5G CATS certifies AI accelerator location with an in-package 5G radio
Original title: 5G CATS: 5G-Based Cryptographic Authorization Against Trafficking of Semiconductor Devices
Unloaded-cell simulations cut 99.9th-percentile RTT spread to 2.60 ms and reject every 300 km fiber-relay round.
5G CATS is a location-certification design for export controls on AI accelerators. A co-packaged 5G radio, cryptographic engine, and hardware interlock terminate verification inside the GPU package, outside the host stack, and use URLLC base stations. In ns-3 and 5G-LENA simulations of an unloaded cell, the profile rejects a 300 km fiber relay and stays safe farther from a 3,000 km attacker than PowerCATS or default 5G.
Unconfirmed: Results are from unloaded-cell simulations, not a live carrier deployment.
- The 99.9th-percentile round-trip spread is 2.60 ms, over ten times narrower than default 5G.
- A tracer rejects every relay round from 300 km, versus 2,000 km for default 5G.
- Against a 3,000 km attacker, the safe tracer distance is up to 1,410 km.
- PowerCATS's published model stays safe at most 619 km in that comparison.
- Short-lived execution leases replace permanent key erasure.
From the original article
1 min read at arxiv.orgEnforcing export controls on AI accelerators requires verifying where a chip operates after deployment. Round-trip time measurements can bound a chip's distance from a trusted landmark, but any jitter in honest replies forces the verifier to loosen its timeout, giving an attacker slack to relay traffic over fiber from a restricted region.
By Sreevikram Rajaraman · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.