A Common Language for Data Privacy with Jack Fitzsimons

The Eyes Off Data Summit 2023 hosted by Oblivious in Dublin was an event that united data enthusiasts, policy and regulatory experts, security professionals, and PETs developers who are committed to staying at the forefront of the rapidly evolving AI landscape and promoting responsible innovation.

Kicking off the event, Jack Fitzsimons from Oblivious delivered an impactful talk with the goal of aligning all stakeholders on the urgent need for a common language in data privacy. This talk positioned the event as the starting point for a community-driven effort to establish that common language.

Dive into this blog article to get a comprehensive overview of the insights and key takeaways from his talk.

The Status Quo

In today’s data-driven world, where personal information is scattered across various organisations and privacy concerns abound, the urgency to safeguard sensitive data has never been greater.

The current status quo is that sensitive personal information is often locked away in silos across multiple organisations, making it difficult to access and analyse. This not only hinders the work of data scientists but also limits the ability of businesses to leverage this valuable data for insights and growth.

Even if Personally Identifiable Information (PII) is masked or hidden, sensitive information can still be pieced together. This risk, together with the ability of individuals to access aggregate data can lead to the reverse engineering of sensitive individual data, posing a threat to privacy.

The lack of trust and transparency in current data privacy practices adds to compliance challenges and exacerbates privacy concerns.

PETs As An Ideal Solution

To address these issues, a clearly defined trust model enforced by technology is needed, one that provides transparency regarding who inputs, receives, and processes data.

Privacy Enhancing Technologies (PETs) emerge as one such solution, enabling the secure and efficient integration of data from diverse sources while ensuring that data usage adheres strictly to the intended purpose.

By preventing the reverse engineering of sensitive information when disseminating data outputs, PETs provide powerful tools to navigate the complex landscape of data privacy.

Zooming In On Different PETs

To better understand how PETs work, it’s essential to differentiate between input and output privacy, as both are needed to preserve data privacy effectively.

Input privacy approaches such as Multiparty Computation and Homomorphic Encryption (MPC/HE) involve encrypting and performing arithmetic and/or boolean operations on data but can be computationally expensive and communication-heavy.

On the other hand, Secure Enclaves follow a different security model based on isolating CPUs and memory. They allow individuals to verify how data will be processed and follow standard security tooling and certifications.

Output privacy methods like Statistical Disclosure Controls involve adding noise, swapping, and aggregating data, but lack formal proofs and guarantees.

Differential Privacy adds calculated noise to data to impede individual identification and has mathematical guarantees around privacy loss.

Synthetic Data generates artificial data based on the statistical model trained on the original dataset, while Federated Learning involves local agents computing on data.

Each PET type has its benefits and challenges and is most appropriate for different scenarios.

PETs Community to Establish a Common Language in Data Privacy

Unlocking the true potential of Privacy-Enhancing Technologies (PETs) and effectively safeguarding sensitive data hinges on collaboration and shared learning.

By showcasing expertise, perspectives, challenges, and opportunities, we can develop comprehensive solutions that balance technical details and broader implications, accessible to a wide audience ranging from experts to researchers, regulators, and enthusiasts alike, which is exactly what Eyes Off Data Summit aimed to do.

Through collaboration, we can establish ethical standards, build trust, and address concerns related to PETs’ responsible implementation. Such collaboration ensures that PETs provide benefits while minimising potential risks, striking a balance that promotes the responsible use of sensitive data and preserves privacy.

This is vital in today’s digital era, where data breaches and unethical use of data have become recurrent and advanced technologies like AI continue to multiply sources of collected data.

You can watch the recording of Jack’s talk at the Eyes Off Data Summit 2023 on our YouTube channel here.