### 🧮 Evaluation

Submissions are evaluated based on the **Area Under the Receiver Operating Characteristic Curve (ROC AUC)** between your predicted probabilities and the actual `attack_detected` target.

This metric ensures a balance between sensitivity (recall) and specificity, helping evaluate how well your model distinguishes between safe and malicious activity.

Your final submission should be a CSV file with two columns: `ID` and `TARGET`. Each row represents a test instance with a probability score for the attack detection.

Example format:

```
ID,TARGET  
2,0.0034  
5,0.7823  
6,0.1890  
...
```

* `ID`: Unique identifier for each test sample.
* `TARGET`: Predicted probability that the given activity is a **cyber attack** (1 = high probability of attack, 0 = safe).

> Note: Ensure that all values in the `TARGET` column are floating-point numbers between 0 and 1.