USI – Models of selected homomorphic encryption, multi-party computation and FL open-source libraries suitable for continuous monitoring
| ID: R13 | Licence: GPL 3.0 |
Owner: USI |
Contacts: francesco.regazzoni@usi.ch subhadeep.banik@usi.ch tom.slooff@usi.ch |
| Short Description | Privacy and security as a driver of orchestration in the MIRTO cognitive engine. The Privacy & Security Manager is responsible for maintaining security and privacy according to the specification at run time: it recommends the nodes able to meet the security requirements of an application and updates the security primitives when conditions change. The Security & Privacy Proxy, attached to each node as a sidecar container, intercepts raw TCP connections and adds a layer of security on top of them, using implementations of security and privacy primitives and a synthesizer that selects the most appropriate ones based on models of their power/energy consumption and performance. | |
| Security & Privacy Proxy | Privacy & Security Manager | |
|---|---|---|
| Key features | – Transparent layer of security added to the application – Security and privacy primitive implementations – Model of security and privacy-preserving primitives exposing figures of merit – Synthesizer which selects the most appropriate primitive from those in adherence with the security and privacy specification | – Recommendation of the nodes able to meet the required security level (low, medium, high), also considering trust and Gaia-X restrictions – Updating security and privacy primitives at run time, to adhere to the given TOSCA specification – Aggregated trust score of each node, published through an API |
| Require | – Open-source libraries of security and privacy primitives – For transparent security: Linux environment and sidecar set-up in Kubernetes – For profiling: access to the RAPL interface and the LIKWID tool – Security and privacy specification (YAML) | TOSCA specification of security and privacy; Security & Privacy Proxies deployed as sidecars in the application (including the appropriate primitives inside). |
| Provide | – Model of the selected libraries reporting the figures of merit – Transparent security at TCP layer – Security and privacy primitives directly callable – Adherence to the security and privacy specification, also considering run-time updates to the specification | Adherence to the specification; if primitives are updated, this is reported. Recommended nodes and aggregated trust levels. |
| Input | Security and privacy specification. | TOSCA specification of security and privacy (security level, optionally target energy and performance); device-level information on the nodes; trust scores from the proxies. |
| Output | – YAML file with the measures of the figures of merit – YAML file with the concrete choice of primitives, determined from the specification and primitive profiling | The updates made to the primitives used by the Security & Privacy Proxies; the recommended nodes; the aggregated trust levels of the nodes. |
| User | – Privacy & Security Manager – Original applications with TCP-based connections | MIRTO Agent. |
| Benefits for the user | – Transparent layer of security added to the application – Enables performance-aware placement for security and privacy-preserving primitives | Adherence to the security and privacy specification. |
| Download | Please contact the developers listed above. | github.com/TSlooff/PS-Manager |
| Role in MYRTUS | Adherence to the security and privacy specification, performance-aware selection of security and privacy primitives. | Adherence to the security and privacy specification. |
| TRL@M0 | 0 | |
| TRL@M36 | 3 | |
General description
Security & Privacy Proxy
The proxy is a container deployed as a sidecar next to each component of an application. It automatically intercepts the outgoing TCP connections and secures them at the transport layer, so that the application above it is protected without modifications. Proxies authenticate each other with certificates issued by a MYRTUS certificate authority and agree on the primitives to use, choosing from a library of open-source implementations for encryption, hashing, digital signature and key exchange. A synthesizer selects the implementations that meet the specification, using profiling information on their energy consumption and throughput.
The proxy also tracks signals on its own integrity, on the behaviour of the node and of its peers, and on the compliance of its configuration, from which trust scores are computed. Through a volume shared with the application, it provides tools for privacy-preserving computing based on homomorphic encryption. A native version of the container, in which the designer provides the source code of the primitives, is also available (see Privacy and Security).
Privacy & Security Manager
The manager is responsible for maintaining security and privacy according to the specification at run time. It knows the security requirements of the application, the security and privacy functionalities available on each node, and the trust level of the nodes. When the MIRTO Agent deploys an application, it filters out the nodes that cannot meet the requested security level and recommends the suitable ones, preferring trusted nodes; further restrictions, such as legal, location or security restrictions coming from Gaia-X, can be applied in the same way.
At run time, the manager reacts to changes in the allocation of the application, in the available resources (for example energy) or in the security requirements themselves: it can issue a new recommendation, or push a new configuration of primitives to the proxies, which update them live without re-allocating the application. Security levels range from lightweight algorithms for constrained devices (low) to quantum-resistant algorithms (high).
MYRTUS Extension/Contribution
Security & Privacy Proxy
Within MYRTUS, the proxy provides a transparent and general way to integrate a large number of open-source implementations of security primitives into existing applications, and to change them at run time: after a configuration update, it re-initiates the handshake with its peers when necessary, without terminating the connection. For privacy-preserving computing, it provides a compiler and three homomorphic encryption back-ends, together with a model that predicts the execution time and energy of a program for different back-ends and configurations, so that the most suitable configuration can be chosen.
Privacy & Security Manager
The manager is developed together with the Security & Privacy Proxy, which adds security and privacy functionality to individual nodes. It allows the security primitives to be changed at run time by communicating directly with the proxies, through an authenticated interface available only to the manager. It also aggregates the trust scores reported by each proxy about itself and about its peers into a trust level for each node, which is used in its own recommendations and published for other managers.
Technology Benefits and Innovations
Privacy and security are often addressed through predefined policies or dedicated services, with limited integration into run-time orchestration decisions. In MIRTO, they become dynamic drivers of orchestration: the Privacy & Security Manager is part of the closed control loop and directly influences node selection and the configuration of the security primitives, which can be adapted to the required security level and to the available energy and performance. Unlike approaches that use trust only to select infrastructure providers, trust is assessed for each node and used in fine-grained placement decisions. The transparent proxy adds this security to existing TCP-based applications without changing their code.
Users
Developers and operators of distributed applications in the edge-fog-cloud continuum that need to meet security and privacy requirements at run time, for example applications with sensitive data or that must move towards quantum-resistant cryptography, without modifying their code. The result is also relevant for developers of orchestration platforms that want to take security, privacy and trust into account in placement decisions.
