Cryptography and Machine Learning

I am vacationing in Madrid, Spain, enjoying the jet lags on top of the excitement of sight seeing, walking and eating. To endure the sleepless nights I decide to read the machine learning books I never had the time to read, and post my thoughts in this journey. How are cryptography and machine learning related? On the surface, these two disciplines may appear unrelated, but a closer look reveals some fascinating similarities: 1. Foundations in Mathematics: Both cryptography and machine learning are rooted in sophisticated mathematical algorithms. The purpose of cryptographic algorithms is to conceal data’s meaning, while machine learning aims to extrapolate and interpret it. A small group of engineers typically develop these algorithms, while the majority of practitioners focus on designing and implementing distributed services so that these algorithms are accessible and provide value to the end-users. 2. Delivering Value: • Cryptography: The value offered to users hinges on (a) the security attributes of the keys driving the algorithms, including their entropy source, protection, and correct usage, (b) the durability of these keys (loss of keys == loss of data), and (c) the operational attributes of the service, such as availability and scalability. • Machine Learning: The emphasis here is on (a) the ML models efficiently addressing customer queries related to their data, whether for generative or discriminatory purposes, (b) the necessity for ML models to evolve with changing data – a concept not existing in cryptography, and (c) maintaining the service’s operational standards, including availability and scalability. 3. Hardware Dependency: Both services are intricately tied to hardware. Cryptography leans on hardware for security and optimization, while machine learning employs GPUs for large-scale parallel computations on floating vectors. 4. The necessity of building multi-disciplinary teams. For cryptography services we need a team of cryptographers, software engineers, data engineers and service reliability engineers. For machine learning services, just replace the cryptographers with machine learning scientists. Richard Sutton, a distinguished research scientist at DeepMind said “The biggest lesson that can be read from 70 years of AI research is that general methods that leverage computation are ultimately the most effective, and by a large margin.… Seeking an improvement that makes a difference in the shorter term, researchers seek to leverage their human knowledge of the domain, but the only thing that matters in the long run is the leveraging of computation.” Seeing how parallelism from quantum computing may change our perception of cryptography, I wonder similar claims can be made on cryptography. The book “Designing Machine Learning Systems” is a nice read. https://lnkd.in/g3mEhDjn