Raspberry Pi has introduced that it’s loosening the foundations for its second RP2350 Hacking Problem, through which researchers are invited to aim to interrupt explicit points of the safety constructed into to firm’s second-generation in-house microcontroller — eradicating the necessity to maintain reminiscence deal with randomization enabled.
“On the finish of July 2025 — so virtually 6 months in the past — we launched the second RP2350 Hacking Problem, looking for sensible side-channel assaults on the power-hardened AES [Advanced Encryption Standard] implementation underpinning RP2350’s safe boot,” explains Raspberry Pi’s Thomas Roth. “To date, we don’t have a winner, so we determined to evolve the problem by eradicating one of many core defense-in-depth options: the randomization of reminiscence accesses. In the event you handle to show a profitable assault on our AES implementation with out the randomization, you win!”
Fancy a shot at $20,000? Raspberry Pi has now made it simpler to qualify for its RP2350 safety bounty. (📷: Raspberry Pi)
The Raspberry Pi RP2350 microcontroller, which options the bizarre potential to allow any two-core mixture from two Arm Cortex-M33 cores and two free and open Hazard3 RISC-V cores, launched again in August 2024 as a successor to the favored RP2040. Along with extra highly effective processor cores, extra reminiscence, and different enhancements, the chip additionally provided improved safety performance — one thing the corporate was so assured in it launched a $10,000 bug-bounty competitors, later elevated to $20,000. Naturally, flaws have been discovered and winners rewarded earlier than Raspberry PI introduced a second safety contest to see if its enhancements withstood scrutiny.
It is this which has now been prolonged, following a scarcity of profitable entrants, with a brand new rule designed to focus particularly on the built-in AES implementation. “Our AES implementation was designed to face up to side-channel assaults through the use of multi-way secret sharing (the place delicate values are break up into random elements that have to be XORed collectively) and by randomly permuting the order of operations and knowledge,” Roth explains. “We hope that even simply the multi-way shares are sufficient to guard us towards side-channel assaults.”
Consequently, the requirement that profitable assaults will solely depend as legitimate entries in the event that they maintain the chip’s separate reminiscence deal with randomization characteristic enabled has been eliminated — and now, even when disabled, entries can be thought of for the $20,000 prize. The deadline can be being prolonged, with entrants now having till April 30, 2026. “Getting shut however don’t have a profitable assault but? Write to us,” Roth says. “We care extra about defending our implementation than about having a full end-to-end assault. In the event you’ve recognized a leak, we wish to discuss to you!”
Extra data is on the market within the RP2040 Hacking Problem 2 GitHub repository.