Can your Private Data b e Secured in the age of Machine Learning?
Can we trust artific ial intelligence to protect our privacy? LLMs like ChatGPT and Claude are trained with both public and private data. How can such a model be designed with privacy by design? These are the questions posed by computer security professor at Centrum Wiskunde & Informatica (CW I) Marten van Dijk in his inaugural lecture. Cryptographic methods li ke Differential Privacy and PAC Privacy offer partial solutions, but they have limitations. Can these be formally described mathematically ? And can training algorithms be adapted to maintain accuracy while g uaranteeing privacy?
Besides privacy, poison and evasion attack s also play a role in undermining the security of AI. There is a need for rigorous comparison of defense mechanisms, something that is oft en lacking. Fairness by design is also crucial: it has been mathemati cally demonstrated that not all fairness definitions are compatible. How can LLMs still be used fairly and responsibly in decision-making?
Ultimately, we strive for Trustworthy AI: systems that are acc urate, secure, fair, and explainable, according to Van Dijk.
DESCRIPTION: Can we trust artificial intelligence to protect our priva cy? LLMs like ChatGPT and Claude are trained with both public and pri vate data. How can such a model be designed with privacy by design? T hese are the questions posed by computer security professor at Centru m Wiskunde & Informatica (CWI) Marten van Dijk in his inaugural l ecture. Cryptographic methods like Differential Privacy and PAC Priva cy offer partial solutions, but they have limitations. Can these be f ormally described mathematically? And can training algorithms be adap ted to maintain accuracy while guaranteeing privacy? Besides privacy, poison and evasion attacks also play a role in undermining the secur ity of AI. There is a need for rigorous comparison of defense mechani sms, something that is often lacking. Fairness by design is also cruc ial: it has been mathematically demonstrated that not all fairness de finitions are compatible. How can LLMs still be used fairly and respo nsibly in decision-making? Ultimately, we strive for Trustworthy AI: systems that are accurate, secure, fair, and explainable, according t o Van Dijk. Can your Private Data be Secured in the age of Machine Le arning? END:VEVENT END:VCALENDAR