TranScope
Hardware footprints reveal what model outputs can hide.
TranScope examines how tokenization and embedding access patterns affect caches, TLBs, and on-core accelerators during inference. With local control of the hardware, operating system, and runtime, an auditor measures these footprints to distinguish input distributions associated with training data.
The study evaluates distributional inference through privileged hardware measurements.
Significant findings
98% reported accuracy
Distributional information in hardware footprints
The abstract reports 98% out-of-distribution detection accuracy, with 1.5% false positives and 2.5% false negatives, for the evaluated large-scale model setting.
0.9 reported AUC
Stronger separation in the reported comparison
The abstract reports 0.9 AUC for TranScope versus 0.6 for PETAL. This is evidence from the paper’s evaluated setting, not a universal accuracy guarantee.
Cache + TLB locality
A mechanism behind the signal
Tokenization and embedding access patterns can change token locality, TLB behavior, and page-table walks even when software paths are alike.
Beyond output scores
Confidence hiding leaves hardware signals
The evaluated output-discretization defenses do not remove TranScope’s signal. Measurements require a local auditor with control of hardware, OS, and inference runtime; distributional inference alone does not establish exact training-record membership.
Explore the evidence
9 figures · Select a preview to enlargeHardware inference under confidence-hiding defenses

The evaluated hardware classifier is compared with a software baseline under different restrictions on model outputs.
Classification accuracy as models scale

Each point reports the highest accuracy at that model scale across the evaluated families.
Socket-power footprints across model sizes

The plot compares normalized per-inference socket-power measurements for the evaluated input groups.
From token locality to memory-hierarchy behavior

The mechanism diagram traces clustered and scattered embedding accesses through caches, TLBs, and page-table walks.
Page-table walks and token structure

The experiments vary zero-token counts and model size to examine their effects on page-table walks.
Token-distance distributions across model sizes

The distributions compare successive token distances for in-distribution and out-of-distribution samples.
Cycle counts and page-table walks

The evaluated input groups can have similar inference cycle counts while producing different page-table-walk footprints.
Counter signals on Haswell

The figure ranks the evaluated BERT-tiny hardware-counter signals by effect size.
Counter signals in a Vision Transformer

Four measured counters illustrate hardware-footprint differences for the evaluated Vision Transformer inputs.
































































































