NIST picked the Ascon family as its lightweight cryptography standard on February 7, 2023. It made that official on August 13, 2025, by publishing NIST SP 800-232. The standard defines four algorithms. Ascon-AEAD128 encrypts and authenticates data. Ascon-Hash256 is a hash function. Ascon-XOF128 and Ascon-CXOF128 produce output of any length you ask for. All four are built on one small permutation, so a sensor, RFID tag, implant or toll transponder can get encryption, integrity checks and hashing from very little code or silicon. AES and SHA-2 are still the right choice wherever they fit. Ascon is for devices where they don’t, or where they cost too much battery and chip area.
What changed since mid-2025?
In July 2025 Ascon was only a draft standard. NIST published the initial public draft in November 2024 and the final version on August 13, 2025, according to the NIST announcement. Use the SP 800-232 algorithm names (Ascon-AEAD128, not “Ascon-128”) when you specify parts or write requirements. The final standard does not include every variant the designers originally submitted.
Some figures that circulated in 2025 have no published source. These include “300–500 KB of memory for traditional crypto vs 50–100 KB for Ascon”, “30% less energy” and percentage cost savings for smart-city pilots. Ignore them. The real footprint of Ascon is far smaller than those numbers suggest (see below). Also, US-CERT no longer exists as a separate brand; its work now sits under CISA.
How did NIST choose Ascon?
The selection was an open competition that ran for about five years. NIST’s final-round status report (NIST IR 8454) records 57 submissions in 2019. Of these, 56 became first-round candidates, 32 went through to round two, and 10 were named finalists in March 2021:
- ASCON
- Elephant
- GIFT-COFB
- Grain-128AEAD
- ISAP
- PHOTON-Beetle
- Romulus
- SPARKLE
- TinyJAMBU
- Xoodyak
Ascon was designed in 2014 by Christoph Dobraunig, Maria Eichlseder, Florian Mendel and Martin Schläffer. Before NIST chose it, it had already been named the primary choice for lightweight authenticated encryption in the CAESAR competition (2014–2019), as noted on the Ascon project site. It had therefore been under public cryptanalysis for nearly a decade by the time it was standardized.
What exactly does the standard specify?
| Algorithm | Job | Key facts |
|---|---|---|
| Ascon-AEAD128 | Authenticated encryption with associated data | 128-bit key, 128-bit nonce, 128-bit tag. Without tag truncation it gives 128-bit security when nonces are never reused. An optional nonce-masking mode supports 256-bit keys. |
| Ascon-Hash256 | Hashing | 256-bit digest, for firmware integrity checks and key derivation inputs |
| Ascon-XOF128 | Extendable-output function | Output of any length |
| Ascon-CXOF128 | Customizable XOF | Adds a customization string so different uses of the same input give separate outputs |
The security claims depend on how you deploy it. Shorten the tag to λ bits and security drops to λ bits. Across many independent keys, security is reduced by log2 of the number of keys. Reusing a nonce under the same key is the most likely way to break a deployment, as with AES-GCM. Plan how nonces are generated before you write any firmware.
How small and fast is it in practice?
The designers publish reference implementations and benchmarks:
- Software, 64-bit server (AMD EPYC 7742): about 4.2 cycles/byte for Ascon-AEAD128 and 12.4 cycles/byte for the hash and XOF functions on long messages.
- Software, older 32-bit ARM (ARM1176JZF-S): about 42.9 cycles/byte for AEAD and 92.2 cycles/byte for hashing.
- RISC-V co-processor: about 2 cycles/byte at roughly 4.7 kGE (thousand gate equivalents, a standard measure of chip area). With fault-attack and power-analysis protection, about 4 cycles/byte.
Footprint is the real advantage. Encryption, hashing and MAC all come from one permutation, so you need one code path or one hardware block instead of a separate AES core plus a SHA-2 core. On a microcontroller with no AES accelerator, that saving can decide whether security fits in the flash and battery budget at all. On a chip that already has an AES engine, AES-GCM or AES-CCM is often still the cheaper choice.
Is Ascon safe against side-channel and quantum attacks?
For side channels, NIST says Ascon “is designed to support side-channel-resistant implementations more easily than many traditional algorithms.” NIST also warns that no algorithm is immune. Masking (splitting secret values into random shares so power traces leak less) and fault countermeasures still cost extra area and cycles, as the protected RISC-V figures above show. Devices an attacker can physically hold, such as meters, locks and wearables, need a protected implementation.
For quantum attacks, Ascon is symmetric cryptography. Ascon-AEAD128 uses 128-bit keys, which NIST’s lightweight project treats as adequate. Ascon does nothing for key exchange or signatures, the parts that quantum computers actually threaten. Pair it with post-quantum key establishment such as ML-KEM (FIPS 203) where the device can afford it, or with pre-provisioned symmetric keys where it can’t.
Who should adopt it, and how?
Ascon is the right default for new designs that fall into any of these groups:
- RFID and NFC tags
- Implantable and wearable medical devices
- Battery sensors that must last years
- Toll and asset-tracking transponders
- Any 8-bit, 16-bit or low-end 32-bit microcontroller without AES hardware
Projects to adopt it are under way outside NIST too. An IEEE 802.1 project has proposed an Ascon cipher suite for MACsec, the IEEE standard for encrypting Ethernet links.
A sensible rollout order:
- Check each device against NIST’s IoT Device Cybersecurity Capability Core Baseline (NIST IR 8259A). Encryption is one capability out of several. Secure updates and device identity matter as much.
- Use the designers’ reference code or a vendor library that implements SP 800-232. Avoid libraries written against the pre-standard Ascon v1.2 specification, because the final standard differs from it. Check against NIST test vectors.
- Decide how nonces will be generated (counter, or random with enough bits) and how long tags must be before you write production firmware.
- Budget for a masked implementation on any device that people can physically get their hands on.
We produce explainer and launch videos for hardware clients, and the same mistake shows up in nearly every script review: Ascon gets called “military-grade” or “quantum-proof”. It is neither. It is NIST-standardized authenticated encryption and hashing for constrained devices, and engineering buyers trust the product more when the marketing says exactly that.
Next step: list every product that ships without hardware AES or runs near the limit of its battery budget. Prototype Ascon-AEAD128 on that hardware with the reference code, measure cycles and flash use on your own MCU, and then decide between Ascon and AES-CCM using your own numbers.





