Content based filtering

Content filtering is often hard to explain to people, and I’m not sure I’ve yet come up with a good way to explain it.
A lot of people think content reputation is about specific words in the message. The traditional content explanation is that words like “Free” or too many exclamation points in the subject line are bad and will be filtered. But it’s not the words that are the issue it’s that the words are often found in spam. These days filters are a lot smarter than to just look at individual words, they look at the overall context of the message.
ISP_tolerances
Even when we’re talking content filters, the content is just a way to identify mail that might cause problems. Those problems are evaluated the same way IP reputation is measured: complaints, engagement, bad addresses. But there’s a lot more to content filtering than just the engagement piece. What else is part of content evaluation?

  1. Does the mail have hashbusters? Hashbusters are blocks of text, sometimes invisible to the recipient, that are put in an email in order to break some types of filtering. Ways to hide text include in HTML comments and by making foreground and background text the same color.
  2. Does the mail have valid HTML? Spammers have frequently used invalid HTML tags as a way to avoid filters by breaking up content or as hashbusters.
  3. Does this mail contain malicious content? These filters look for virus signatures or code that may compromise a recipient’s computer. Very few legitimate mailers have mail caught in virus filters, but every incoming mail is still  scanned for viruses or malicious code.
  4. Does this mail look like a phish? These filters look at the domains and authentication, but also look for common words and tricks phishers use. This filter is most likely to catch legitimate mail using tracking links with different URL content in the text portion of the HTML. An example of this kind of trigger is <a href=”http://tracking.example.com/login.html”>http://paypal.com</a>. Making sure there aren’t URLs, email addresses or hostnames in the text portion of a link generally avoids this kind of filter.
  5. Is this an industry with a bad reputation? The most obvious examples here are payday loans. There are so many horrible players in the online payday loan industry that it doesn’t really matter how good or clean individual mailers are. Payday loans are filtered heavily. Stock and financial messages also have challenges because there are so many pump-n-dump spammers out there.

Changing content can cause an improvement in delivery. But if that content was flagged because of user complaints or bad recipient profiles, the content filters will catch up. Continuing to attempt to evade filters by changing content can result in IP based filtering.
These are just a few of the things companies look at when evaluating content.
 

Related Posts

Looking for message labs help?

There’s a common bounce error from the Message Labs’ filtering appliance that goes no where.

Read More

Abuse it and lose it

Last week I blogged about the changes at ISPs that make “ISP Relations” harder for many senders. But it’s not just ISPs that are making it a little more difficult to get answers to questions, some spam filtering companies are pulling back on offering support to senders.
For instance, Cloudmark sent out an email to some ESPs late last week informing them that Cloudmark was changing their sender support policies. It’s not that they’re overwhelmed with delisting requests, but rather that many ESPs are asking for specific data about why the mail was blocked. In December, Spamcop informed some ESPs that they would stop providing data to those ESPs about specific blocks and spam trap hits.
These decisions make it harder for ESPs to identify specific customers and lists causing them to get blocked. But I understand why the filtering companies have had to take such a radical step.
Support for senders by filtering companies is a side issue. Their customers are the users of the filtering service and support teams are there to help paying customers. Many of the folks at the filtering companies are good people, though, and they’re willing to help blocked senders and ESPs to figure out the problem.
For them, providing information that helps a company clean up is a win. If an ESP has a spamming customer and the information from the filtering company is helping the ESP force the customer to stop spamming that’s a win and that’s why the filtering companies started providing that data to ESPs.
Unfortunately, there are people who take advantage of the filtering companies. I have dozens of stories about how people are taking advantage of the filtering companies. I won’t share specifics, but the summary is that some people and ESPs ask for the same data over and over and over again. The filtering company rep, in an effort to be helpful and improve the overall email ecosystem, answers their questions and sends the data. In some cases, the ESP acts on the data, the mail stream improves and everyone is happy (except maybe the spammer). In other cases, though, the filtering company sees no change in the mail stream. All the filtering company person gets is yet another request for the same data they sent yesterday.
Repetition is tedious. Repetition is frustrating. Repetition is disheartening. Repetition is annoying.
What we’re seeing from both Spamcop and Cloudmark is the logical result from their reps being tired of dealing with ESPs that aren’t visibly fixing their customer spam problems. Both companies are sending some ESPs to the back of the line when it comes to handling information requests, whether or not those ESPs have actually been part of the problem previously.
The Cloudmark letter makes it clear what they’re frustrated about.

Read More

Email filters

What makes the best email filter? There isn’t really a single answer to that question. Different people and different organizations have different tolerances for how false positives versus false negatives. For instance, we’re quite sensitive to false positives here, so we run extremely conservative filtering and don’t block very much at the MTA level. Other people I know are very sensitive to false negatives and run more aggressive filtering and block quite a bit of mail at the MTA level.
For the major ISPs, the people who plan, approve, design and monitor the filters usually want to maximize customer happiness. They want to deliver as much real mail as possible while blocking as much bad mail. Blocking real mail and letting through bad mail both result in unhappy customers and increase the ISP’s costs, either through customer churn or through support calls. And this is a process, filters are not static. ISPs roll out new filters all the time, sometimes they are an improvement and sometimes they’re not. When they’re not, they’re pulled out of production. This works both for positive filters like Return Path and negative filters like blocklists.
Then there is mail filtering that doesn’t have to do with spam. Business filters, for instance, often block non-business mail. Permission of the recipient often isn’t even a factor. Companies don’t often go out of their way to block personal mail, but if personal mail gets blocked (say the vacation plane ticket or the amazon receipt) they don’t often unblock it. But when you think about why a business provides email, it makes perfect sense. The business provides email to further its own business goals. Some personal usage is usually OK, but if someone notices and blocks personal email then it’s unlikely the business will unblock it, even if the employee opted in.
In the case of email filters, the free market does work. Different ISPs filter mail differently. Some people love Gmail’s filters. Other people think Hotmail has the best filtering. There are different standards for filtering, and that makes email stronger and more robust. Consumers have choices in their mail provider and spamfiltering.

Read More