Security Summaries

Computer vision training system for robust artificial intelligence image classification
Tong, Liang ; Vorobeychik, Yevgeniy ; Wu, Tong
T-019303

— Technology Description Researchers in Prof. Yevgeniy Vorobeychik’s laboratory have developed a new adversarial model and training methods to defend deep neural networks against physical attacks that corrupt image classifications. This system outperforms previous state-of-the-art techniques a…

Error correcting codes for robust, secure deep neural networks
Raviv, Netanel
T-019252

— Technology Description Researchers in Prof. Netanel Raviv’s laboratory have developed a provable error correcting code (ECC) system that could guarantee deep neural networks (DNN) are resilient to noise for robust, secure and private neural computation. This technology could be deployed in n…

Blockchain-based algorithms for secure, collaborative risk assessment and decision making
Jain, Raj ; Salman, Tara
T-018710

— Technology Description Researchers in Prof. Raj Jain’s laboratory have developed a new type of algorithm that extends the security of decentralized blockchains from consensus validation to efficient group decision making. This technology, called “probabilistic blockchains”, provi…