The Signal
JPMorgan Researchers Combine QKD Authentication and Key Expansion
JPMorganChase researchers tested a quantum authenticated key expansion protocol that combines authentication and key generation, with each of three runs producing between 177,000 and 213,000 key bits. The demonstration used QKD hardware wit
Quantum.world Team · Written with AI · fact-checked ·
Three test runs of a JPMorganChase quantum key protocol each produced roughly 200,000 secret-key bits while combining client authentication with key generation. Published October 28 in Quantum, the protocol also lets the parties reuse authentication keys after a successful session rather than refresh them after every run.
The work comes from eight researchers in JPMorganChase’s Global Technology Applied Research group: Wen Yu Kon, Jefferson Chu, Kevin Han Yong Loh, Obada Alia, Omar Amer, Marco Pistoia, Kaushik Chakraborty and Charles Lim. They call the method quantum authenticated key expansion, or QAKE.
“We experimentally implemented the protocol with appropriate post-selection.”
— The paper’s authors write. Wen Yu Kon et al., authors of the paper
Two authentication checks inside the QKD session
Conventional QKD generates shared secret bits, but the communicating parties still need to authenticate one another and their classical messages. The JPMorgan team’s design brings those tasks into the key-distribution session: it authenticates Alice and the exchanged transcript, then checks Bob, using two authentication steps near the end of the protocol.
The design also combines error verification with authentication, reducing the amount of shared randomness consumed during post-processing. A pre-shared privacy-amplification seed can be reused, too. The authors say the approach can reduce the overhead of refreshing authentication secrets across a service provider’s servers—but key recycling applies when the protocol succeeds; the protocol handles failed sessions by moving to new secret labels rather than continuing to use the same keys.
About 200,000 bits from a post-selected test
The researchers tested the protocol using commercial ID Quantique Clavis XGR QKD systems. An optical attenuator imposed 10 decibels of channel loss, representing the loss of a 45-kilometer fiber link; the setup did not send quantum signals over that length of actual fiber. The team fixed the test-round probability at 25% and collected data in three runs.
After post-selection, each run used roughly 20 billion signals. The measured detection probability was about 1.03 × 10-4 per signal, and the quantum bit error rate was about 2%. The resulting key lengths were 177,000 bits in the first run and 213,000 bits in each of the next two. The paper calculates those lengths using a secrecy parameter of 10-15.
Offline processing limits what the demonstration shows
The experiment demonstrates a protocol using real QKD hardware, but it does not show a complete, real-time system. The authors post-selected the recorded raw data to balance detection between the two measurement bases, then ran the remaining protocol steps—including error correction, authentication and privacy amplification—offline in software. The paper also describes a Toeplitz-hashing implementation on a Virtex UltraScale+ FPGA, clocked at 332 MHz.
The security analysis treats repeated sessions and separates conditions for authentication from those for data privacy. It also relies on the assumption that shared secrets do not leak through side channels between sessions. The test therefore establishes a post-processed implementation under a specified loss and error model, not unattended operation across a deployed network. Questions for practical deployment include how to integrate the processing in real time and how the protocol performs on a live fiber connection.