Machine Learning Security

Scoreboard · My Scores · Submit Flags · Take a Quiz

Canvas Outage

CNIT 430 students should have received
an email from sam.bowne@agentmail.to
showing how to take the final exam
without using Canvas.

Since Canvas is down and I cannot access quiz
scores, your grade will be based on projects
and the final exan only, assuming you got
20 on all the quizzes.

Here is the new grading system:

A: 271
B: 216
C: 107
D: 53
F: 52 or less

I am working on gathering all the extra
credit I sent to Canvas and adding it to
my new scoring system.

If you submitted projects in Canvas, or
have any other issues to discuss, please
email sbowne@ccsf.edu or come to one of my
Twitch sessions shown at the top of samsclass.info

  

Presentation: Securing AI Systems

KEY · PPTX· PDF

Understanding Prompts

ML 130: Prompt Injection (25 pts + 40)

Google Learning

GL_Badges: Google Learning (30 pts + 60 or more)

Security Risks

ML 150: OWASP Machine Learning Security Top Ten (15 pts)
ML 151: OWASP Top 10 for LLM Applications (15 pts)
ML 152: Microsoft Copilot Security (15 pts)

Awareness: Demonstrating Capabilities

ML 100: Machine Learning with TensorFlow (20 pts + 45)
ML 101: Computer Vision (10 pts)
ML 102: Breaking a CAPTCHA (10 pts)
ML 103: Deblurring Images (10 pts + 30)

Technical: Inner Components

ML 104: Analyzing Input Data (20 pts)
ML 105: Classification (15 pts + 10)
ML 112: Support Vector Machines (40 pts)
ML 113: Decision Trees (15 pts)
ML 114: Ensemble Learning and Random Forests (15 pts)
ML 115: Dimensionality Reduction (20 pts)
ML 116: k-Means Clustering (30 pts)

Attacks

ML 106: Data Poisoning (10 pts)
ML 107: Evasion Attack with SecML (15 pts + 25)
ML 108: Evasion Attack on MNIST dataset (20 pts + 20)
ML 109: Poisoning Labels with SecML (20 pts + 10)
ML 110: Poisoning by Gradients (15 pts + 15)
ML 111: Poisoning the MNIST dataset (20 pts + 20)

Defenses

ML 140: Deep Neural Rejection (45 pts)
ML 132: Llama Guard (10 pts)

Large Language Models

ML 120: Bloom LLM (15 pts + 15)
ML 121: Prompt Engineering Concepts (20 pts)
ML 122: Comparing LLMs on Colab (10 pts + 10)
ML 123: Running Llama 3 Locally (15 pts)
ML 124: Evaluating an LLM with Trulens (15 pts)
ML 126: Building RAGs (15 pts)
ML 127: Encoding Text with BERT (10 pts)
ML 128: Using AnythingLLM to Embed Custom Data (10 pts)
ML 129: Embedding Words with BERT (40 pts)

ML 125: Jupyter Notebook on a Mac M1 (10 pts)

Agents

ED 33: Ubuntu Linux Desktop (10 pts)
ML 135: OpenClaw on Ubuntu Linux (15 pts)
ML 136: OpenClaw: exec and Email (20 pts)
ML 137: OpenClaw: Web Browsing and Remote Access (20 pts)

ML 133: MCP with Claude (15 pts)
H 112: Cloudflare Quick Tunnel (25 pts)
ML 134: IronClaw on Ubuntu Linux (15 pts)

ML 190: Personal AI Infrastructure (PAI) (10 pts)
ML 200: Claude Code Agent on Ubuntu (10 pts)

Generating Code

ML 160: GitHub Copilot (15 pts)
ML 131: Generating Python Code with Gemini (40 pts)
Violent Python Challenges (extra)

Quantum Computing

C 510: Quantum Computing (20 pts)
ML 170: Modeling Chemical Reactions with ML and Quantum Computing (10 pts)

Kolmogorov-Arnold Networks (KANs)

ML 180: Fitting Polynomials to Data (30 pts)
ML 181: B-Splines for Kolmogorov-Arnold Networks (KANs) (15 pts)

Attack References

It’s disturbingly easy to trick AI into doing something deadly
GhostStripe attack haunts self-driving cars by making them ignore road signs
MadRadar hack can make self-driving cars 'hallucinate' imaginary vehicles and veer dangerously off course
Two big computer vision papers boost prospect of safer self-driving vehicles

References

SecML: Secure and Explainable Machine Learning in Python
ChatGPT Prompt Engineering for Developers
Prompt Engineering Guide
Google's Generative AI learning path
A jargon-free explanation of how AI large language models work

Pinecone Makes Accurate, Fast, Scalable Generative AI Accessible to Organizations Large and Small with Launch of its Serverless Vector Database

Pinecone Vector Database
Free Training Building Applications with Vector Databases

The Databricks Data Intelligence Platform
Attention in transformers, visually explained

  

Archives

Spring 2023 Scores
July 9, 2023 Scores
CCSF CyberCamp July 23, 2023 Scores
CCSF Adv. CyberCamp July, 2023 Scores
Videos from TX State Working Connections Summer 2023
Final Scores from DEF CON 31
Scores Archived 1-20-25
Scores Archived 5-15-25

Old Scoreboard · Submit Flags · Old Details

ML 170 added 8-2-24
Scoreboard re-initialized 1-20-25
ML 180 and 181 added 2-8-25
Scores archived 5-15-25
Moved to index4 12-6-25
ML 130 points changed 1-29-26
ML 132 and 133 added 2-12-26
ED 33, H 112, and ML 134 added 3-2-26
ML 135 and ML 136 added 3-7-26
ML 137 added 3-19-26
ML 190 and 200 added 4-16-26
New scoring system with quizzes added 5-10-26
Grading system added 5-12-26