Comparative Timing Characterization of Early-Exit and Fixed-Work Secret Comparisons on ESP32 and ESP8266

Abstract An early-exit secret comparison can expose matching-prefix length through execution timing. We quantify this behavior for 16-byte early-exit comparisons and a fixed-work XOR/OR alternative on ESP32 and ESP8266 microcontrollers. Internal CPU-cycle measurements were collected for prefix lengths 0–16 under two interrupt/barrier configurations, yielding 40,800 observations. Across prefix lengths 0–15, early-exit median timing rose by 40 cycles per matching byte on ESP32 and 25 on ESP8266. In the recorded sessions, 63 of 68 early-exit groups contained 300 identical readings, and all 17 group medians were distinct within each configuration. Fixed-work median timing did not vary with prefix length, although occasional elevated observations remained. These measurements characterize implementation-specific on-device timing leakage and the cost of the fixed-work routine; they neither establish binary-level constant-time behavior nor demonstrate a remotely observable attack.