The Three Pillars of Cyber Security Automation: Addressing the Resource Shortfall

Share on Facebook0Share on Google+0Tweet about this on TwitterShare on LinkedIn0Email this to someone

Cyber security operations require an automated response to keep up with threats.  This blog looks at the three pillars of cyber security automation and gives insight into how they help sustain cyber resilience.

The speed of change in cybersecurity is incredible, with new malware variants continually emerging alongside new hacking techniques that change attack profiles beyond recognition. The tools we employ, such as antivirus (AV) scanners and intrusion prevention systems (IPS) struggle to keep up with this rate of change, and while these technology controls are still required to banish legacy threats, modern attackers integrate evasion techniques that easily bypass our aging ICT immune systems.

Over the past few years, contemporary technologies have appeared integrated into cybersecurity products, purporting to solve these problems with cutting edge artificial intelligence, machine learning and user behavioural analysis. On the surface of it, this sounds great, but what are these technologies, and do they really do what they say they do? Furthermore, how do you upgrade your operational security capability to meet today’s challenges using this new technology? The reality is that these technologies will help, that is a fact, but they all achieve one end goal – incident response automation. Let’s look at how these modern technologies play their part in the cybersecurity response and how this can afford operational teams an edge over their adversaries.

The 3 Key Technology Categories in Operational Security

The three technology categories that have found their way into operational security products over the past few years are Artificial Intelligence (AI), Machine Learning and User Entity Behavioural Analysis (UEBA).

AI is often misunderstood and misrepresented. At its heart, the goals of AI are to build computer systems that have innate human traits, such as reasoning, perception, intuition, learning, planning and the ability to turn data into knowledge. AI is an entire branch of academic research, modelling mountains of data to make sense of it in the real world. This results in the formation of automated medical doctors, lawyers and even autonomous vehicles. It has massive application in the real world and is a paradigm shift for computers from old-school number-crunchers to analytical thinkers, capable of complex diagnostic work. Research into AI in cybersecurity is prolific; event log information, network traces and system-to-system conversations provide the perfect information base for AI algorithms to mine. We do, however, need to understand the goals of integrating AI capabilities into cybersecurity products and how its capabilities fundamentally help us achieve our desired outcomes.

Machine Learning serves as a feedback system into operational security detection systems, allowing increasingly accurate detections as algorithms learn from their mistakes (and successes). The most well-known machine learning systems are recommender systems used by online retailers such as Amazon and Alibaba. These suggest products that you might be interested in looking at, based on your purchase history, browsing habits and searches. Facebook also uses machine learning to improve the value of your News Feed, coupling interesting posts from connections with the adverts for products you have shown in an interest in. In cybersecurity, machine learning is invaluable. Security Operations Centre (SOC) analysts waste far too much time chasing false positives, so any system that improves the odds of an alert being real, while weeding out the chaff, improves the SOC’s overall efficiency.

UEBA helps identify unusual user behaviour that could indicate an insider threat. This approach is underpinned by profiling systems that learn what normal looks like from the perspective of end users, then variations in behaviour, such as a user trying to get to a file store or database they don’t normally access, can be flagged as risky. The  SOC can be alerted to this anomaly and can subsequently decide to investigate. UEBA packages the profiling, baselining and alerting of insider threats in a way that makes it easy (relatively speaking) to deploy in a large enterprise, however, it still takes time to learn the context of the organisation (systems and architecture) and to learn what normal looks like. There is no doubt that UEBA is sound technology and should be considered on any SOC’s operational threat management roadmap.

SOC Operations and Automating Threat Responses

Each data modelling and feedback system has one primary goal: automating the threat response process to reduce the time between detecting and eradicating the threat. In most SOCs, at least half of an analyst’s time can be spent investigating false flags, with the other half writing reports. Even with a highly-tuned security information and event management (SIEM) system, analysts are responding to hundreds (if not thousands) of potential incidents per week. This is time consuming and demoralising for analysts when 99.9% of their investigations are against false positives, hence they often feel their job is futile. Interestingly, the latest shift in integrating threat intelligence into the SOC’s ecosystem has only served to make this problem worse, since millions of additional correlation items are introduced into the mix every month, pushing the number of false positives up rather than down. The value we get from hiring expert cybersecurity analysts is their ability to leave the confines of the SOC and investigate real incidents, piecing together disparate lines of investigation inside and outside the ICT systems – yet it’s this aspect of the job that is overwhelmingly underserviced since they spend most of their time chasing down false leads.

The three pillars of cyber security automation we’ve looked at here give us the prospects of a brighter future, where the technology supports our beleaguered SOC analyst by:

  1. Identifying false flags and removing them from the analyst’s workload;
  2. Identifying real threats and performing automated responses (running scripts, making changes to network architecture and quarantining compromised systems);
  3. Pre-packaging the investigation to help the analyst resolve the incident.

To move forward, one of the biggest hurdles to overcome when introducing automation is that of trust; remediation efforts can cause a loss of service to the business, something that is rarely acceptable if it’s unnecessary. If the SOC automatically quarantines computers, removes user-rights and turns off customer-facing websites, the business will invariably lose money. If these service disruptions are due to false positives, it has a cumulatively detrimental effect on the business’s bottom-line and long-term trust in the SOC is eroded.

Automating security processes and incident response capabilities must become the goal of every SOC and the only way to do this effectively is through technology. Using a combination of proactive threat identification and security testing (penetration testing can be used for this purpose as it appears to SOC systems as a real attack) SOC analysts can support these AI, Machine Learning and UEBA systems to quickly learn what normal looks like, and as more automations are introduced (and tested) the SOC can build trust with the business that the security investments are worthwhile.

Cyber Security Automation – stay in control

AI, Machine Learning and UEBA are perfect for automating the incident response process. However, automation introduces risk, so comprehensive planning and testing is required to develop trust in any automatic response capability.

AI, Machine Learning and UEBA technologies will undoubtedly improve your organisation’s overall cyber security posture, but the fundamental role of the SOC analyst needs to change as the technology platforms evolve, helping them become master rather than slave to the underpinning technology. If you start by focusing on automating the response to simple, high-confidence alerts, you’ll afford your analysts the time they need to model more complex threats and continually fine-tune their systems, while helping your SOC management team build confidence with the business.

Cyber security automation will eventually be an integral component of every organisation’s defensive security posture, when this happens your SOC analysts will finally gain an edge over their adversaries.

You can read more Automation blog posts here

Understanding the deepening malware crisis