[Original] views on spam text message Detection

I read what Yu Kai wrote.

Talk about the solution of junk stickers in interactive websites

I also talked about my thoughts on text message filtering. In my previous text message sending project, I encountered million spam messages a day. Later, I gradually improved the system, and the number of spam messages per day was limited to less than 1 million.

SMS sending mechanism:

In the past few years, many sites have provided the free mobile phone text message function, through the web site to send text messages directly to the user's mobile phone. The implementation process is to submit text messages to the carrier's SMS gateway through CMPP (mobile) Protocol through cooperation with the carrier.

Spam SMS analysis:

1. Because Spam messages also have a certain cost (even if the website has free sending, there is a certain amount of free sending restrictions), and now there are generally image verification codes, therefore, it is generally not easy to implement robot sending.

Therefore, spam messages are generally sent manually, which is much more costly than posting.

2. The recipient of spam messages is a mobile phone user. Generally, the contact information (phone number or mobile phone number) is left in the text message, which is generally not an email or QQ number. A spam message without any contact information is of little significance. (This is very important. We also discovered this rule more than a month after spam messages)

Technical implementation:

1. Implement a dedicated spam text message detection system. When users click to send a message, the website will send the text message (including the sender, content, IP address, and recipient) submitted to the SMS Detection System for detection. The SMS is submitted to the background only after the detection is passed.Program

2. The following lists the main detection aspects:

I. Text message content: Convert all text messages to ANSI, extract the numbers in the middle (including letters such as numbers, such as O and L), and combine the numbers, if more than five numbers are displayed, the data is put into the memory. A total of more than five text messages are combined in the same way, and subsequent text messages of the same combination cannot be detected.

2. Sender: Limit the number of senders sent per minute, hour, and day

3. Sender IP Address: Limit the number of sender IP addresses sent per minute, hour, and day

4. Send duplicate content: hash all sent content and put it into the memory. If the SMS of the same content reaches a certain value within one hour, the detection fails.

 

This is a project that we participated in a few years ago, and some aspects are hard to remember. The focus is to discuss it together. Thank you.

 

-----------------------

Retained the source for forwarding