1. Differential privacy (差分隱私 ,我看過交大/中山大學 官大智教授有研究這個): where noise is added to an analytical system so that it is impossible to reverse-engineer the individual inputs

其實Google / Apple都把它們的差分隱私的Library開源了. 我想未來的各類API應用, 恐怕得.... 隱私工程(Privacy Engineering) 還是得上場了, ISO那一套制度與驗證, 必須與隱私工程互為表裡才能有比較好的保障, 可別只看一張證書吧....

(上面這篇不錯)

2. Federated analysis (聯合分析) : where parties share the insights from their analysis without sharing the data itself

Homomorphic encryption: where data is encrypted before it is shared, such that it can still be analysed but not decoded into the original information

(我是看上面這兩篇, 歐盟的這個站值得瀏覽)...

3. Zero-knowledge proofs: where users can prove their knowledge of value without revealing the value itself

Secure multiparty computation: where data analysis is spread across multiple parties such that no individual party can see the complete set of inputs

(零知識驗證就不必說了, 現在作ML的, 這個是顯學...)

WEF世界經濟論壇這樣建議, 代表商用地產用需求就會有一波市場. 所以產學合作呢. 可以考量投入. (但必須說 BigTechs都已經有應用了....)

但是有能力稽核的, 我就不多贅述了....