Sharing attack telemetry across organisations boosts collective defence but raises compliance headaches. We apply differential‑privacy noise mechanisms and secure enclaves (SGX) to sanitise NetFlow and IDS alert feeds before aggregation. The framework guarantees probabilistic anonymity while preserving statistical fidelity for machine‑learning models. Pilot deployments include three universities and one regional ISP.

Generative AI systems are vulnerable to crafted prompts that bypass safeguards, leak data, or induce malicious actions. We curate a living corpus of real‑world injection strings, automate fuzz testing across major LLM APIs, and prototype semantic filters and policy‑based output rewriting. Deliverables include a public “Risk Atlas,” mitigation guidelines for Bangladeshi fintechs and government chatbots, and conference presentations on guardrail engineering.

The project tailors the MITRE ATT&CK for ICS matrix to the power‑generation and SCADA ecosystem. We are methodically mapping 94 network‑detectable techniques to both open‑source and proprietary energy‑sector datasets, highlighting protocol blind spots (Modbus, DNP3, IEC 61850, OPC UA) and hybrid IT/OT attack paths. The outcome will be an open “ICS Dataset Coverage Checklist” and a peer‑reviewed journal paper that utilities can use to baseline their detection posture.