Eliminating Zero-Day Chemical Threats (EZ-DCT)

Zero-day threats in the software industry are vulnerabilities that do not match known malware signatures, and upon discovery, developers have zero days to develop a patch. With the rapid progress in the fields of machine learning (ML) and computer aided drug design, algorithms can now be used to design novel chemicals that are optimized for harm. These computer-generated chemicals pose a “chemical zero-day” threat to detection equipment, whose algorithms are not trained to identify the potentially thousands of novel chemicals that bad actors could deploy. In Phase I we focused on using gas phase infrared (IR) spectra to demonstrate the feasibility of using ML to detect chemical zero-day threats. In Phase II we will extend the ML algorithm for condensed and gas phase IR and Raman spectra. Phase II will result in the development of two prototype products that will have wide applicability for chemical applications in academia, industry, and government. The two prototype deliverables of the Phase II project are: 1) the first novel, toxic IR and Raman reference spectral datasets for researchers and chemical equipment vendors to develop their own proprietary novel chemical detection algorithms, and 2) end-to-end novel, toxic chemical detection software that provides a rapid upgrade for existing chemical equipment. These deliverables maximize commercialization potential by 1) providing reference spectral datasets that can be licensed and 2) providing vendors with software to implement a state-of-the-art approach for detecting novel chemicals.