Search Results - adversarial+attacks

2 Results Sort By:
Decorrelation Mechanism and Dual Neck Autoencoders for Deep Learning
RPI ID: 2022-018-201 Innovation Summary: A dual-neck autoencoder architecture is introduced to improve feature separation and reduce redundancy in deep learning models. The decorrelation mechanism embedded in the encoder layers enhances generalization by minimizing overlap in learned representations. This design allows for more effective training in...
Published: 3/10/2026   |   Inventor(s): Christopher Wiedeman, Ge Wang
Keywords(s): Adversarial Attacks, Attack Transferability, computer vision, Decorrelation, deep learning
Category(s): Computational Science and Engineering
Synthesize Effective Ensemble Based Dynamic Defenses to Adversarial Attacks
Reference #: 01490 The University of South Carolina is offering licensing opportunities for Synthesize Effective Ensemble Based Dynamic Defenses to Adversarial Attacks Background: A lot of learning-based AI techniques are very prone to adversarial attacks. However, these attacks are detectable and there are algorithms that can defend against adversarial...
Published: 12/2/2025   |   Inventor(s): Biplav Srivastava, Jianhai Su, Ying Meng, Pooyan Jamshidi Dermani, Jason O'Kane
Keywords(s): Adversarial attacks, AI safety, Ensemble based defenses
Category(s): Software and Computing