Security needs to be better automated, but while detecting attackers is great, all too often automation means that security teams are left with chasing down a list of security events that turn out not to be an attack but unexpected system, network, or user behavior.
These “false positives” are the bane of most machine-learning systems: valid e-mail messages blocked by anti-spam systems, unexploitable software defects flagged by software analysis systems, and normal application traffic identified as potentially malicious by an intrusion detection system. First-generation security information and event management (SIEM) systems, for example, would often deliver lists of potential “offenses” to security teams, leading to a lot of work in wild goose chases, says Jay Bretzmann, market segment manager for security intelligence at IBM Security Systems.
SOURCE: darkreading.com
