Phishing is one of the biggest cybersecurity threats today: attackers send emails that look legitimate in order to trick users into revealing sensitive information.
Your challenge is to develop an AI system that classifies emails as either Phishing (fraudulent) or Legitimate (safe), while also providing explanations for its decisions.
You will receive:
- A public training set of labeled emails .
- A public validation set for local testing.
- At the end, your solution will be evaluated on a hidden test set for the official leaderboard.
WHEN: on the November 13th 10:30 – November 14th 14:00
PREVIOUS EXPERIENCE
- Basic knowledge of Python programming.
- Familiarity with machine learning or NLP techniques is helpful but not mandatory.
- Participants should be comfortable working with JSONL datasets and building scripts that handle text data.
Goal
Goal of the Competition
Can you outsmart phishing?
Build an AI system that detects fraudulent emails and explains its decisions. Train on public data, compete on hidden tests, and climb the leaderboard — only the most accurate and efficient solution wins!
Rules
Rules of Engagement
TBD
Scoring
Scoring
- Primary metric: MCC (Matthews Correlation Coefficient).
- Tie-breakers: F1-macro → Balanced Accuracy → runtime per email.
- No subjective judging: the best score on the hidden test set wins.
PRIZES
TBD
REGISTRATION
- You need to have an account on CyberEDU. Register here or login here.
- Authenticate in your account and click on this link (this is required only the first time). This will give you access to a private space called “DefCamp”. See the picture below.
- Go to section Compete as instructed. You should see all the active contests.
- Click on any contest of interest and make the Pre-Registration.
- Have fun!