The Kafka Summit brought together industry experts, developers, and enthusiasts to discuss the latest advancements and practical applications of event streaming and microservices. In this article, our CTO Sabri Skhiri delves into some of the keynotes and talks that took place during the summit.
Continue readingFor Large Language Models (LLMs), Azure confidential computing offers TEEs to protect data integrity throughout various stages of the LLM lifecycle, including prompts, fine-tuning, and inference. This ensures that all data used by LLMs remains encrypted and protected within the TEE, enhancing overall data security and privacy.
The recent Privacy Enhancing Technologies (PETs) London Conference convened stakeholders across Legal, IT, and various business domains to delve into the evolving data privacy and security landscape. With a focus on collaborating and sharing data through cutting-edge PETs, the summit provides valuable insights into how organisations can optimise the value of their data while ensuring privacy and security. Here, we distil the key themes and insights shared during the conference, shedding light on the interesting talks.
Tags: <privacy, synthetic data, encryption, sharing data >
The PET Summit 2024 showcased significant advancements and key trends in Privacy Enhancing Technologies (PETs), reflecting the growing importance of data privacy and security in the digital age. Major themes included:
Data privacy is a highly regulated landscape, riddled with ongoing legal and political disputes regarding its interpretation. While regulations like GDPR (General Data Protection Regulation) strive to provide clear guidelines, companies face the ongoing challenge of balancing compliance with innovation. This keynote emphasised the importance of a risk-based approach to data privacy, advocating for measures that safeguard individual rights while enabling data movement within legal frameworks.
The keynote acknowledges the challenges but doesn’t offer definitive answers. Instead, it serves as a call to action for ongoing dialogue and collaboration. By fostering open discussions and exploring creative solutions, stakeholders can translate the fundamental right to privacy into a more secure and responsible digital ecosystem. Moving forward, collaboration across legal, IT, data scientists, and business users is essential to bridge the gap between regulations and practical implementation.
PwC Germany showcased their journey in leveraging PETs, particularly anonymisation solutions from Anonos to unlock client value. Their approach involved gathering user requirements, conducting successful proof-of-concepts, and implementing use case-based strategies.
Challenges included communication gaps between stakeholders (understanding stakeholders’ perspectives across diverse fields with differing terminologies and expectations) and the need for agile delivery to be able to meeting the business’s need for rapid implementation.
Another hurdle involved catering to various use cases requiring different anonymisation approaches. This is where Anonos data embassy proved valuable, offering a comprehensive suite of anonymisation methods.
This session delved into the legal implications of utilising PETs, focusing on privacy principles like transparency, anonymisation, and data sharing:
This highlights the need for a clear understanding of which legal frameworks apply and how they translate to specific constraints, IT requirements, roles, and responsibilities. A balanced approach, combining PETs with robust legal frameworks and clear, user-friendly communication, is essential for fostering responsible data use.
Mastercard’s collaboration with the Singaporean government showcased the potential of Homomorphic Encryption (HE) for secure information sharing in financial crime investigation. The core use case focused on enabling banks to share financial crime intelligence across borders while guaranteeing robust privacy protection.
The technical details involved utilising a Multi-Party Computation (MPC) encryption scheme, allowing banks to verify the validity of International Bank Account Numbers (IBANs) without revealing any sensitive user or transaction data. However, scaling this solution even within the limited scope of financial crime investigation presented significant challenges, including computational intensity and regulatory complexities. While HE enables cross-border data exchange, questions lingered regarding its effectiveness in resolving data privacy concerns.
Azure Confidential Computing tackles the critical gap in data security by focusing on protecting data during processing within the cloud environment. It utilises Trusted Execution Environments (TEEs), such as Intel SGX, to create secure enclaves within the CPU and memory. This ensures that sensitive data remains encrypted and isolated from the main operating system, even during active processing.
