OTRO combines replicated square-root ORAM instances, epoch rotation with dummy padding, and KV-cache-aware chunking to make tokenizer lookups oblivious with at most 4.5% TTFT overhead and under 0.5 GB extra memory in TDX + H100 setups.
About Private Processing
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
OTRO: Oblivious Tokenization Path with Square-Root ORAM
OTRO combines replicated square-root ORAM instances, epoch rotation with dummy padding, and KV-cache-aware chunking to make tokenizer lookups oblivious with at most 4.5% TTFT overhead and under 0.5 GB extra memory in TDX + H100 setups.