Spammers use invisible Unicode text to bypass email filters

In short: Spammers are now using “ASCII smuggling,” a way to hide invisible text inside emails, to sneak past spam and phishing filters.

What's going on

ASCII smuggling first became known as a method for hiding “prompt injections,” which are secret instructions meant to trick AI assistants into doing something unsafe. The same hiding method is now showing up in regular spam emails.

The trick uses a special part of Unicode, which is the big catalog of characters used by computers for text in many languages. In this case, spammers use a set of Unicode “tag” characters that computers can read but people almost never see. It is like writing a message in ink that a machine can detect, while a human sees a normal sentence.

Microsoft says it saw a major spike in this activity earlier this year. Microsoft Defender for Office detections went from about 21,000 per day to more than 1.3 million in early February. Within four days, detections reached 2.5 million. The wave lasted for months, then dropped sharply in mid-May.

Spammers use the invisible characters to break up “trigger” words that filters look for, such as “credit,” “term,” or dollar amounts. For example, a message can look like it says “funding” to you, but a filter might read it as “fun” and “ding” because invisible characters are inserted in the middle.

What to watch

Email services will likely update their filters to spot these invisible characters, but spammers often move on to new tricks once old ones are blocked. Microsoft also noted that some modern filters rely on models that split text into pieces before checking it, and hidden characters can confuse that step. Expect more defensive updates, and more experiments by spammers to dodge them.

Source: Arstechnica