Chapter Summary

Chapter Summary

Key Points

  • 1.

    Secure delivery model (Sengupta-Tandon-Clancy 2015): external eavesdropper observes the broadcast XX. Goal is I(W;X)=0I(\mathcal{W}; X) = 0 β€” perfect info-theoretic secrecy of the library.

  • 2.

    Minimum memory requirement: Mβ‰₯1M \geq 1. One file's worth of random key material per user is necessary (cut-set/OTP lower bound).

  • 3.

    STC achievable rate Rsec=K(1βˆ’(Mβˆ’1)/N)/(1+K(Mβˆ’1)/N)R_\text{sec} = K(1 - (M-1)/N) / (1 + K(M-1)/N) β€” MAN rate with effective memory Mβˆ’1M - 1.

  • 4.

    Two-tier placement underpins STC: tier 1 stores 1-file key material (via Shamir sharing); tier 2 stores (Mβˆ’1)(M-1)-file MAN subfiles. Delivery XORs MAN messages with fresh keys.

  • 5.

    Secrecy is not free. Unlike demand privacy (Wan-Caire, Ch 12), which costs nothing, secrecy costs 1 file of memory. At scale (N≫1N \gg 1), the rate penalty is small.

  • 6.

    Shamir secret sharing achieves the Mβ‰₯1M \geq 1 bound exactly by compactly storing shares such that (t+1)(t+1) users can reconstruct keys, tt users learn nothing.

  • 7.

    Joint privacy + secrecy schemes combine Wan-Caire masking with STC XOR keys β€” full protection against both insiders and outsiders, at MAN-with-effective-memory rate.

Looking Ahead

Chapter 18 covers multi-access coded caching, where each user accesses multiple caches (not just one). Cyclic wrap-around access and combinatorial design schemes extend MAN to the multi-access setting. This completes Part IV's survey of coded- caching extensions before moving to open research problems in Part V (Ch 19-22).