P031 Line-rate capture and hashing hardware

Can hardware hash, timestamp and sign every packet at full line rate with zero loss, and at what resource and power cost?

Line-rate capture, hashing and signing hardware.

Terms: FPGA (field-programmable gate array): a chip whose logic can be reconfigured, often used for line-rate packet processing. DPU / SmartNIC: a network card with its own processor that can inspect and process traffic.

Status

Started The authors' own CPU/DPDK benchmark at 400 Gb/s: line-rate hashing for frames of 1500 B and above; at 64 B minimum frames no algorithm reaches line rate within the physical core count; hashing only, with no timestamping or signing, zero loss on real datacenter traffic not shown, and no trusted or simple hardware; not 800G or 1.6T (limited hardware availability); the authors say an FPGA or ASIC is likely needed. Peigne-Lefebvre et al. (2026) still pose formally verified FPGA hashing at 400+ Gbps as open.

For realistic traffic profiles at 1500B and above, AES-128-GMAC and SipHash-1-3-128 are both capable of sustaining line-rate hashing
Reynolds & Foster (2026), Network Traffic Hashing (Amodo Design note), DPDK Summary (2026-07-03), L547

Posed by

Also raised in: Baker, Kulp, Marks, Brundage & Heim (2025), Verifying International Agreements on AI: Six Layers of Verification for Rules on Large-Scale AI Development and Deployment (RAND WR-A4077-1); Cankaya (2026), A System Overview for Near-Term, Low-Trust AI Compute Verification (v0.2); Cankaya (2026), All Hands on Deck to Build the Datacenter Lie Detector; Canary Institute research; AI 2040 (2026), Get Involved in Verification; AI Security Forum, Requests for Discussion; Dean, Verification Plan (AI 2040); Milton & Reynolds (2026), Network Tapping for AI Verification: A Technical Assessment; Proofworks, Living Verification Documents; Reynolds & Foster (2026), Network Traffic Hashing.

Sub-questions

IDSub-questionWhat the sources report
P031.1Can an FPGA hash (SHA-2/SHA-3), timestamp and sign every packet at 400 Gb/s line rate, including 64 B minimum frames, and at what resource/power cost?partly answered
P031.2Has per-packet or per-batch hashing with >=128-bit digests been shown at 800G or 1.6T line rate?open
P031.3Can a SmartNIC or DPU (e.g. 32-64 core) hash all traffic at 400G with zero loss, given inline DPU crypto cannot expose the tag and RSS spreading may not be uniform on real traffic?open
P031.4At what sampling rate can a commodity 400G switch port-mirror without loss?a source suggests an answer (not checked)
P031.5Is a 64-bit digest sufficient for a high-bandwidth hash log?a source suggests an answer (not checked)
P031.6Can a COTS FPGA relay and log/hash all I/O of an inference server at line rate with under a stated added latency, and what changes make it trusted?open
P031.7Can a tap or logger capture/hash east-west (backend) traffic at 400-800 Gb/s per link or node with under a stated added latency and under a stated training slowdown?open

Sub-questions not yet stated precisely enough to check

IDSub-question
P031.8When will passive taps, SmartNICs and active tap appliances be available and installable at 800GbE and 1.6TbE?

Possibly relevant work found by automated search

Found by AI agents searching for this problem's sub-questions, mostly by keyword and citation in the OpenAlex scholarly database and on the web; a second AI pass reviewed each one, and we spot-checked some by hand. They seemed reasonably related, as far as one can tell from outside each paper's field. Some come from neighbouring fields, such as arms control and financial audit, and bear on the problem by analogy. Not reviewed against the status above; a paper listed here does not change that status. 928 of 1052 quotes checked word for word against the paper's text.

For P031.1: Can an FPGA hash (SHA-2/SHA-3), timestamp and sign every packet at 400 Gb/s line rate, including 64 B minimum frames, and at what resource/power cost?

For P031.3: Can a SmartNIC or DPU (e.g. 32-64 core) hash all traffic at 400G with zero loss, given inline DPU crypto cannot expose the tag and RSS spreading may not be uniform on real traffic?

Data