Case Study: Cisco improves network security machine learning with Julia Computing

Cisco - Customer Case Study

Cisco, a global developer and manufacturer of network hardware and telecommunications equipment, needed a faster way to tailor machine learning algorithms for computer security. Researcher Tomas Pevny wanted to build scalable, optimizable message-passing models for graph data to help detect infected computers and improve network security, and he used Julia and the Flux.jl machine learning library as his primary prototyping tools.

With Julia Computing’s Julia language and the Flux.jl ecosystem, Cisco was able to consolidate development into a single language instead of combining Python, TensorFlow, and C. Cisco extended Flux through Mill.jl for nested multi-instance learning and used Julia’s sparse matrix support to optimize one-hot encoding, delivering a 7.5x performance improvement in just two lines of code. According to Cisco, this made rapid prototyping much more efficient and scalable.

Cisco

Tomas Pevny

Prague and Technical Lead