Side Channel Attacks

What You Will Learn

What a side channel attack is and how it differs from direct attacks
The main types of side channel attacks
How timing attacks work in practice
How to defend against side channel leakage

What Is It?

A side channel attack does not attack the algorithm or logic of a system directly. Instead, it exploits physical information leaked during execution — things like how long an operation takes, how much power it consumes, or what electromagnetic signals it emits.

Even if the cryptographic algorithm is perfect, a bad implementation can leak the secret key through these physical side channels.

Why It Matters

Side channel attacks have broken real-world hardware and software:

Smart card keys extracted via power analysis
AES keys leaked via cache timing on CPUs
RSA keys recovered from electromagnetic emissions

These attacks are used in hardware hacking, embedded security research, and advanced CTF challenges.

Key Concepts

Types of Side Channels

Type	What It Measures
Timing	How long an operation takes
Power Analysis	How much power a chip consumes
Electromagnetic (EM)	EM emissions from a device
Cache	CPU cache access patterns
Acoustic	Sound emitted during computation

Timing Attack

A timing attack measures how long code takes to run. If a comparison function returns early when it finds a mismatch, an attacker can determine correct bytes one at a time by measuring response time.

Example of vulnerable string comparison:

def check_password(input_pw, real_pw):
    for i in range(len(input_pw)):
        if input_pw[i] != real_pw[i]:
            return False  # returns early — leaks position of first mismatch
    return True

Example of safe constant-time comparison:

import hmac
def check_password(input_pw, real_pw):
    return hmac.compare_digest(input_pw, real_pw)  # always takes same time

Cache Timing Attack (Flush+Reload)

Modern CPUs cache memory accesses. An attacker sharing a CPU can measure whether a memory address is in cache (fast access) or not (slow access) to infer what data a victim process touched.

flush(target_address)       # evict from cache
wait for victim to run
t1 = time()
access(target_address)      # measure reload time
t2 = time()
if (t2 - t1) < threshold:
    # address was accessed by victim (it was in cache)

Simple Power Analysis (SPA)

During cryptographic operations, different operations (multiply, square) consume different amounts of power. An attacker with a power probe can read RSA private key bits directly from the power trace.

Differential Power Analysis (DPA)

DPA uses statistical analysis across many power traces to extract key material even with noise. It is far more powerful than SPA.

Defenses

Use constant-time algorithms for comparisons and crypto
Add random delays to obscure timing
Use power noise generators in hardware
Apply masking — XOR intermediate values with random data
Use hardware security modules (HSMs) with built-in countermeasures