A side-channel attack learns secrets from observable properties of an implementation rather than from a weakness in the cryptographic algorithm.
Meltdown and Spectre show how shared CPU microarchitecture creates a practical side channel.
Leakage sources include:
- Execution time
- Branch behaviour
- Cache accesses
- Power consumption
- Electromagnetic emissions
- Fault responses
Timing Attacks
A timing attack measures how long an operation takes. A difference correlated with a secret bit, byte, or comparison result leaks information over repeated measurements.
An attacker may average many noisy observations. A small but repeatable timing difference can then reveal the secret-dependent path.
Square-and-Multiply
Naive modular exponentiation processes each bit of a private exponent .
- A bit squares the accumulator.
- A bit squares the accumulator, then multiplies.
The extra multiplication changes the execution time, power trace, or electromagnetic trace. An observer can distinguish the bits of .
Mitigation
Cryptographic code must not expose secret-dependent observable behaviour.
- Use constant-time algorithms with a fixed instruction pattern and memory-access pattern.
- Replace secret-dependent branches with constant-time selection.
- Use exponent blinding or message blinding for private-key operations.
- Use vetted cryptographic libraries and hardware implementations.
- Test implementations for timing, cache, power, and fault leakage.
Constant time reduces a major source of leakage. It does not eliminate power analysis, electromagnetic analysis, or fault injection by itself.
