A fault attack deliberately makes a device compute the wrong result, then uses that error to reveal or compromise secret information. Faults may come from heat, lasers, voltage changes, or repeated memory access. section 13.4.3 section 13.2
A fault can matter because:
- A bad cryptographic result can leak a secret key, including with some optimized RSA signature implementations. section 13.4.3
- Faults can target randomness: algorithms requiring fresh randomness each time may be especially exposed. section 13.4.3
- Defenses include repeating and comparing operations, verifying results before release, and verifying signatures with the public key. section 13.4.3 section 13.5
A practical threat
An attacker with physical access might aim a laser at a chip or alter CPU voltage to force an erroneous computation. No defense is foolproof, so systems combine mitigations and limit impact with unique keys per device or operation. section 13.4.3 section 13.5
