English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Recent research has shown that bit-flip attacks (BFAs) can manipulate deepneural networks (DNNs) via DRAM Rowhammer exploitations. Existing attacks areprimarily launched over high-level DNN frameworks like PyTorch and flip bits inmodel weight files. Nevertheless, DNNs are frequently compiled into low-levelexecutables by deep learning (DL) compilers to fully leverage low-levelhardware primitives. The compiled code is usually high-speed and manifestsdramatically distinct execution paradigms from high-level DNN frameworks.  In this paper, we launch the first systematic study on the attack surface ofBFA specifically for DNN executables compiled by DL compilers. We design anautomated search tool to identify vulnerable bits in DNN executables andidentify practical attack vectors that exploit the model structure in DNNexecutables with BFAs (whereas prior works make likely strong assumptions toattack model weights). DNN executables appear more "opaque" than models inhigh-level DNN frameworks. Nevertheless, we find that DNN executables containextensive, severe (e.g., single-bit flip), and transferrable attack surfacesthat are not present in high-level DNN models and can be exploited to depletefull model intelligence and control output labels. Our finding calls forincorporating security mechanisms in future DNN compilation toolchains.

Unveiling Single-Bit-Flip Attacks on DNN Executables