EZ Arena
PrototypeLearn by doing, compete, and level up
EZ Arena brings together gamified learning challenges - ML scoring, crossword puzzles, cyber awareness assessments, and live events. Every activity reinforces the course material and earns you points on the leaderboard.
This is a new, actively evolving part of the platform. If something looks off, it is being worked on. Check back soon or let us know on EZ Forum.
Unit 1 · Introduction to ML in Cybersecurity
Malicious URL Detection
Train a classifier on 200 labeled URLs, then flag malicious vs benign among 100 you have never seen. The lexical signal is in the URL string, not the weak reputation score.
Start challengeUnit 2 · Supervised Learning
Phishing Threshold Tuning
Train a classifier on 200 labeled emails, then predict phishing vs legitimate on 100 you have never seen. Scored against the real hidden answers, not a slider you can eyeball.
Start challengeUnit 2 · Supervised Learning for Cybersecurity
Malware Detection from PE Features
Train a classifier on 200 labeled Windows executables, then predict malware vs benign for 100 you have never seen, using static features like entropy, imports, and signing status.
Start challengeUnit 3 · Unsupervised Learning for Anomaly Detection
Network Intrusion Anomaly Threshold
Train an anomaly detector on 200 labeled network events, then flag intrusions among 100 you have never seen. Scored against the real hidden answers, not a slider you can eyeball.
Start challengeUnit 3 · Unsupervised Learning and Anomaly Detection
Credit Card Fraud Detection
Fraud is rare, which is the whole challenge. Train on 200 labeled transactions, then catch the handful of frauds among 100 you have never seen without drowning in false alarms.
Start challengeUnit 4 · Reinforcement Learning in Cybersecurity
Adaptive Firewall Defense
For every connection, a defender must choose: block or allow. Learn that defensive policy from 200 labeled connections, then decide the right action for 100 you have never seen.
Start challengeUnit 5 · Malware Detection and Feature Engineering
DGA Botnet Domain Detection
Botnets generate thousands of random domains for command-and-control. Train on 200 labeled domains, then spot the algorithmically generated ones among 100 you have never seen.
Start challengeUnit 6 · Model Evaluation and Adversarial Attacks
Adversarial Spam Robustness
Spammers swap letters for lookalikes (FR33 iPh0ne) to dodge filters. Train on 200 labeled messages, then stay accurate on 100 you have never seen, including obfuscated ones.
Start challenge