The significant computational complexity of Fully Homomorphic Encryption (FHE) has prompted numerous accelerator designs. However, existing FHE accelerators often implicitly assume that all computations are executed reliably, overlooking the fact that modern FHE schemes can be highly vulnerable to hardware faults: even a single-bit error in a ciphertext can cascade into widespread plaintext corruption.

Embedding aggregation in large-scale recommendation systems creates severe memory bandwidth bottlenecks, as each query sums many high-dimensional vectors. Bitwise-operation-based PIM can exploit subarray bandwidth, but traditional designs struggle with summation because long carry propagation limits parallelism. We propose PIM-SUM, an in-DRAM summation primitive for Sparse Length Sum (SLS) in recommendation workloads. PIM-SUM reformulates vector summation as column-wise accumulation via popcount, truncating carry propagation and avoiding redundant in-DRAM computation. This enables highthroughput summation using native DRAM bitwise primitives. PIM-SUM integrates a Reed-Solomon-inspired error correction at the DRAM row level. Its linearity supports parity propagation, enabling integrity checking and multi-bit error correction with low overhead. On DLRM workloads, PIM-SUM achieves up to$5.14 \times$speedup, logarithmic I/O reduction, and large energy savings. It also reduces silent data corruption by$1778 \times$and improves detection by over$170 \times$. These results show PIM-SUM is a scalable, fault-tolerant, and energy-efficient solution for memory-bound inference at data center scale.

This work proposes an On-Off Keying (OOK) coding scheme for covert communication over complex Gaussian channels. In particular, a transmitter Alice employs phase deflection to covertly transmit information to a receiver Bob, simultaneously ensuring that the communication intent is concealed from a warden Willie. The utilization of phase deflection allows Alice to improve the transmission rate by leveraging Willie’s uncertainty about the received phase, without changing the codebook construction. Considering the asymmetry of the OOK codebook’s input distribution and shape constellation, we first analyze the relationship between the input distribution and the signal amplitude, and then propose a scheme that can achieve covert transmission with the input distribution of the “on” symbol$a_{n}=\mathcal {O}\left ({{\frac {1}{\sqrt {n}}}}\right)$and an average transmission power$\beta ^{2}=\mathcal {O}({1})$. We quantify the improvement brought from the phase resource as phase deflection gain and derive its closed-form expression by approximating the Kullback-Leibler (KL) divergence and mutual information through Taylor expansion. Numerical results show that our scheme achieves significant phase deflection gain, and the maximum gain can be achieved by fully utilizing the phase resources through three stages.