Human-Powered Spam Filtering

arturs writes "A company called eProvisia started an unusal business: they filter out spam not by using complicated algorithms, but human beings... It costs around $20/year - is the war against spam over?" It's an interesting idea - the privacy concerns are big of course, but how would this stack up to, say SpamAssassin or a suite like Barracuda's Spam firewall. We tested the Barracuda device - great integration of OSS software, with a nice interface. Update: 09/20 15:12 GMT by J : Corrected price of Spam Eradicator.

Won't work

[seanyboy]

1. The only way that they'll be able to do this at a good cost is to use some kind of third world labour with a first language that isn't english. 2. Given that my baysian spam solution seems to be better at sorting spam than me (I've accidently deleted items which were not spam before), then I'm not entirely sure that a stranger could do better. They offer a 100% guarantee. I doubt they'll be able to offer a 100% service.

Re:Buzzword Bingo

[Nos.]

No kidding, that second paragraph (well, actually sentence) has high buzz word content yet actually only says, we do anti-spam. I have a feeling marketing and technical don't agree on much at this company.

You've been beaten to it!

[DaHat]

MailWasher Pro from Firetrust has an option similar to that known as FirstAlert, you sign up for a year membership and use their app to submit spam type messages to them, later, a human verifies that the e-mail is indeed spam and adds it to their database. Once added, any MWP users using FirstAlert will hit the database, see that the message is there and act accordingly (often times deleting without even showing it to the user). It works... ok, about 30% of my spam is nuked this way, the built in Bayesian filtering catches another 40% or so, and the DNS blacklists catch most of the rest. Of the last two groups I verify manually, but have come to trust FirstAlert... it's just a shame it's not getting the high %'s it used to.

Re:Buzzword Bingo

[garcia]

That second paragraph, in particular, would keep me as far away from them as possible.

And yet those are exactly the words that HR departments are looking for on resumes. As long as you can use stupid words that really don't work in any other situation you can get your foot in the door.

Business model?

[Empiric]

"24 hours a day" * 30 days/month = 720 hours

$20 per month / 720 hours = about 3 cents an hour.

Since they say they begin "manually reviewing, hand-picking and approving important correspondence", how does this work? To pay someone $6/hour, they'd need to be reviewing at least 200 mailboxes simultaneously. My confidence level of their accuracy under these circumstances would be considerably -lower- than a software solution.

hilarious

[alatesystems]

Look at this [dione.cc] about page.

Im going to include their footnotes on that pge in parentheses and bold.

Privately funded in 1993, now with customers in 40 countries(Not all currently recognized by UN) and over $67 million(Palmyra Atoll dollars) in cash reserves, the company experienced a phenomenal growth

Spamassassin works great

[Stone316]

if its setup properly and bayes is enabled. Since Aug. 1st i've recieved 1800 emails flagged as spam... A few false positives but I have my threshold set pretty low. About 140 emails have gotten through and soon i'll take that folder and process it. Before I started processing spam that wasn't caught by spamassassin about 4-5 were getting through a day.. Now its once every few days.

isn't this similar to what gmail does?

[laard]

from my understanding, all gmail users that click the little "report spam" button are essentially helping build the database and increase the effectiveness of its filters... though I guess they must be careful because this could potentially generate a lot of false positives.

Re:Yay

[Buzz_Litebeer]

There can be a huge advantage of looking over data manually.

My job is focused around looking at data to find problems. Many companies by high cost and very fancy data verification tools, address cassing software etc... Then they run these things on automatic on all of there data. What happens is that there data slowly becomes unreliable over time because they are paying their data entry a pittiance or are simply outsourceing it to a foreign nation who has people who do not speak the language, as much as spell it.

So, having someone like me who can clean the data (even sensitive data) to quickly check for errors, find common errors, and write novel solutions for them is a pretty big business. I deal with crap sent from data entry day after day and turn it into data that can be used, and is highly accurate.

So, I can definately see someone using a human as an intermediary between spam.

