Collaborative Self-Learning for Security, Fun and Profit
summary
Describes building a production federated learning platform for mobile cyber risk intelligence that runs across nine companies and trains on 24 million users, integrating industry partners like Cisco Intel and Huawei; shows a concrete pipeline where local risk models train on-device, upload updates, and produce a Federated risk model that other parties deploy immediately; reports that the system detects IoT/DoS and mobile app risks at scale, documents poisoning and model-leakage attacks, and provides engineering defenses that survive large-scale evaluation.
AI-written summary · about
chapters (14)
- 0:00Start
- 2:49Can Machine Learning Improve Security in a Meaningful and Sustainable Way
- 5:22Federated Learning
- 7:59Is Federated Learning Useful in Security Applications
- 9:16Can We Use Federated Learning for Intrusion Detection in a Large-Scale Iot Network against Denial of Service Attacks
- 9:50Sharing Cyber Risk Intelligence
- 10:12Mobile Cyber Risk Intelligence for Mobile Applications
- 11:06Cyber Threat Intelligence Systems
- 11:41Challenges
- 12:54What Is a Risk Analysis Framework in a Mobile Application
- 13:45Risk Categories
- 16:59Data Set Overview
- 19:28Problem of Data Leakage
- 20:58Conclusion
related talks
25mData collaboration with Federated Learning
27mCreate Value and Insight Through Privacy-preserving Data Collaboration [...]
38mFederated Learning and Its Application...
41mFederated Learning: Machine Learning on Decentralized Data
42mFederated Learning: Unlocking the Values of Discrete Data for Enterprise
27mTalk by Micah Sheller (Flower/ML Commons)
34mLFN AI Taskforce: Platform-Driven Fede...
44mFederated ML on iOS: Protecting User Data in the Age of AI