Skip to main content
Definition Website Computer and Technology

Side-Channel Attack

Subject: Computer and Technology noun /ˈsaɪd ˌtʃæn.əl əˈtæk/ 1 definition

  1. Definition (noun)

    Entry #344 · Updated

    A Side-Channel Attack is a security exploit that recovers sensitive information—such as cryptographic keys, passwords, or protected memory contents—by observing unintended physical or operational artifacts emitted by a computing system rather than breaking mathematical algorithms or exploiting software logic flaws. Whenever hardware or software branches on, indexes with, or computes over secret values, those data-dependent operations subtly modulate measurable properties like execution timing, electrical power draw, electromagnetic emissions, or shared CPU cache states. By capturing repeated observations and applying statistical methods such as differential analysis or machine learning, an adversary separates secret-correlated signals from system noise to reconstruct hidden state incrementally. Because side channels arise from how a design is physically and microarchitecturally implemented rather than from its abstract specification, even provably secure ciphers and formally verified protocols require explicit countermeasures—such as constant-time code, cryptographic blinding, power masking, or hardware partitioning—to close leakage paths.

    Common Types

    • Timing Attacks: Measure tiny variations in how long an operation takes—such as early-exit string comparisons, non-constant-time modular exponentiation, or branch mispredictions—to deduce secret inputs locally or across a network.

    • Power Analysis Attacks: Monitor fluctuations in a device's electrical current or voltage draw during computation, using Simple Power Analysis (SPA) or Differential Power Analysis (DPA) to extract keys from smart cards, microcontrollers, and hardware security modules.

    • Cache and Microarchitectural Attacks: Exploit shared CPU hardware structures—such as L1/L3 caches, branch target buffers, or speculative execution paths—using probing techniques like Prime+Probe or Flush+Reload to infer another process's memory accesses across isolation boundaries.

    • Electromagnetic and Acoustic Attacks: Capture unintentional radio-frequency emissions from processor buses or faint high-frequency vibrations from voltage regulators and keyboards using nearby probes or microphones to recover cryptographic keys or typed input without physical tampering.

    Example

    On a Linux host where two isolated processes map the same read-only libcrypto.so binary, an unprivileged attacker runs a Flush+Reload loop using the x86 clflush and rdtsc instructions to measure L3 cache-hit latencies while a victim service generates ECDSA signatures with a non-constant-time scalar multiplication routine. By correlating nanosecond timing differences across thousands of signing operations, the attacker infers secret bit patterns in the ephemeral nonces and derives the private key without ever reading the victim's protected memory.