The concept of security information and event management (SIEM) has its origins in the need to have a unified view of security events across multiple technologies. From antivirus software detecting malware on some endpoint to a firewall blocking suspicious traffic on an unauthorized port, SIEM gives security operations teams a central dashboard from which they could assess the "security posture" of their organizations. The dashboard serves as a kind of one-stop shop for finding and acting on security-related information, hence the name.
Over the years, expectations from these systems have grown, with marketers beginning to claim that their SIEM platforms can even be used to correlate seemingly isolated events and identify threats that individual security products would otherwise miss. Given the rising importance of cybersecurity, this technology quickly catapulted to the top of every CISO's wish list.
What followed was a rush fueled by paranoia and efficient marketing to set up SIEM-powered security operations centers. SIEM platforms evolved into multifaceted tools for monitoring, reporting, and forensic investigation. Suddenly, SIEM had become such a crucial component of cybersecurity strategy that regulations began requiring the use of such platforms in banking institutions.
SIEM, however, is hardly a panacea for all security problems. Issues like complex implementation, lack of flexibility, slow response times, and sky-high costs became seemingly necessary evils. Moreover, the torrent of alerts kept security teams under constant pressure to "close issues." It was only natural for users to notice the elephant in the room: The reality of SIEM systems had fallen far short of all the hype.
In hindsight, it seems obvious that SIEM systems were unable to deliver on their marketers' promises. They were built on a back-end technology — relational databases — that wasn't designed for use like this. The volume and variety of data generated in today's environments shows that traditional RDBMS-based SIEM systems cannot find the proverbial needle in the haystack. The volume of alerts generated cannot be investigated manually by SOC analysts. Correlation-based rules alone cannot identify the ever-increasing number of scenarios in which a cybersecurity incident can penetrate an organization.
Advancements in technologies managing unstructured data (i.e., big data), toolkits leveraging data science in machine-assisted decision-making (i.e., machine learning), and frameworks that allow multiple isolated security applications to work together (i.e., security orchestration) have given us the firepower needed to scale up the capabilities of the next generation of SIEM systems.
Security analytics and orchestration solutions available today are capable of processing large volumes of data at high speeds in a scale-as-you-grow architecture. This processing capability allows users to build models using data science. These models can identify both patterns of normal activity and outliers or potentially malicious events without writing specific rules. The systems can be integrated with other applications to exchange information and investigate the outliers identified. Orchestration can also be paired with automated threat response mechanisms, reducing human involvement.
Now, marketers have begun raising the bar on what technology alone can deliver, furthering the competition between vendors to prove technical superiority. Today's automated black-box solutions claim to leverage advanced artificial intelligence and orchestration to provide security at the click of a button. There is a sense of déjà vu with banking regulators prescribing big data-based analytics solutions to manage security operations.
What is needed is an understanding of what capabilities users should expect of the technologies they use, and how to correctly leverage them to improve security. Every IT environment is unique, so expecting software to automagically secure them all with artificial intelligence is still beyond the realm of what technology can deliver.
Technology plays a crucial role in processing large volumes of data, performing repeated tasks, making disconnected systems work together, and using complex math to identify the potentially malicious. This is already the bulk of the heavy lifting in ensuring security; what is needed further is human intelligence for correct navigation.
No doubt, this calls for a comprehensive skill set, which is evidently a scarce resource. While it might appear difficult to pull off, being blown away by marketers boasting about one-size-fits-all solutions is certainly not the best way to start.