduction	Free
Lakera - Real World LLM Exploits	Free
SpyLogic Prompt Injection Attack Playground	Free
Offensive ML Playbook	Free
Snyk OWASP top 10 LLM	Free
Prompt Injection Games from Secdim	Free
Large Language Model (LLM) Pentesting	Free
LLM Pentest: Leveraging Agent Integration for RCE	Free

Exam Syllabus

The exam will cover the following topics

Prompt Injection

Direct Prompt Injections
Indirect Prompt Injections

Insecure Output Handling

Training Data Poisoning

Supply Chain Vulnerabilities

Traditional third-party package vulnerabilities, including outdated or deprecated components.
Using a vulnerable pre-trained model for fine-tuning.
Using outdated or deprecated models that are no longer maintained leads to security issues.
Use of poisoned crowd-sourced data for training.

Sensitive Information Disclosure

Incomplete or improper filtering of sensitive information in the LLM responses.
Overfitting or memorization of sensitive data in the LLM training process.
Unintended disclosure of confidential information due to LLM misinterpretation, lack of data scrubbing methods or errors.

Insecure Plugin Design

Excessive Agency

Excessive Functionality.
Excessive Permissions.
Excessive Autonomy.

Overreliance

Model Theft

System Prompt Leakage

Sample Question?

Evaluate the application’s anti-reversing checks. Which of the following statements is true in the context of the Jailbreak Detection?

1. The Application has a robust Jailbreak Detection implementation.
2. The Application lacks any Jailbreak Detection implementation.
3. The Application has insufficient Jailbreak Detection, which can be bypassed.
4. The Application has implemented Jailbreak Detection, and it can not be bypassed.

Prerequisites:

Host Operating System:

Windows/Linux/MacOS with minimum 8GB RAM (MacOS Preferred).

Physical Device with Minimum iOS Version Supported:

iOS 14 or higher (Jailbroken).

Note: Please make sure you have your iOS pentesting environment ready (Jailbroken Physical Device, Burp Suite, or any similar proxy tool along with Frida, objection and other similar pentesting tools) prior to starting the exam. The IPA build will be distributed via TestFlight.