Full Program
Summary:
Cyberattacks pose a significant threat to food supply chains, which are essential infrastructure. The increasing digitization of the food industry, with IoT sensors and data collection throughout the supply chain, expands the attack surface.This paper proposes a three-pronged approach to securing digital food supply chains:
1. End-to-end encryption: Protecting data collected by sensors, especially on resource-constrained IoT devices, requires a model-based resource estimation framework to identify suitable security mechanisms.
2. Trust scoring: Addressing the risk of contradictory or false information from various sources by incorporating data quality and plausibility metrics into a trust scoring method.
3. Machine Learning (ML) pipeline security: Protecting ML models used in critical applications like food quality assessment from attacks like data poisoning. This involves a training-time data separation technique to identify and mitigate backdoor effects.
Author(s):
Marten Fischer
Germany
Ralf Tönjes
Germany
Rohit Bohara
asvin GmbH
Germany
Mirko Ross
asvin GmbH
Germany
Achyut Hegde
Karlsruhe Institute of Technology
Germany
Christian Wressnegger
Karlsruhe Institute of Technology
Germany
Matthias Brunner
Germany