The integration of Inception-ResNet into autonomous robotics and
manufacturing systems has significantly enhanced decision-making
and operational efficiency. However, vulnerabilities in the model, such
as susceptibility to adversarial attacks and reduced performance in
dynamic environments, pose critical challenges for secure and
transparent deployment. Ensuring robustness and reliability is
essential for avoiding costly errors and maintaining operational safety
in high-stakes applications. This work presents a comprehensive
framework for secure and transparent deployment of Inception-ResNet
by addressing its vulnerabilities. The proposed approach involves
incorporating an adversarial training pipeline, optimized for
autonomous robotics scenarios, and a blockchain-based logging
mechanism to enhance transparency and traceability. Additionally,
performance optimization is achieved through hyperparameter fine-
tuning and the integration of dropout layers to reduce overfitting. The
model is evaluated on a benchmark robotics dataset comprising 50,000
samples, split into 70% training and 30% testing datasets, to assess its
performance. The proposed framework demonstrates significant
improvements in both security and performance metrics. The enhanced
model achieves an accuracy of 96.4%, a 4.7% increase compared to the
baseline, with a robustness score against adversarial attacks improving
from 73.2% to 89.6%. Deployment transparency is reinforced through
blockchain implementation, ensuring data integrity and reducing
unauthorized model access attempts by 92%. These results underline
the potential of the proposed framework to set a new standard for
deploying Inception-ResNet in critical robotics and manufacturing
applications.
John Chembukkavu1, P.C. Remya2, M.K. Rachana3, P. Santhi4, P.S. Suvitha5
IES College of Engineering, India1,2,3,4,5
Inception-ResNet, Adversarial Training, Blockchain Transparency, Autonomous Robotics, Secure Deployment
| January | February | March | April | May | June | July | August | September | October | November | December |
| 4 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| Published By :
ICTACT
Published In :
ICTACT Journal on Soft Computing
( Volume: 15 , Issue: 3 , Pages: 3589 - 3597 )
Date of Publication :
January 2025
Page Views :
60
Full Text Views :
4
|