Speakers
Description
As scientific workloads migrate to the cloud, securing research infrastructure against automated threats, opportunistic cryptominers, agentic bad actors, and even non-agentic LLMs is a growing challenge. This security must be balanced with the accessibility required for reproducible science.
To address this, we share insights from operating cloud-based scientific platforms, detailing an automated monitoring pipeline that aggregates logs, detects anomalies, and generates daily security reports. These reports enable rapid patching against emerging threats and identify orphaned resources to reduce cloud costs.
Furthermore, we explore integrating Large Language Models (LLMs) to transform unstructured logs into actionable insights, streamlining threat response. Ultimately, we provide practical, time-saving strategies to help researchers and administrators maintain secure, proactive environments without distracting from their core scientific work.