Editor’s note: This is part two of a series on generative AI in cybersecurity.

Part one: Understanding and Leveraging Generative AI in Cybersecurity

Part three: Building a Generative AI Strategy for SecOps

Part four: CISO’s Guide: Six Steps to Start Adopting AI

In the first part of this series, we explored the topic of generative AI and what it means generally within the field of cybersecurity. We’ll now dive a bit deeper into the current security landscape, some common challenges we’re seeing, and ways in which generative AI is positioned to address these challenges.

Current Security Landscape and Challenges

Many challenges within cybersecurity can be attributed to data—the amount of it, where it resides, and how to extract the most value from it. Let’s break down each of these challenges and see how generative AI can be applied to provide additional value.

Challenge #1: Data Volume

Pulling together telemetry from across the enterprise—including endpoint, network, and cloud—results in an extremely large quantity of data. Billions of individual events must be processed in the context of all the other signals surrounding them. This challenge is compounded by the immense number of external data sources, including threat intelligence. By combining all of these signals, complex narratives begin to form, which security teams can then analyze to determine overall risk to the business. This elevates detection coverage, moving up the “pyramid of pain” from simple indicators of compromise (IoCs) towards behavioral analysis and identification of tactics, techniques, and procedures.

Today’s tooling does a decent job at tackling some of these problems, but certainly leaves room for improvement. For example, SIEMs can collect and process large quantities of information, but contextualizing this data into narratives and processing them through workflows remains an issue. SOAR excels at executing workflows but falls short when it comes to analyzing narratives in the context of the business and performing more advanced reasoning and decision-making.

How Can Generative AI Address This Problem?

Generative AI can untangle the data volume challenge by automating the contextualization of disparate pieces of information and identifying complex patterns. It can quickly analyze massive amounts of data and generate summaries of its findings. Security analysts can then chat interactively to further unpack these findings and answer any lingering questions. By building these narratives, generative AI enables analysts to identify potential threats more quickly and accurately. Moreover, the models can adapt to new information as it becomes available, such as threat advisories, and update their interpretations of ongoing security situations.

Challenge #2: Disparate Artifacts

As organizational data continues to grow in volume, it’s also becoming increasingly decentralized. Logs may reside in separate silos but need to be combined as part of the investigative process. Additionally, these logs may not always be normalized properly; a vendor may push an update that alters a log’s formatting and breaks parsing. On top of this, artifacts sometimes begin to overlap, such as private IP address ranges. Segmenting these artifacts, while still having the ability to stitch them together in the context of a larger investigation, is absolutely critical.

How Can Generative AI Address This Problem?

For disparate artifacts, generative AI can identify relationships and correlations among them, even when they are decentralized. This allows aggregation and normalization of data from various sources for a comprehensive analysis, regardless of any changes in vendor formatting. Generative AI can also intelligently segment and analyze overlapping artifacts while maintaining connections to the broader context of an investigation.

Challenge #3: Incident Enrichment

To further improve contextualization and extract the most relevant narratives, data must be enriched as part of the broader investigation. This includes individual artifact enrichment, such as whether an external IP address is malicious. The answer to this question oftentimes requires input from external sources, such as threat intelligence data. Building out these workflows is trivial, but the challenge lies in contextualizing the results and acting on the outputs.

This enrichment also applies at the incident level—for example, evaluating historical incidents to drive the current investigative process. A given activity may look suspicious, but historically has been closed as a false positive when certain conditions were met (such as being sourced from this user or being run by this specific process). These insights can be used to provide a more accurate analysis of the incident. Conversely, two open incidents may be merged if they appear to be part of a larger campaign. This would expand the overall scope of the incident and ensure that all relevant information is included in the analysis.

How Can Generative AI Address This Problem?

To improve data enrichment, generative AI can efficiently automate workflows that leverage external sources like threat intelligence data. By learning from past incidents, generative AI can also improve incident response by accounting for historical context. For example, it can flag false positives or link related open incidents, thereby streamlining analysis and decision-making for cybersecurity teams.

Conclusion

Generative AI has the potential to greatly enhance the speed, accuracy, and efficiency of cybersecurity efforts in dealing with massive amounts of data and complex security challenges. Stay tuned for part 3 of this blog series, where we’ll explore different ways in which ReliaQuest is leveraging AI to address these challenges.