I could see this working this way. If the email comes from someone "trusted" in the company, it gets passed along without a human filter. Any email addresses that are not trusted would come by a physical person.

This way, if "Bob, your investment advisor" emails you, then that data would be sent directly to the company without human interference.

but, if a mail comes to [e-mail], then you would have a person scan that e-mail and verify its spammyness. This way, any spam that would get through would have to be from someone trusted. If someone is trusted, then they are not likely to be spam.

It could be a very good system, but I doubt companies would want to allow email to be looked at this way. Consumer responce data, address data, etc.. I can see, but not really e-mails unless there was a pressing need to stop it.

LLC Companies

[jeffs72]

Actually, LLC's are the smart way to do a partnership. In a partnership, all principles enjoy equal responsibility for mishaps. In LLC's, all principles enjoy shared responsibility.

I guess the best way to sum it up would be to quote my Business Legal Environment professor: "...and I hope that now you all have a clear understanding of partnerships. Now let me give you a word of advice, never form one."

GMail outcry

[TVC15]

And people were upset when it was 'discovered' that GMail was going to programatically 'read' your email to provide contextual advertising (and spam filtering) as an invasion of privacy? Here we will have actual _people_ reading your private correspondences. No thanks.

I'd be a bit suspicious

[J Mack Daddy]

A piece of software that is filtering your spam is not being paid to do it. Yes it's creators were perhaps paid, but at least they are one step removed. A paid human-based service doing spam-filtering, however, would have a direct motive to want as much spam flying around as possible, so as to have as many potential customers as possible. I wouldn't be surprised if this company was created by and is being funded by the spammers themselves as just one more way to make money from spam! .

Filtering using spelling checker...

[baywulf]

Why can't we use spelling checkers to filter spam? It seems all the spam now uses mispelled words and numbers in words to trip other filtering methods. So measure percent words mispelled or with numbers in them and above a particular threshold consider it spam.

Guaranteed?

[bluntmanspam]

From the home page (in BIG H1 font):

For the first time ever: 100% reliability in combating spam. Guaranteed.

But from the first point on the TOS page:

(1) Warranties and waivers. You understand that there are no guarantees, either expressed or implied, regarding the accuracy, confidentiality or availability of the service.

Hmmm...

It gets better...

[alakon]

Screams IP theft.

(1) Warranties and waivers. You understand that there are no guarantees, either expressed or implied, regarding the accuracy, confidentiality or availability of the service.

eProvisia LCC may choose to share any information acquired in the course of providing its services with other entities, and may, at its sole discretion and based on this information, take whichever actions the company, its affiliates, subsidiaries, or representatives, consider to be appropriate. You henceforth void your reasonable expectation of privacy, and your constitutional rights to a fair and speedy trial.

Not only that, but the contract is "governed by and construed in accordance with the laws of Uninhabited Sovereign Territory of Palmyra Atoll, ..."

Many Solutions Already (indirectly) Human-Powered

[tabdelgawad]

Many popular anti-spam systems, e.g. those implemented by webmail services, are already indirectly human-powered. Users classify their own spam emails and the everyone benefits system-wide without privacy concerns.

I'd say the system works pretty well. My Yahoo account, which was unusable after being harvested from my Usenet postings, is usable again. I just checked, and I have 426 messages in my bulk (spam) folder and 9 in my inbox. Of the nine, half (ok, 4!) are auto-responses from mail daemons to messages I never sent, while the other half are spam that escaped the filters. Not bad at all for a few days' worth of mail.

I think a sensible business model is for the webmail services to leverage their huge, continually updated, spam database and license them to ISPs, who can then filter spam at the server level before users download anything. I think that's much more elegant than software+community based solutions implemented at the user level.

