Authors: Tsz On Li, Dong Huang, Xiaofei Xie, Heming Cui

Published on: May 15, 2024

Impact Score: 7.8

Arxiv code: Arxiv:2405.09314

Summary

• What is new: Themis introduces an automatic workflow to achieve full coverage of fault-inducing data flows in DLSs, improving fault detection significantly.
• Why this is important: Current DLS testing methods require manual effort and detect only a small proportion of faults.
• What the research proposes: Themis automates the process with a novel workflow that systematically identifies data flows likely to induce faults.
• Results: Themis detected 3.78 times more faults than other techniques and increased DLS accuracy by approximately 14.7 times after retraining with detected faults.

Technical Details

Technological frameworks used: Deep Learning Systems (DLS)

Models used: Themis

Data used: Perturbed inputs

Potential Impact

Automotive (autopilot systems), aerospace, medical devices, and other industries relying on safety-critical DLS applications.

Want to implement this idea in a business?

We have generated a startup concept here: FlowSecure.