One of our key objectives is …
… to bring people together. To make sure that we can do this, it is important to prevent spammers and fakers from getting contact to real users. In this transparency report – the sixth already – we explain what changed in the anti-spam systems over the last quarter and which observations we made.
What has changed since the last anti-spam report?
Except for training a new behavioral model and making slight adjustments to our algorithms, we focused this quarter mainly on migrating our AntiSpam System into the cloud. This will allow us to add new features to our system with less engineering effort. Moreover, the AntiSpam System can now be scaled more easily.
In comparison to the last quarter the amount of spammers has remained stable. However, in early December we switched on a new feature in AntiSpam. This allows us to punish spammer profiles that show properties which were identified to be malicious in the meantime even if those profiles are in a dormant state. In addition, spammers have been more active in the first three quarters of December.
As expected, the percentage of spammers among all daily active users remains constant as we did not add substantially different algorithms to our system.
Although there are only a few spam profiles, they are active above average and generate a fair amount of likes, i.e. positive votes in the Match.
The percentage of spam votes compared to the total votes is 3.42%. This is almost 1% less votes induced by spammers than in the previous quarter. We attribute this fact mostly to reduced spam activity in the last quarter of December.
Users meet fewer spammers
Spammers can be reported by users as well as identified automatically by the anti-spam system. Our goal is to block as many spammers as possible automatically, before the critical number of reports is reached or the support staff have to intervene manually.
In this quarter we succeeded in blocking 85% of all spammers automatically using the anti-spam system. The remaining 15% spammers that got reported by users was also helpful. These reports help us to continuously improve our system. If it is a new type of spam or a completely new behavior of spammers, we include this information in the system. With this we make sure to be able to recognize spammers sooner and we can even retroactively block profiles which showed the same characteristics in the past. Our main goal, however is to get the anti-spam system to find 100% of all spammers so that no user is forced to encounter them.
What happens to a report?
|Quarter||Reports||Users||Of which were no spammers|
A question often asked is why a user is not blocked immediately after the first report. Very simple: users report other users for the most diverse reasons. Not all of these reasons justify blocking the reported user. Every day we receive a huge number of reports: In this quarter alone we got 1.292.563 reports that reported a staggering 1.122.601 of different user profiles. However, 65% of these profiles were, in fact, no spammers.
The anti-spam system is this fast on average:
We have to be very careful when blocking users automatically. On the one hand, we want to catch as many spammers as possible, but we must not be too quick to judge. Otherwise, we risk blocking real users who only briefly exhibit spam-like behavior. This can happen, for example, if Match is played very fast. The anti-spam system therefore waits until a user has exhibited negative behavior on several occasions.
On average in this quarter, we took 1.2 hours from the first event executed, e.g. a vote in the Match, till the profile got blocked through the anti-spam system. This is significantly faster than in the previous quarter and, though also influenced by the spammer’s behavior, mainly a result of our continuous effort to train better machine learning models.
The new behavioral model, launched this quarter, was trained with the explicit goal to reach decisions faster than our previous model.
Are spammers more active on male or female accounts?
|Q3/2016||♀ 84%||♂ 16%||♀ 29||♂ 29|
|Q4/2016||♀ 70%||♂ 30%||♀ 31||♂ 29|
|Q1/2017||♀ 77%||♂ 23%||♀ 29||♂ 29|
|Q2/2017||♀ 80%||♂ 20%||♀ 29||♂ 39|
|Q3/2017||♀ 81%||♂ 19%||♀ 29||♂ 40|
|Q4/2017||♀ 81%||♂ 19%||♀ 30||♂ 40|
As in the previous quarters, spammers had predominantly female profiles. From all spam profiles a staggering 81.05% were female. On average these profiles were 30 years old. The male spam profiles were on average 10 years older. Why this discrepancy exists is a mystery.