References & Further Reading

References

  1. B. Choi, J.-y. Sohn, D.-J. Han, J. Moon, and G. Caire, Communication-Computation Efficient Secure Aggregation for Federated Learning, 2022

    The CommIT-group CCESA paper. Headline reference for this chapter — the sparse-graph construction, privacy and reliability analysis, and complexity bounds.

  2. K. Bonawitz, V. Ivanov, B. Kreuter, and others, Practical Secure Aggregation for Privacy-Preserving Machine Learning, 2017

    The complete-graph baseline (Chapter 10). CCESA's sparse-graph construction is structurally a modification of Bonawitz — same primitives (DH, Shamir), different graph topology.

  3. A. R. Elkordy, A. S. Avestimehr, and G. Caire, On the Information-Theoretic Optimality of Secure Aggregation in Federated Learning with Uncoded Groupwise Keys, 2022

    Chapter 10 §10.4. Establishes Bonawitz's $O(n^2)$ optimality within the uncoded groupwise-key class — the result that motivates CCESA's move outside the class.

  4. S. Kadhe, N. Rajaraman, O. O. Koyluoglu, and K. Ramchandran, FastSecAgg: Scalable Secure Aggregation for Privacy-Preserving Federated Learning, 2020. [Link]

    Predecessor to CCESA. Uses regular graphs to reduce overhead to $O(n \\log n)$, but at weakened privacy guarantees. CCESA improves over this with random graphs and full Bonawitz-equivalent privacy.

  5. A. Shamir, How to Share a Secret, 1979

    Foundational secret-sharing primitive used in CCESA's dropout-handling phase (inherited from Bonawitz §10.3).

  6. P. Erdős and A. Rényi, On Random Graphs I, 1959

    The Erdős–Rényi random-graph model used in CCESA. Classical reference; the connectivity-threshold result is the foundation of CCESA's reliability analysis.

  7. T. Jahani-Nezhad, M. A. Maddah-Ali, and G. Caire, Byzantine-Resilient Secure Aggregation for Federated Learning Based on Ramp Sharing, 2023

    ByzSecAgg (Chapter 11). Complementary to CCESA: handles Byzantine adversaries at higher cost; CCESA handles passive adversaries efficiently. Together they span the privacy-preserving FL design space.

  8. K. Bonawitz, H. Eichner, W. Grieskamp, and others, Towards Federated Learning at Scale: System Design, 2019. [Link]

    Google's production FL paper. Documents the $O(n^2)$ wall in production deployments and the $n \\leq 500$ ceiling for Bonawitz.

  9. P. Kairouz, H. B. McMahan, B. Avent, A. Bellet, and others, Advances and Open Problems in Federated Learning, 2021

    Comprehensive FL survey including secure-aggregation variants and their scaling.

  10. T. Boyd and S. Mathuria, Protocols for Authentication and Key Establishment, Springer, 2003

    Standard reference for cryptographic randomness protocols, including blockchain-based randomness for the graph-seed generation in §12.2.

  11. W. Diffie and M. E. Hellman, New Directions in Cryptography, 1976

    Foundational Diffie–Hellman key exchange, used in CCESA's sparse-graph DH phase.

  12. A. Frieze and M. Karoński, Introduction to Random Graphs, Cambridge University Press, 2015

    Standard textbook on random-graph theory, including the Erdős–Rényi model connectivity threshold and concentration results used in CCESA's analysis.

Further Reading

Resources for going deeper into CCESA, sparse- graph secure aggregation, and the privacy-preserving FL frontier.

  • Random-graph theory for protocol design

    Frieze and Karoński, *Introduction to Random Graphs*, Cambridge UP 2015

    Comprehensive treatment of random-graph analytical techniques. Essential for rigorously analyzing CCESA-style constructions.

  • Privacy-preserving ML frontier

    Kairouz et al., *Advances and Open Problems in FL*, FnT 2021

    Survey including secure-aggregation variants. Excellent context for placing CCESA in the broader research landscape.

  • Information-theoretic FL converse bounds

    Caire et al. 2022 IEEE T-IT (Chapter 10 §10.4)

    The optimality result that motivates CCESA's existence. Reading both papers together clarifies which scheme classes are bounded and which can be exceeded.

  • Practical secure aggregation deployments

    Bonawitz et al. 2019 MLSys (Google's production paper)

    Engineering perspective on the $O(n^2)$ wall and how Google navigates it. Useful context for understanding CCESA's production motivation.

  • Cryptographic randomness for protocol seeds

    NIST SP 800-90B; Boneh, Boneh, and Franklin, *Identity-Based Encryption from the Weil Pairing*, 2001

    For the graph-seed generation in CCESA, cryptographic randomness sources need to be honest and unpredictable. Useful for deployment.

ExercisesCh 13 →