Phishing remains one of the most reported cybercrimes, and attackers now target hotels, banks and their customers with messages that look real. Develust Technology has released phywall, a free open-source toolkit that helps mail administrators running Rspamd and Dovecot detect phishing, clean existing mailboxes and train their filters, without sending any data unless they approve it.
Baku, November 2026 — Develust Technology has released Phywall, a free, open-source phishing defense toolkit for mail servers running Rspamd and Dovecot. The tool was built on production mail servers that Develust manages for hotels and other businesses, after the team saw how many convincing phishing emails were still reaching staff mailboxes despite standard spam filtering.
The numbers explain why the problem matters. According to the FBI's Internet Crime Complaint Center (IC3), the center received 1,008,597 cybercrime complaints in 2025, the first time the figure passed one million, and reported losses reached a record $20.877 billion, up 26% from the previous year. Phishing and spoofing was again among the most frequently reported crimes, making up nearly one in five complaints.
Business email compromise (BEC), a scheme in which attackers impersonate executives, suppliers or partners by email to redirect payments, caused about $3 billion in reported losses from almost 25,000 complaints. 2025 was also the first year IC3 tracked artificial intelligence in complaints: more than 22,000 reports mentioned AI, with $893 million in losses.
These figures cover only incidents reported to the FBI in the United States. The real global damage is higher, because many victims never report what happened.
Phishing succeeds because it targets people rather than software. Verizon's 2025 Data Breach Investigations Report, based on more than 22,000 security incidents and 12,195 confirmed breaches, found that a human element, such as clicking a malicious link or being manipulated by a fake message, was involved in about 60% of breaches.
Modern phishing emails rarely contain the spelling mistakes people learned to look for. They copy the design and tone of real brands, refer to real orders or reservations, and create pressure: an account will be suspended, a booking will be cancelled, a payment has failed. Generative AI makes it easier to write fluent messages in any language, which removes another warning sign.
The hospitality industry has become one of the clearest examples. Security researchers and consumer groups have documented a pattern that repeats across countries:
Because these messages arrive through a trusted channel and contain correct names and dates, they are far more convincing than ordinary scams. DomainTools reported that one attacker has generated nearly 1,000 spoofed booking and hotel reservation domains since May 2025, and researchers at Gen have named the trend the "Reservation Hijack Scam". Booking.com itself says it never asks customers to share payment information by email, chat, text message or phone.
For a hotel, the first step of this chain is almost always the same: one phishing email that reaches one employee's inbox.
The damage from phishing goes well beyond the first victim:
Most phishing messages follow the same pattern, which is exactly what filters can learn to recognize:
A single element on its own proves little: legitimate companies also send urgent emails and links. It is the combination that gives phishing away, and that idea is at the heart of phywall.
phywall adds a focused phishing layer on top of Rspamd, the open-source spam filtering system used on many Linux mail servers. It installs curated phishing phrase lists for 12 sectors, from hotels and travel to finance, healthcare and government, with more than 1,100 phrases in 16 languages, together with phishing URL feeds from OpenPhish and PhishTank.
Each signal on its own scores zero. Points are added only when signals appear together:
PHISH_URL link from a phishing URL feed 0 points
PHISH_TEXT phishing phrase in the body 0 points
PHISH_SUBJECT phishing phrase in the subject 0 points
PHISH_LINK_AND_TEXT PHISH_URL + (PHISH_TEXT or PHISH_SUBJECT) +6
PHISH_DEFAULT_PLUS_CUSTOM Rspamd PHISHING + any phywall signal +3This design keeps false positives low: a legitimate newsletter with an urgent subject, or a sender who shares a link, is not punished. A message that combines a known phishing link with typical phishing language is.
phywall can also scan mailboxes that already contain mail. Phishing found in the Inbox is moved to Junk, while the Sent folder is checked in report-only mode, which helps administrators spot phishing that staff forwarded to colleagues. A separate command trains Rspamd's Bayes filter using the server's own mail, learning ham only from Sent folders and spam only from Junk.
phywall is a single command-line tool written in Bash. Every command previews its changes first:
# See the available sectors
phywall configure --sectors
# Preview, then apply the Rspamd rules for your sectors
phywall configure --sector hotel,finance
phywall configure --sector hotel,finance --apply
# Check a mailbox, then move phishing to Junk
phywall scan [email protected]
phywall scan [email protected] --move
# Train the Bayes filter from your own mail (Sent = ham, Junk = spam)
phywall learn example.com --days 30
phywall learn example.com --days 30 --learnConfiguration changes are written as clearly marked blocks, a backup is taken first, and phywall configure --rollback removes everything phywall added.
Phishing campaigns change quickly, so phywall includes an optional way for administrators to send anonymized samples back to Develust and improve the phrase lists for everyone. Nothing is sent automatically: phywall contribute shows a preview, and only phywall contribute --apply sends it. Recipient addresses and the server's own domains are replaced with placeholders, links are reduced to host names, long numbers are masked, and no hostname, IP address or mailbox name ever leaves the server.
phywall is one layer of defense, not a complete solution. Mail administrators should also publish SPF, DKIM and DMARC records, enable multi-factor authentication on email and booking accounts, keep staff informed about current scams, and give users an easy way to report suspicious messages.
For everyone else, a few habits block most attacks: be suspicious of any message that creates urgency, check the real address of a link before entering a password, never send card details in reply to an email or chat message, and confirm unusual requests through a channel you already trust.
phywall is available now on GitHub under the MIT license. It has been tested on AlmaLinux 9 with Rspamd 4.2.2 and requires Rspamd and Dovecot with Maildir storage. Administrators can report false positives and missed phishing through GitHub issue templates, and security vulnerabilities can be reported privately.
Project page: github.com/develust/phywall. Full documentation is available on the phywall tool page on develust.com.