#Hacktivity2021 // The History of Hacking AI Systems

Eugene Neelou - The History of Hacking AI Systems

This presentation was held at #Hacktivity2021 IT security conference on 8th October 2021.

Machine learning has proven its value in many industries. Yet, due to fast implementations, most real-world ML systems affecting our lives are catastrophically insecure. The state of ML security today is similar to computer security in the 90s.

To show the whole picture, this presentation will connect many signals from governments, academia, and industry in their efforts to secure AI systems. The talk is based on the research that covers the past 10 years of progress in adversarial machine learning.

While the field of adversarial ML is primarily driven by academia, the talk will also focus on the industry and cover the rising number of real-world vulnerabilities in AI systems, emerging ML security tools and frameworks, and the growth of AI red teams.

#HACKTIVITY is the biggest event of its kind in Central & Eastern Europe. About 1000 visitors are coming from all around the globe every year to learn more about the latest trends of cybersecurity, get inspired by people with similar interest and develop themselves via comprehensive workshops and training sessions.