phishing_detection: the web_traffic feature calls the retired data.alexa.com and marks every site as -1 #1822

Cybersecurity_Tools/phishing_detection/features_extraction.py, check_web_traffic(), requests https://data.alexa.com/data?cli=10&url=.... Amazon retired Alexa Internet in May 2022; today the host does not accept a connection at all, so the request raises and every URL gets web_traffic = -1.

The training set (Phising_Training_Dataset.csv) has three values for this feature: 1 (4,739 rows), 0 (2,090) and -1 (2,126), i.e. the usual definition of the UCI phishing dataset: 1 = rank up to 100,000, 0 = ranked but lower, -1 = no traffic rank. The model therefore always sees the "no traffic" value now, which pushes legitimate sites towards the phishing class.

Fix: take the rank from a ranking that still exists. The free NetAPI Top 1M has a JSON method for it, no API key needed (https://netapi.com/api-json/?method=domain-rank&domain=example.com answers rank or null), which maps 1:1 onto the three values. Pull request follows.

Disclosure: I run NetAPI.

Slava