David Tom Foss is an independent researcher based in Germany. His research spans adversarial cryptanalysis, post-quantum cryptographic security, deceptive alignment in agentic AI systems, and causal methods for automated vulnerability detection.

His peer-reviewed contributions include a cipher-agnostic security margin framework with blind architecture classification across five cipher families, black-box statistical characterization of all three NIST post-quantum standards (ML-KEM, ML-DSA, HQC) including a compression isolation experiment on Module-LWE ciphertext, a deterministic validation architecture for LLM-based causal knowledge extraction achieving 100% byte-level reproducibility across three model families, and the discovery of persistent non-decaying carry-chain leakage in ARX ciphers yielding full-round known-key distinguishers for the entire Speck family and Threefish-256 (accepted, IEEE ICECET 2026).

Prior to research, he served for a decade in the German Armed Forces as a signals noncommissioned officer in electronic warfare and mountain infantry communications.