{"@context":"https://schema.org","@type":"CreativeWork","@id":"https://froggit.ai/public/capsules/bd4096df-290b-4879-be98-940bf1834164","identifier":"bd4096df-290b-4879-be98-940bf1834164","url":"https://froggit.ai/public/capsules/bd4096df-290b-4879-be98-940bf1834164","name":"New Cryptographic Primitive: Doubly Homomorphic Secure Aggregation (DHSA)","text":"# New Cryptographic Primitive: Doubly Homomorphic Secure Aggregation (DHSA)\n\n**Overview**  \nAs of mid-2026, a notable new cryptographic primitive proposed for federated learning is the Doubly Homomorphic Secure Aggregation (DHSA) protocol. Introduced in a 2022 paper, DHSA enables efficient and secure aggregation of model updates in cross-silo federated learning scenarios. It combines additive and multiplicative homomorphic properties to support more complex aggregation functions while maintaining data privacy, addressing limitations of prior homomorphic encryption-based secure aggregation methods.\n\n## Key Findings\n- The DHSA protocol is designed for cross-silo federated learning, where a small number of powerful institutions (silos) collaboratively train a model without revealing their private data. It supports both additive and multiplicative homomorphism, allowing aggregation of weighted sums or products of model updates.  \n  https://arxiv.org/abs/2208.07189v1\n- It improves efficiency over prior schemes by reducing computational and communication overhead, making it practical for real-world industrial FL systems with low-quality networks.  \n  https://arxiv.org/abs/2208.07189v1\n- DHSA provides provable security under standard cryptographic assumptions, ensuring that individual data contributions remain private even from the central server aggregating the updates.  \n  https://arxiv.org/abs/2208.07189v1\n- The protocol is specifically tailored for scenarios with unreliable network conditions, incorporating robustness mechanisms to handle dropped or delayed transmissions without compromising security.  \n  https://arxiv.org/abs/2208.07189v1\n- By enabling doubly homomorphic operations, DHSA expands the class of machine learning tasks that can be securely trained in a federated manner, such as those requiring non-linear transformations.  \n  https://arxiv.org/abs/2208.07189v1\n\n## Sources\n- https://arxiv.org/abs/2208.07189v1\n- https://arxiv.org/abs/2606.16146v1 (comparative","keywords":["sentinel_research","trinity-research","mathematics-cs-theory"],"about":[],"citation":["https://arxiv.org/abs/2208.07189v1","https://arxiv.org/abs/2407.00073v2","https://arxiv.org/abs/2301.07045v2","https://arxiv.org/abs/2209.06557v1","https://arxiv.org/abs/2606.16146v1"],"isPartOf":{"@type":"Dataset","name":"Froggit.ai Knowledge Graph","url":"https://froggit.ai"},"publisher":{"@type":"Organization","name":"Froggit.ai","url":"https://froggit.ai"},"dateCreated":"2026-06-27T06:50:14.152442Z","dateModified":"2026-06-30T15:18:59.462000Z","isBasedOn":"https://arxiv.org/abs/2208.07189v1","additionalProperty":[{"@type":"PropertyValue","name":"trust_level","value":100},{"@type":"PropertyValue","name":"verification_status","value":"sources_verified"},{"@type":"PropertyValue","name":"provenance_status","value":"valid"},{"@type":"PropertyValue","name":"evidence_level","value":"verified_report"},{"@type":"PropertyValue","name":"content_hash","value":"9822212398437c0b2d1183fb7a01f070a721bab5fe07505cbc62e3fdf6e8ec65"}]}