By adapting a clever technique from Filippo Valsorda, we can exploit the structure of ML-DSA to improve our sampling efficiency, getting more precise performance estimates in 10 hardware simulations than we could before in 100.
Take a look at how we made our benchmarks both faster and more precise in our latest deep-dive here:
https://t.co/t1S4AJG5aJ
Benchmarks are hard. Cryptographic benchmarks are harder. 📊⏱️
New post-quantum algorithms like ML-DSA can be incredibly difficult to benchmark, because their runtimes can vary depending on random values, meaning lots of time-consuming tests are needed to estimate performance.
Performance, versatility, and maintainability can be at odds unless you design carefully.
Take a peek here at our new blog post for the tricks, tips, and tools we use in maintaining the cryptolib cryptographic library: https://t.co/oFhx0Btprn
[3/3]
Fast. Flexible. Future-proof. Pick three. 📷📷
After sharing how we achieved ~3x performance on core RSA operations, we’re now giving a tour of how our cryptographic library can support a root of trust in everything from IoT to a BMC.
[1/3]
Expect formal verification, static analysis, modular design–and of course, more performance tricks.
While implementing cryptography can be an easy engineering exercise, implementing production-quality cryptography is not.
[2/3]
Last week, we discussed the mathematical details of why CRT modular exponentiation works.
Check out Part 2 below for all of the implementation details, from dealing with register pressure to FIPS certifiability!
https://t.co/jik2XDvqzi
[2/2]
Lighting-fast crypto in pure assembly? 🏎️💨
Building on our post from last week, we’re sharing the technical details behind our 3.43x RSA speedup from implementing Chinese Remainder Theorem (CRT) modular exponentiation in our embedded cryptography library.
[1/2]
Modular exponentiation is the core of signing & decryption, and we've sped it up significantly. Check out Part 1 of our series for the mathematical deep dive!
https://t.co/bArVZztdry
#ZeroRISC#RSA#Security#Crypto [2/2]
3.43x faster RSA! 🚀
ZeroRISC just achieved a huge performance boost for core RSA operations by implementing Chinese Remainder Theorem (CRT) modular exponentiation for our embedded cryptography library. [1/2]
Learn how we’re advancing certification-aligned open silicon in @GlobalPlatform’s Trusted Open Silicon Task Force (#TOS): https://t.co/666XCstmQw [2/2]
ZeroRISC and Rivos implemented FIPS 140-3 compliant OTP zeroization, showing why building for certification early in silicon design is critical, because do-overs are pricey in hardware. [1/2]
This summer, our MIT intern Yeabsira Hawaz dove into open source silicon integration⚡He designed a TL-UL to AXI bridge + AXI AdapterReg, enabling seamless AXI compatibility. Testing, debugging, and iterating paved the way for smoother adoption & real-world hardware innovation🚀
This summer, our MIT intern Beshr Bouli improved flexibility & memory efficiency in ZeroRISC’s embedded Rust-based secure OS. PIC + a new memory allocator enables dynamic app loading with less downtime & fragmentation. Learn more in Beshr’s blog: https://t.co/9OMKT3QWyR
Congratulations to our board member Zakir Durumeric, co-founder of Censys, on winning the Industry Leadership award in the 2025 #CyberScoop50! His groundbreaking work continues to shape the future of security. 👏🔒