The Invisible Security Paradigm: Protecting Data in Use and Confidential Computing in the Age of AI and Edge Computing

In today’s digital ecosystem, information has become the most critical strategic asset for companies, public administrations and industries. The widespread deployment of distributed architectures, ranging from the public cloud to Edge Computing nodes and Internet of Things (IoT) environments, has multiplied the capacity to collect, process and exchange data at scale. Added to this is the rapid progress of Artificial Intelligence (AI), whose analytical accuracy depends on the availability of large volumes of representative, high-quality data.

However, this dynamic creates a strategic paradox: the greater the need to share, analyse and collaborate on data in order to generate value, the greater the exposure of confidential information, user privacy and corporate secrets. Organisations are caught between two opposing demands. On the one hand, the need to collaborate in multi-organisation environments and third-party infrastructures; on the other, strict regulatory frameworks such as the General Data Protection Regulation (GDPR) and the AI Act, together with the constant risk of information leakage.

In this context, traditional security approaches are incomplete. Solving this challenge requires adopting the principle of Security & Privacy by Design, integrating technological safeguards from the initial design phase and throughout the entire data lifecycle. In this scenario, data protection and confidential computing are emerging as fundamental pillars for guaranteeing confidentiality, privacy and digital sovereignty without sacrificing analytical value.

The fundamental problem: from data at rest to the challenge of “data in use”

To understand the importance of advanced solutions, it is necessary to examine how information protection has evolved. Traditionally, cybersecurity has effectively addressed two of the three states in the data lifecycle:

Data at Rest: Information stored on servers, databases or storage systems, protected through encryption and access controls.

Data in Transit: Information moving across local networks or global communications, protected through cryptographic communication protocols.

However, the real blind spot in information security has historically been Data in Use. For IT infrastructure to run a program, perform data analysis or train an AI model, the information traditionally has to be decrypted in order to be processed in system memory. It is precisely during this active processing stage that data becomes exposed to vulnerabilities.

Some of the threats affecting data while it is being processed include:

• Untrusted environments and third-party infrastructures: When sensitive workloads are deployed in public cloud or shared environments, organisations lose direct control over the physical and logical execution environment.
• Exposure in RAM: Decrypted information held in memory during active processing may be vulnerable to unauthorised analysis or improper access in multi-user infrastructures.
• Barriers to inter-organisational collaboration: The inability to guarantee that data will not be viewed or copied during processing has historically hindered shared analytics initiatives between organisations in sensitive sectors such as healthcare, banking and industry.

Addressing this problem requires technologies capable of protecting information while it is being processed and objectively verifying where and how sensitive workloads are executed.

Some of the key challenges and threats affecting data protection include:

Exposure of sensitive data during AI training

Artificial Intelligence requires large datasets. When these datasets contain personal information, medical records or banking data, there is a risk that models may learn or memorise confidential information. This creates a security risk because third parties may be able to extract private data through inference attacks.

Loss of data sovereignty in cloud and edge computing

In distributed architectures where data is processed at the edge of the network — through Edge and IoT nodes — or across multiple clouds, maintaining traceability and data sovereignty becomes increasingly complex. The absence of mechanisms to verify where and how workloads are being executed leaves organisations without sufficient assurance during audits.

Conflict between regulatory privacy and analytical utility

Compliance with regulations such as the GDPR imposes strict data minimisation principles. Traditional anonymisation, when applied in a basic or rigid manner, can destroy statistical relationships and reduce the analytical value of data. The challenge is to apply adaptive anonymisation techniques that preserve data utility without compromising the identity of data subjects.es. 

Gradiant’s response: advanced data protection and confidential computing solutions

To provide a comprehensive response to these challenges, Gradiant researches and develops advanced solutions combining cryptography, privacy-preserving technologies and trusted execution environments. This enables an optimal balance between privacy and data utility.

Confidential and verifiable computing: TrusTEE

To protect data in use and address the lack of trust in underlying infrastructure, Gradiant offers TrusTEE, a confidential computing platform based on hardware-level Trusted Execution Environments (TEEs).

• Protection of Data in Use: TrusTEE isolates sensitive workloads within a secure memory enclave, ensuring that neither the host operating system, the hypervisor nor the cloud infrastructure administrator can access the data or code while it is being executed.
• Remote Attestation and Trust Policies: The platform performs a remote attestation process that cryptographically verifies the integrity of the code and the authenticity of the environment before authorising the release of data or cryptographic keys for processing.
• Verifiability of Results: TrusTEE provides execution results together with verifiable technical evidence, guaranteeing where and how the information was processed and ensuring that the results have not been altered. This is particularly relevant in AI scenarios, shared medical analytics and the processing of sensitive information.

Collaborative AI and privacy preservation: PAC-AI

Gradiant addresses the security requirements of Artificial Intelligence model training across different organisations through PAC-AI, without compromising corporate confidentiality or user privacy.

• Privacy-Preserving Machine Learning (PPML): Based on Secure Federated Learning techniques and advanced cryptography, PAC-AI enables multiple organisations to train shared analytical models without the need to centralise or exchange the original data.
• Full control over data: Each organisation keeps its datasets within its own local systems and under its direct control. Only cryptographically protected model update parameters are shared, reducing the risk of private information leakage.

Adaptive and intelligent anonymisation: AnonymiX

For the secure exchange and use of datasets in analytics or Open Data initiatives, Gradiant offers AnonymiX.

• Preserving value and privacy: AnonymiX is an intelligent and adaptive anonymisation solution designed to transform sensitive data by eliminating the risk of user re-identification while preserving the essential statistical properties required for analysis.
• Privacy-Enhancing Technologies (PETs): It integrates privacy-enhancing technologies that allow the level of protection to be adapted according to the type of data, the context of use and the applicable regulatory requirements, ensuring maximum analytical utility with minimum risk.

Ensuring data sovereignty without slowing innovation

Data protection and confidential computing are fundamental pillars for building the digital society of the future. The ability to process confidential information without exposing it to the underlying infrastructure removes traditional barriers to collaboration between companies, institutions and public administrations.

By integrating verifiable confidential computing, privacy-preserving Artificial Intelligence, post-quantum cryptographic management and sovereign identity, Gradiant provides the tools needed to address both current and future cybersecurity challenges.

This approach not only protects information assets against sophisticated threats, but also enables organisations to innovate with greater confidence and regulatory compliance.