Ask HN: Have you noticed an increase of spam evading Gmail's filters recently? | Hacker News Reader

Ask HN: Have you noticed an increase of spam evading Gmail's filters recently?

34 pointssupermatou16 comments

I've had a Gmail account since the service was in its early beta and you could get one by invitation only.

That was nearly 20 years ago. Over the years, the number of messages in my Spam folder grew constantly - last year, for example, there were some 1,000 messages there, waiting for auto-deletion. I didn't care, as I very seldom saw one in my inbox.

Something must have changed in the past months. Now I'm getting 3-4 messages per day and those messages are very obvious spams, that even a primitive mid-Nineties filter would have caught. I keep reporting them as spam, but nothing changes - I get the same messages the next day. Even more infuriating, they seem to be emanating from the same bot, as they arrive in my inbox between at ~10am PDT, ~12pm PDT and ~3pm PDT each and every day.

How is is possible for Google's excellent (until now) spam fighting system to miss out on those ones?! I'm reluctant to copy'n'paste some fragments here, as they're so spammy that HN's own filter would be triggered and my message would go straight to HN's purgatory - but, please, believe me when I say they're in the spammiest format one can imagine.

Any explanation?

"On June 21, Google parent Alphabet Inc. floated a plan to fix a problem it had identified in its spam filters. The solution was specifically aimed at political spam, the kind likely to pour in during an especially noisy midterm election season. Unfortunately for all of us, Google seems to have decided the problem is that it filters too much spam—and the solution is to let politicians run wild with polling updates, merch solicitations, and frantic fundraising pleas."

https://www.bloomberg.com/news/articles/2022-07-05/gmail-spa...

No moreso than usual, but it comes and goes almost seasonally.

This particular Ask HN comes up almost monthly:

https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...

I've experienced the same. Here's a larger collection of obvious spam failing to get flagged by Gmail's spam filters that we gathered: https://www.surgehq.ai/blog/are-the-spammers-winning-failure...

For example, a spammy email that is strangely uncaught:

Subject: ///Y0urFREEFlashlight(NeedYourAddress)///1851

Sender: UBzjwbFb-lyZSC3-noReply@gwhsi.lairpro.com

Within the past year (I don't recall how many months ago but at least for a few months) I started getting more that made it into the inbox: spam with subject lines similar to 'StUfF L1k3 Th!s_00'

Things that as a layperson, I would think wouldn't be too difficult to make an algorithm to recognize.

> Things that as a layperson, I would think wouldn't be too difficult to make an algorithm to recognize.

Also, they have the same subjects, day after day: "notification", "coupon", "winner", pretending to be "Costco", "some-insurance-company" and "Lowe's". Day after day.

I have, yes. I've also had an account since invite-times, and the spam filtering was the main feature that has kept me from migrating off.

I now get terrible, obvious-spam messages. Huge images to evade text filtering, subject lines with all special characters, the classic hallmarks. OCR on the images would make it obvious--not that I want Google OCR'ing images in emails.

I've been getting them for at least 6 months now, maybe as long as a year.

I will say that it peaked for a while, and lately has improved again.

Can only speculate -- but perhaps they continually train some spam detecting ML model and swap them out every few weeks. Wondering if there are some hidden regressions or something to that effect which are evading the new model. Wouldn't surprise me if they originally started with a rule-based engine and are A/B testing ML models for it instead now.

Today GMail added a large "this message maybe is a spam, please tell us" banner to my mom's messages, halving my viewport, and I have never see this banner before. Come on Google, we have hundreds of dialogues in years, how dare you to ask me some extra clicks.

I've been consistently getting 3 or 4 repeat messages that evade spam filters. They all seem to come from 'life' TLD domains. Marking them seems to have little effect as very similar message will arrive straight to my inbox, but a new domain is used.

I wish I could stop winning makita drills.

I have noticed this also. I have a few friends at Google who I pointed this out to, so let's see what happens. If anyone else has friends at Google, be sure to pester them about it.

This strategy of "Pester friends at Google" instead of them having legitimate customer feedback avenues seems odd to me. Every few days we see something rise to the top of HN and other social media sites about areas where Google/YouTube is doing problematic things and every time we "solve" it by having people pester their Google friends.

Welcome to the Dystopian Algorithmic Future.... where your only solution to a pain point is to go viral on HN and get enough people who can actually bug the people who manage the algo to fix your issue.

I agree. Further, in this case it did not work: one of my friends at Google said that even if he reported it, he doubts they would do anything about it. What are we going to do about it though?

For me, it's the opposite. Recently more and more not-spam messages land in the spam folder. Mostly notifications from Github.

The last few weeks have been really bad. I get 3-4 obvious spam emails a week now where I would maybe get 1 every 6 months prior.

I was surprised to notice something similar last month, for the first time in memory.

Reply on news.ycombinator.com