Adoption of Tiny Machine Learning for Industrial Predictive Maintenance Solutions
Abstract
Predictive maintenance (PdM) is a result of the amalgamation of advanced versions of some of the most popular branches of technology, viz. Internet of Things (IoT) and machine learning (ML) have revolutionized equipment maintenance strategies across industries, thereby saving time, effort, and money by anticipating probable issues and avoiding taxing, abrupt, unplanned breakdowns. A common concern in applying machine learning techniques in real-world applications is the need for heavy computing resources and dependence on slow and cost-inefficient cloud services. By relying on TinyML, PdM overcomes this challenge, which is a much lighter version of ML suitable for edge computation. However, this solution is alarming for data security. Its limited memory and processing power render traditional security mechanisms redundant in such cases. The risk of data leakage, insecure interface exposure, revealing sensitive information stored in weights, and side-channel attacks also increases owing to its easy physical accessibility, adding to its challenges.
