Bug #435
closed
  
  
    
    
    
    
       - Status changed from New to In Progress
- Assignee set to Marc Dequènes
I am looking at Rspamd which seems promising and using a similar learning system. It would also replace policyd-weight and add other filtering methods.
I am going to look deeper into it and probably experiment a bit.
 
   
  
  
  
  
    
    
    
    
       - % Done changed from 0 to 40
Rspamd has been deployed with Ansible rules. So far it seems to work well and I've activated a few more features.
	I've setup and DNSBL which is not a blacklist but a way to query IP information based on AS and BGP lists.
	Also I'm working on replication of the stack on Toushirou. Currently Redis is syncing fine. I had to use stunnel because there is not concept of security in this database. I'm working on deploying Rspamd now.
	Also I deactivated Policyd-weight on Orfeo and intend to get rid of it very soon.
 
   
  
  
    
    
    
    I forgot to say I modified the training script to train both DSPAM and Rspamd, so we can later disengage DSPAM and see if the trainign is sufficient.
 
   
  
  
    
    
    
    
       - % Done changed from 40 to 80
The stack on Toushirou was done 3 days ago and works well.
	Outgoing mails are also filtered so the dirty redirection of www-data mails from Toushirou to DSPAM on Orfeo is deactivated.
	I disengaged DSPAM ~2 days ago and so far apart for more learning needed it seem to work fine.
 
   
  
  
    
    
    
    
       - Status changed from In Progress to Resolved
- % Done changed from 80 to 100
It's working. No user complaints :-).
	DSPAM was removed.
 
   
  
  
  
 
  
  
 
Also available in:  Atom
  PDF