Fraud Reports: Aggregate and Rank User Fraud Reports Within a Time Window

Question

## Problem

You are given a list of fraud reports, each with a `reporter_id`, `reported_user_id`, and `timestamp`. Find all users with more than `threshold` unique reporters within any rolling 24-hour window.

Return a list of `(reported_user_id, max_unique_reporters_in_any_window)` sorted by count descending, then by user id ascending.

```python
from typing import List, Tuple

def flagged_users(
    reports: List[Tuple[str, str, int]],
    threshold: int
) -> List[Tuple[str, int]]:
    # reports: [(reporter_id, reported_user_id, timestamp_sec), ...]
    # return: [(reported_user_id, max_unique_reporters), ...]
    pass
```

**Example:**
```
reports = [
  ("u1","bob",0), ("u2","bob",3600), ("u3","bob",7200),
  ("u1","bob",90000),  # outside 24h window of t=0
  ("u1","alice",0)
]
threshold = 2

bob: window [0,86400) has 3 unique reporters -> flagged (count=3)
alice: window [0,86400) has 1 -> not flagged

Output: [("bob", 3)]
```

## Follow-ups

1. What if a single reporter can submit multiple reports and they should only count once per window?
2. How do you scale this to process 10M reports/day in near real-time?
3. How would you prevent false positives from coordinated mass-reporting attacks by a single group?
4. What additional signals would you combine with report count to improve fraud detection accuracy?

Fraud Reports: Aggregate and Rank User Fraud Reports Within a Time Window

Question Details

Full Details

Topics

More from Stripe