Very attentive readers of our blog will surely be able to count how many times we have written about SPAM or even warned against it. We have provided clear instructions, tightened spam filters, and introduced new mechanisms time and again. We’ve taken up this challenge once more.
Both you, as email users, and we – admins and consultants – know how spam or phishing can be... annoying, and more often even dangerous. It's safe to say that we now live in the age of artificial intelligence, which is being used by spammers to ensure their emails no longer look like "person must send money" but instead resemble messages written by your closest friend, sister, or colleague. Deceptive!
That’s why we decided – fight AI with AI!😊
The email service at WEBD.pl has undergone a complete overhaul of its mail filter. Thanks to the implementation of cutting-edge technologies, including advanced language models based on artificial intelligence, the mail filter now has new capabilities to detect suspicious messages. Nothing gets past us!
This comprehensive modernization has allowed us to refine algorithms and rules, making us more precise and effective in identifying even the most sophisticated fraud attempts. Our goal is not just to eliminate spam but also to accurately distinguish between spam and legitimate messages, ensuring that you only receive relevant information without the risk of losing important emails. And that’s where the real challenge lies.
How does it work?
Suspicious messages are forwarded to our LLM system based on artificial intelligence, where they undergo a detailed analysis. The system evaluates them for potential threats such as phishing or spam, using advanced algorithms to identify suspicious content, links, and even the sender's intent.
If a potentially dangerous message is detected—one that might contain phishing elements but isn’t entirely clear-cut—our system marks it with a **PHISHING ALERT**. This way, you are properly informed about the risk and can make a conscious decision while remaining cautious before opening the message or clicking on any suspicious links. This additional warning enhances security without disrupting the flow of important information.
Such an email will look like this:
We operate locally!
Our system, powered by a local language model (LLM), works exclusively within our closed server infrastructure. This means that all data is processed on-site by advanced artificial intelligence, without the need to send it to external services or cloud platforms.
Because all operations are carried out within our infrastructure, user data never leaves the system, completely eliminating the risk of it leaking to the web or external entities. By using a local LLM, we achieve effective message filtering while ensuring maximum protection of user data.
What’s next?
We are continuously working on improving how spam is categorized so that our system can keep up with the ever-evolving methods of attacks. In the near future, you’ll notice a significant improvement in the effectiveness of message filtering. We’re already seeing the results—here’s a fresh statistic from the past few days: on average, 200 emails per hour are flagged by the AI as phishing attempts.
As you can see, the scale of this issue is substantial. On a side note, we’ve had instances where over 700,000 SPAM emails were blocked in a single day!
If you’d like to contribute to the fight against SPAM once and for all (or at least for a while), we encourage you to forward suspicious messages (using the forward option) to our dedicated email address: spam@webd.pl.
The end. 🎭
