Mamun Akand

Scalable Multi‑Recipient Encryption

Problem: Confidential file sharing cost scales linearly with recipients (O(N)).
Action: Designed scheme with pairing‑based cryptography; optimized keying.
Result: Reduced encryptions from O(N)→O(1) and cut key size by 60%+.

Risk‑Weighted Secret Sharing

Problem: Uniform shares enabled insider collusion risk.
Action: Introduced dynamic, risk‑based weights (MCDM) with threat modeling.
Result: Lowered collusion risk; preserved O(1) reconstruction; integrated into product line.

DLP: Forensic Watermarking & Tamper‑Evident Metadata

Problem: Insider leak investigations were slow and storage‑heavy.
Action: Built robust watermarking and metadata chaining; production‑integrated.
Result: Faster investigations and reduced storage footprint.

Privacy‑Preserving AI Proxy (PPAP)

PII‑aware proxy for safer AI inference.

Action: Reverse proxy that strips PII before routing to on‑prem/cloud LLMs; combined deterministic redaction and context‑aware masking.
Result: Reduced data leakage risk; containerized for self‑service deployment.

View Repository

AI Research Assistant Pipeline

Action: Auto‑extracts technologies and gaps; generates reports; optimized for conference research.
Result: Cut manual analysis time by 50%+.

AI & Agentic System Security

Security research and evaluations for agentic systems.

Action: Built threat taxonomy (perception, cognition, memory, action); evaluated defenses incl. TEE, privacy‑preserving inference, tamper‑resistant deployment.