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

Currently, there is a burgeoning demand for deploying deep learning (DL)models on ubiquitous edge Internet of Things (IoT) devices attributed to theirlow latency and high privacy preservation. However, DL models are often largein size and require large-scale computation, which prevents them from beingplaced directly onto IoT devices, where resources are constrained and 32-bitfloating-point (float-32) operations are unavailable. Commercial framework(i.e., a set of toolkits) empowered model quantization is a pragmatic solutionthat enables DL deployment on mobile devices and embedded systems byeffortlessly post-quantizing a large high-precision model (e.g., float-32) intoa small low-precision model (e.g., int-8) while retaining the model inferenceaccuracy. However, their usability might be threatened by securityvulnerabilities.  This work reveals that the standard quantization toolkits can be abused toactivate a backdoor. We demonstrate that a full-precision backdoored modelwhich does not have any backdoor effect in the presence of a trigger -- as thebackdoor is dormant -- can be activated by the default i) TensorFlow-Lite(TFLite) quantization, the only product-ready quantization framework to date,and ii) the beta released PyTorch Mobile framework. When each of the float-32models is converted into an int-8 format model through the standard TFLite orPytorch Mobile framework's post-training quantization, the backdoor isactivated in the quantized model, which shows a stable attack success rateclose to 100% upon inputs with the trigger, while it behaves normally uponnon-trigger inputs. This work highlights that a stealthy security threat occurswhen an end user utilizes the on-device post-training model quantizationframeworks, informing security researchers of cross-platform overhaul of DLmodels post quantization even if these models pass front-end backdoorinspections.

Quantization Backdoors to Deep Learning Commercial Frameworks