Side-channel attack papers borrow benchmarks instead of retesting

Side-channel attack researchers compare new exploits against numbers borrowed from earlier papers, rather than retesting them, per a paper on side-channel benchmarking.

These attacks hinge on exact hardware and cache setup, and a small difference between machines can swing results enough to make comparisons meaningless. The proxies show up at top-tier security venues, the ones setting the bar for what counts as progress.

Treating a borrowed number as an apples-to-apples comparison risks crediting a new attack with an improvement that's really just a different test machine.

The proxies include covert-channel bandwidth and key recovery against naive AES and RSA implementations, standing in for a retest on the attack's own hardware.

The paper is a methodology critique, not a new attack, which is rarer in a field that tends to reward novel exploits over process work. It doesn't name which published results the comparisons undermine or propose a replacement benchmark, so the diagnosis arrives without a fix.