Harassing and threatening messages are detected and blocked by Simility, preventing them from getting to the members of a large social network.
If Simility can detect fraud as specialized as this,
there’s no limit to what we can do for your company.
What is message spamming? Platforms that allow users to message one another are trying to create an ecosystem whereby one user with a relevant connection to another user can communicate directly with them with for the benefit of all. Unfortunately, some fraudsters use these platforms to send unwanted messages to recipients, often en masse. These spammy messages can be hurtful or just plain annoying, or can be a means for the fraudster to poach business from the platform. They can take the form of direct private messages between users or as public messages or comments for anyone to see that damage your business and online reputation.
Since the message content is written in natural language, it’s very hard to separate the good messages from the bad. There can be thousands of data points in a single short message and multiple possible interpretations for each. Furthermore, there is a lot of gray area. Some messages may be deemed as just an annoyance and merit a minor sanction, while others might warrant more serious action. The same is true of businesses interactions and distinguishing between a partner and a competitor.
Simility takes a three-pronged approach in its harmful messaging model. We first parse the content to look for known keywords and sequences of words that fraudsters use. We then analyze the the sender’s account to see if it represents a fake persona. And finally we look for particular victim account profiles known to be attractive targets for the fraudsters. This complete process is done in real-time on big data sets and lets you prevent harmful messaging from getting to your users.
PARSE CONTENT OF MESSAGE
- Search for known keywords, linguistic patterns, and sequences of words that fraudsters use
- Search for multiple variations of words deemed important
- Completed with machine learning–no manual intervention needed
- Customization can be done by adding additional rules manually
ANALYZE SENDER’S PROFILE
- Use Device Recon to determine if the same person has created multiple accounts
- Check the natural language and images to determine if the message includes a stolen online profile or photo
- Compare the credibility of senders’ actions to their profiles
IDENTIFY VICTIM ACCOUNT PROFILES
- Identify potential targets of spammers
- Check profiles against the risk models for varying fraud types
- Continually adapt the models and apply regular expression rules
- Use real time review that is invisible to your users