Addressing the Data Challenge for Trust and Cybersecurity: Malcolm Harkins, Cybersecurity expert and Chief Security and Trust Officer, HiddenLayer

The rapid expansion of semi-structured and unstructured data, combined with the accelerated adoption of AI and Gen AI, has prompted a fundamental reassessment of Trust and the economics of Cybersecurity.  In traditional threat detection and remediation: the explosion of devices, apps, and various environments is generating an exponential increase in the log data that security organizations must swiftly curate and process at wire speeds.  AI models in this security context are the most capable of identifying and addressing cyber threats in real-time.  However, AI models themselves are not secure – existing enterprise controls do not protect against the various forms of attacks against AI.  At best, they provide indirect protection to the models by protecting the infrastructure where an AI model operates.  Thus, a new breed of cyber security solutions is required - purpose-built to protect AI models themselves.  As AI models themselves need protection – logging of model interactions is needed, model scanning is needed, and AI detection and response will be required. Combined these trends create even more security data we need to process to manage our risks.  But how do we scale our security data processing?  How do we manage the costs of more and more data? 

We need to change the economics of the data processing pipeline.  That is why I am excited about the DataPelago platform and how it is optimized for the entire data stack and more specifically designed for Gen AI and analytics workloads.  The potential to achieve 10x to 100x performance gain goes far beyond the traditional theory of Moore’s Law by refactoring data processing to capitalize on proprietary accelerated computing capabilities to achieve significant performance gains and thus dramatic compute cost reductions.   In a world of heterogeneous computing, we need a solution that can dynamically map operations to execution units, we need plug-and-play integration with open-source components (for example Apache Spark & Gluten) to ensure maximum flexibility.  Finally, we need frictionless solution deployment that enables seamless integration into existing environments without changes to data, workflows, or tools, with no vendor lock-in. DataPelago is the platform capable of changing the economics of the data processing pipeline which can enable the cost-effective expansion of AI, GenAI, and cybersecurity.