The increasing reliance on data for AI development raises critical questions about data privacy and sovereignty. Azure confidential computing enhances data security by encrypting it at rest, in transit, and in use, which protects data throughout the AI lifecycle, and ensures users maintain control over their data, complying with various regulations.
On top of these considerations, Microsoft proposes a multi-layered approach to achieving responsible AI with strong privacy safeguards:
For Large Language Models (LLMs), Azure confidential computing offers TEEs to protect data integrity throughout various stages of the LLM lifecycle, including prompts, fine-tuning, and inference. This ensures that all data used by LLMs remains encrypted and protected within the TEE, enhancing overall data security and privacy.
The panel discussed the balance between the value of data and associated risks, emphasising principles like lawfulness, trustworthiness, and privacy. They highlighted the importance of data governance as a foundational PET and advocated for a combined approach of data governance with a risk-based assessment. Challenges to PET adoption included:
This panel emphasised the crucial role of cross-functional collaboration in successfully implementing PETs. It addressed stakeholder involvement, technology gaps, and the need to shift focus from technical implementation to unlocking value across different functions. Mastercard’s experience showcased the importance of a dedicated PET team and understanding industry-specific regulatory constraints.
The panel explored the significant role of data in unlocking the potential of PETs across sectors. It highlighted the shift in perspective within healthcare to view data as an opportunity and discussed strategies for navigating collaboration and value creation.
Building trust in data processing, particularly regarding transparency and explainability, was identified as a crucial challenge. LLMs were recognised as potential catalysts for PET adoption, but the lack of clear technical implementation guidelines hinders regulatory impact.
The PET Summit 2024 provided valuable insights into the technical challenges and opportunities surrounding PET adoption. Azure confidential computing exemplifies Microsoft’s commitment to enhancing data privacy and security, particularly in the context of AI. However, challenges such as complexity, regulatory tension, and the need for cross-functional collaboration persist. Clear guidance and standards are essential to drive PET adoption and ensure responsible data practices across industries. As organisations continue to navigate the evolving landscape of data privacy and security, collaboration, transparency, and innovation will be key to unlocking the full potential of PETs. Collaboration across disciplines and a balanced approach that integrates PETs with robust legal frameworks are essential for navigating the evolving landscape of data privacy and security effectively.
Our CTO, Sabri Skhiri, recently travelled to Sorrento for IEEE Big Data 2023. In this article, Sabri explores for you the various keynotes and talks that took place during the conference, highlighting the noteworthy insights and the practical applications shared by industry leaders.
Continue readingPresent study highlights the significance of dataset size in ICI microbiota models and presents a methodology to enhance the performances of a multi-cohort-based ML approach.
Continue readingThis document presents internships supervised by our consulting department or by our research & development department. Each project is an opportunity to feel both empowered and responsible for your own professional development and for your contribution to the company.
Continue readingIn this paper, we study graph representation learning and show that data augmentation that preserves semantics can be learned and used to produce interpretations. Our framework, which we named INGENIOUS, creates inherently interpretable embeddings and eliminates the need for costly additional post-hoc analysis.
Continue readingIn this paper, we present SANGEA, a sizeable synthetic graph generation framework that extends the applicability of any SGG to large graphs.
By first splitting the large graph into communities, SANGEA trains one SGG per community, then links the community graphs back together to create a synthetic large graph.
Les modèles d’apprentissage de représentations sont de plus en plus utilisés, mais des modèles d’IA explicables et de confiance sont nécessaires. Ce travail présente l’adaptation aux séries temporelles d’une méthode d’interprétation de représentation initialement conçue pour les images.
Continue readingEarly identification of patients with relapsing follicular lymphoma (FL) is critical but remains elusive. We initiated a collaboration between the academic CALYM Carnot Institute aiming at developing interpretable artificial intelligence (AI) models based on PET images to predict POD24.
Continue readingIn this paper, we study the behaviour of the triplet loss and show that it can be exploited to limit the biases created and perpetuated by machine learning models.
Continue reading