Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Comment spam random text template

gist.github.com

Post reply on HN
Print view View original

61–70 of 76 posts

Re: Comment spam random text template

#61

Post by robrenaud »

klodolph wrote: post #51
tikhonj wrote: post #6

Earlier quoted context omitted.

Sure, there can be millions of unique texts, but they all follow certain very similar patterns. That's what would make even a simple naive Bayes filter very effective against them.

I think you're misunderstanding how naïve Bayesian spam filters work. These techniques explicitly do not recognize patterns at all, they just look at vocabulary, and treat each word as a completely independent feature—so they don't even recognize "discount viagra" as a pattern, but merely two independent features "discount" and "viagra".

You can build a naive bayes classifier with n-gram features.

Re: Comment spam random text template

#63

Post by tikhonj »

klodolph wrote: post #51
tikhonj wrote: post #6

Earlier quoted context omitted.

Sure, there can be millions of unique texts, but they all follow certain very similar patterns. That's what would make even a simple naive Bayes filter very effective against them.

I think you're misunderstanding how naïve Bayesian spam filters work. These techniques explicitly do not recognize patterns at all, they just look at vocabulary, and treat each word as a completely independent feature—so they don't even recognize "discount viagra" as a pattern, but merely two independent features "discount" and "viagra".

First of all, you can have any sort of features you want. So you could certainly look at things like sentence structure, pairs of words and so on.

More importantly, I used "pattern" in a very general sense: there is strictly more structure to a post generated with this system than a random post or even a human-authored one. So repeated words between posts are a pattern, and a very important one at that.

Re: Comment spam random text template

#64

Post by Narretz »

I especially hate the kind of spam you find in the comments over at rockpapershotgun.com, which looks similar, but is more annoying as it interrupts the comment tree:

Joshua. I just agree… Bonnie`s postlng is good, on sunday I bought a gorgeous Acura after bringing in $7140 thiss month and-more than, $10,000 last-munth. this is certainly the coolest job I’ve ever done. I began this seven months/ago and straight away was bringin in over $81 per-hour. I follow instructions here

It was very pervasive for some months, but it looks like the guys finally found a way to block most of it. Shouldn't be too hard with alle the numbers and dollar signs.

Re: Comment spam random text template

#65

Post by A1kmm »

brador wrote: post #62
Titanous wrote: post #28

It looks like this factors out to 4,351,250,624 unique comments.

How did you calculate this?

The set of strings is isomorphic to a Cartesian Product of sets of the same cardinalities as the set of options (for example, if the template was "{I,We} like {HTML, CSS, Javascript}", you can make a set {0,1}x{0,1,2}, where each element in the product set maps to one string and vice versa. For example (0,0) might represent I like HTML, and (1,2) might represent We like Javascript).

Because there is a one to one map (bijection) between the Cartesian Product set and the set of strings, the size of the product set is the same as the size of the set of strings.

The cardinality of a cartesian product AxB, where A and B are sets, is |AxB|=|A|x|B|, so to find the size of the set of strings, you just need to multiply together the number of options at each point in the template where you have a choice.

Re: Comment spam random text template

#66

Post by A1kmm »

klodolph wrote: post #52
bcoates wrote: post #3

Isn't a trivial templating system like this a very easy thing to train a bayesian spamfilter on, even without the source template? I guess blogspammers mostly prey on entirely unmaintained sites for something like this to work. But I'm going to have a hard time avoiding using "fastidious!" as a general expression of approval now.

Naïve Bayesian filtering might not work very well on this kind of text. It basically looks like a regular comment, until you start recognizing that it always follows the same pattern. Your basic Bayesian classifier will throw all of the words in a set before analyzing them, which loses all of the information about patterns and word order. The resulting words are considered "independent" which means that even though t…

You can run a Bayesian filter over pairs (or even triplets) of words (although this could cause the probabilities to be a bit off, because pairs like "although this" and "this could" are not truly independent). The downside is that as you do this, you drastically increase the size of the model and the amount of training data required.

Bayesian filters can also take more than just the words in the text into account - for example, they can take the submitting IP address (or perhaps /24 or ASN) into account, or a spam classification from external sources.

There are certainly better methods that could be built for recognising unknown templates - a simple known-state Markov model would be sufficient for the cases where templates substitute one word at a time, and you could conceivably use an unsupervised learning algorithm to discover an unknown number of models from a large corpus of comments.

Re: Comment spam random text template

#67

Post by brador »

A1kmm wrote: post #65
brador wrote: post #62

Earlier quoted context omitted.

How did you calculate this?

The set of strings is isomorphic to a Cartesian Product of sets of the same cardinalities as the set of options (for example, if the template was "{I,We} like {HTML, CSS, Javascript}", you can make a set {0,1}x{0,1,2}, where each element in the product set maps to one string and vice versa. For example (0,0) might represent I like HTML, and (1,2) might represent We like Javascript). Because there is a one to one map…

Multiply the count in each set, gotcha. But did he just count it for each set? It seems like that would be time consuming. Is there some shortcut? or did he just write a quick program to do it?

Re: Comment spam random text template

#68

Post by csmattryder »

YokoZar wrote: post #39
OGC wrote: post #35

Earlier quoted context omitted.

From http://codex.wordpress.org/Comments_in_WordPress Depending on your site's settings, comments display slightly differently from site to site. The basic comment form includes: Name Email Website Comment

I have the website field turned off -- it's not visible if you are a real user using the form, and "website" doesn't show up next to comments. But I still get a lot of spam containing the website field; it seems like the bot is just automatically submitting POST requests with website included. This should be trivial to detect -- is there a plugin that just immediately deletes all of these out of hand before passing t…

There's a way of setting the 'hidden' property on an input field and checking if the field is blank.

If it's a bot sending POST requests, there's a high chance a "Website:" field would be filled out. Innocent users will not be able to fill out the field, in the case of a legitimate comment.

Re: Comment spam random text template

#69

Post by soult »

b0rsuk wrote: post #59

Don't captchas solve this ? Are spam tools able to bypass them too often, or is it because people find captchas annoying ?

Even a captcha that can separate humans from machines with 100% accuracy does not defend against cheap labour ($1-5 / 1000 captchas) in developing countries like India.

Re: Comment spam random text template

#70

Post by D4N_ »

klodolph wrote: post #52
bcoates wrote: post #3

Isn't a trivial templating system like this a very easy thing to train a bayesian spamfilter on, even without the source template? I guess blogspammers mostly prey on entirely unmaintained sites for something like this to work. But I'm going to have a hard time avoiding using "fastidious!" as a general expression of approval now.

Naïve Bayesian filtering might not work very well on this kind of text. It basically looks like a regular comment, until you start recognizing that it always follows the same pattern. Your basic Bayesian classifier will throw all of the words in a set before analyzing them, which loses all of the information about patterns and word order. The resulting words are considered "independent" which means that even though t…

Something like a Hidden Markov Model might be better then? That way you can can keep the information about word order.
Post reply on HN