When the implementation betrays the secret

The idea: the physical leak

An encryption algorithm can be mathematically sound — RSA, AES, everything we have seen — and yet be broken without anyone touching the mathematics. The weak link is not the theory: it is the physical machine that runs the code.

The main channel and the side channels

When a processor encrypts a message, it receives an input (the plaintext, the key) and produces an output (the ciphertext). This is the main channel, the one the attacker is supposed to observe.

But physical execution also produces secondary, unintended effects: it takes time, consumes current, heats up, emits waves, makes noise. These effects form as many side channels (in English, side-channels).

                 +-----------------------------+
   texte clair ->|      BOÎTE CRYPTO           |-> texte chiffré
        clé  --->|   (mathématiquement sûre)   |   (canal principal)
                 +-----------------------------+
                    |      |       |      |
                    v      v       v      v
                 temps  courant   EM    bruit / chaleur
                 \____________ FUITES ____________/
                        (canaux auxiliaires)

The principle of the attack is as follows: these leaks depend on the data being processed, and therefore on the secret key. By measuring them precisely, the attacker reconstructs the secret without ever solving the mathematical problem.

The attacker does not attack the maths

This is the shift in perspective to remember. Classical cryptanalysis attacks the algorithm: factoring n, breaking a permutation. A side-channel attack ignores all of that and targets the implementation.

The factorization of RSA remains out of reach? Never mind: we measure the computation time of the decryption.
AES is proven resistant to cryptanalysis? Never mind: we listen to the power consumption of the smart card.

A catalogue of leaks

The channels exploited are varied:

Channel	What is measured	Typical target
Timing	computation duration	servers, TLS
Power	electrical current	smart cards
Electromagnetic	emitted waves	connected devices
Acoustic	component noise	processors
CPU cache	memory access time	shared machines

Why it is a serious threat

These attacks are passive: very often, the attacker simply observes, without modifying the system, and therefore without leaving a trace. They affect everyday objects — bank cards, biometric passports, car keys, server cores — everywhere a secret is handled by accessible hardware.

The underlying lesson: security does not play out only in the mathematical proof, but in the way the computation is carried out. A good algorithm badly implemented is a bad system.

In summary