Seven guards, six languages, and the payment keyword that would have refused every listing on Idealista

Описание и применение

Внедрение семи уровней проверки безопасности в системе автоматической отправки ответов на объявления о сдаче жилья в Европе позволяет избежать действий на платежных и входных страницах, используя многоязычные регулярные выражения и специфические маркеры платежных провайдеров.

Задача

Automating replies to rental listings risks acting on unsafe pages; a naive keyword check can block legitimate listings.

Результат

Implementing a layered guard system with language‑aware patterns and specific payment‑provider markers reduced false positives and prevented automated replies from being sent to payment or login pages.

Рабочий процесс

• 1. Evaluate page domain to ensure it belongs to the original site (domain lock).
• 2. Detect payment pages via PSP script hosts and card input markers, not generic keywords.
• 3. Check for anti‑bot walls and back off if present.
• 4. Verify session validity; refuse if expired.
• 5. Detect login or signup walls using URL, title, and body patterns across six languages.
• 6. Detect paid‑plan requirements on the site itself.
• 7. Confirm the page is a valid listing; otherwise refuse.

Доказательства

The first one, the domain lock, has its own post: it refuses to act on a page that has left the site the saved search belongs to, which is the aggregator hand-off case.

Источник

The payment guard does not look at payment words at all: a bare "ideal" keyword would match Idealista (a real listing site), so we key off PSP script hosts and card input attributes instead of payment-method words.

Источник

Three signals with three different levels of trust. The URL is the strongest, because a path segment of /login or /inloggen is not decoration.

Источник