Palmyra Atoll Facts

[omahajim]

From the CIA World Factbook on Palmyra Atoll [cia.gov]:

Location: Oceania, atoll in the North Pacific Ocean, about half way between Hawaii and American Samoa ( 5 52 N, 162 06 W)

Area: about 50 islets covered with dense vegetation, coconut trees, and balsa-like trees up to 30 meters tall
total: 11.9 sq km
land: 11.9 sq km

Population: no indigenous inhabitants; 4 to 20 Nature Conservancy staff, US Fish and Wildlife staff (July 2004 est.)

Dependency status: incorporated territory of the US; privately owned, but administered from Washington, DC, by the Fish and Wildlife Service of the US Department of the Interior; the Office of Insular Affairs of the US Department of the Interior continues to administer nine excluded areas comprising certain tidal and submerged lands within the 12 nm territorial sea or within the lagoon

Legal system: the laws of the US, where applicable, apply

Flag description: the flag of the US is used

Economy - overview: no economic activity

This page was last updated on 14 September, 2004

Translation of eProvisia's four-page web site : We're from the Government, and we're here to help you.

Re:Buzzword Bingo

[jrumney]

Its amazing how many people, even here on Slashdot, can't spot such an obvious scam. There was another company called "Edge Corporation" that used to sell a service where for $14.95/month they would let you know if your credit card number showed up on the lists being traded by fraudsters. Of course, you had to give them your credit card details so they could check on your card. The reason I found out, is a mysterious 39.95GBP showed up on my girlfriend's credit card bill, which I traced back to the same organization, and she hadn't even been stupid enough to sign up for their service.

I expect the same is going on here. Sign up for their service, and not only do you lose your $20, but you end up on a load more spammer's lists as well.

What currency?

[PaybackCS]

As seen here [nature.org] the Palmyra Atoll is nothing more then an inhabited ring of island a thousand miles south of Hawaii. There is no population, no government, no money, and very likely no eProvisia. The site is a complete farce, and is probably run by spammers collecting email addresses and names.

Re:LLC Companies

[Anonymous Coward]

There are valid reasons to form an LP. And lumping LPs together with GPs is overgeneralization.

My employer's company runs about twenty LPs. They invest in other business entities that carry the actual risk, so the "equal responsibilities for mishaps" is quite limited.

At least I think that's how it works. But I guarantee you that the lawyers & execs running this boat understand it better than you & me, & maybe even your BLE professor too.

Boy, I hope this "Post Anonymously" works right.

Re:synergy!

[Glog]

Ah the good old times - the last company I worked for was called "Synergize Solutions". It was during the "boom" times so we had "free donuts Friday".

Re:Palmyra Atoll dollars

[jonadab]

> One Palmyra Atoll dollar = 17 pieces of mithril, or approximately twenty
> kilograms of fairy dust. There's no such thing, people. This is a joke.

Actually, mithril does exist. It's also known as titanium ore or titanium
steel (depending on whether it's being mined or whether a worked object is
being described).

It's Adamantium that hasn't been invented yet. Unlike mithril, adamantium is
not vulnerable to the liquid nitrogen freeze-and-shatter attack. It is
speculated to contain plastic polymer in addition to metal alloy, but we don't
know how to actually make it.

HTH.HAND.

Clever Ad for Barracuda

[Titusdot Groan]

This article is not a troll. This is a very cleverly written ad for Barracuda.

I considered starting such a business...

[Flyboy Connor]

...for a brief moment.

I immediately realised such a business would never thrive, because:

• Businesses would never allow an outside firm to scan potentially sensitive emails.
• Even for humans it is sometimes very difficult to distinguish spam from real mail, especially if you are not an insider to the business.
• It could not be supported by an automated mechanism, because you could be sued if you would filter out real mail as spam, even if it happens rarely.
• If it takes about 10 seconds to scan a piece of mail, any human could scan a maximum of about 300 mails an hour. Since this is the amount of spam I receive in a single day, I have to pay at least an hour's fee per day to scan my mail, or probably something like $3000 a month. Nobody will pay this.

Yes, it's a hoax. Which could be immediately deduced from the fact that it is not viable business (especially with the price they quote).