ABUSE AI is a prototype API that uses machine learning techniques to classify complaints and extract offending IPs.
The type field returns a JS object containing the confidence level of prediction for each type of complaint. For example: a .95 confidence level for malicious content means the model is 95% sure the sent text was from a malicious content complaint.
The project is currently a prototype & the base data set's label accuracy is questionable. The text classification process in place is also rudimentary. Therefore, event though the API can classify emails successfully in the wild, use it in production at your own risk.
Live version at: http://abuseai.us-east-1.elasticbeanstalk.com/
Due to the sensitive nature of the data involved. If you wish to contribue, please open an issue requesting the data and a description of your intention. I will provide you with the necessary data set upon receipt.
With more data more distinct classification would be possible. If you wish to provide additional data, please open an issue as well.
To use the api POST a subject and body to www.AbuseAI.com/api/email.
curl --data "subject=Phishing attack&body=Phising attack from site emulating bank at ip 8.8.4.4" http://abuseai.us-east-1.elasticbeanstalk.com/api/email
There are currently 3 prediction classes that encompass a variety of complaints.
Any complaint relating to websites which are serving malicious content such as malware, fraudulate sites, phishing, pedophilia, or other criminal activity.
Any complaint about servers performing brute force attacks, DDOS, portscans, intrusions and other network attacks as well as playstation abuse, pharming, spam, and other malicious activity.
Any complaint pertaining to copyright, trademark, or intellectual property infringment as well as DMCA take down notices.
git archive -v -o api-$(git rev-parse --short HEAD).zip --format=zip HEAD --worktree-attributes.- Upload a new version in the ElasticBeanstalk console.