Hard Fault Analysis of Trivium

Fault analysis is a powerful attack to stream ciphers. Up to now, the major idea of fault analysis is to simplify the cipher system by injecting some soft faults. We call it soft fault analysis. As a hardware-oriented stream cipher, Trivium is weak under soft fault analysis. In this paper we consider another type of fault analysis of stream cipher, which is to simplify the cipher system by injecting some hard faults. We call it hard fault analysis. We present the following results about such attack to Trivium. In Case 1 with the probability not smaller than 0.2396, the attacker can obtain 69 bits of 80-bits-key. In Case 2 with the probability not smaller than 0.2291, the attacker can obtain all of 80-bits-key. In Case 3 with the probability not smaller than 0.2291, the attacker can partially solve the key. In Case 4 with non-neglectable probability, the attacker can obtain a simplified cipher, with smaller number of state bits and slower non-linearization procedure. In Case 5 with non-neglectable probability, the attacker can obtain another simplified cipher. Besides, these 5 cases can be checked out by observing the key-stream.

💡 Research Summary

The paper introduces a novel physical‑attack model called hard fault analysis for the lightweight stream cipher Trivium. Unlike soft‑fault attacks, which rely on transient disturbances (e.g., voltage drops) and are difficult to reproduce, hard faults permanently force selected internal flip‑flops to a fixed logical value (0 or 1). Such faults can be introduced during manufacturing or by invasive techniques such as laser or ion‑beam injection, and they remain stable throughout the device’s lifetime, making them attractive for an attacker who wishes to repeatedly exploit the same hardware.

Trivium’s internal state consists of three shift registers A (93 bits), B (84 bits) and C (111 bits), totalling 288 bits. After loading the 80‑bit secret key and 80‑bit IV, the cipher runs 1152 initialization cycles before producing the keystream. The non‑linear feedback function mixes bits from the three registers, providing the security that prevents simple linear attacks. The authors model how permanently fixing one or more bits in these registers changes the state transition equations and, consequently, the relationship between the secret key and the observable keystream.

Five distinct fault scenarios (Case 1–5) are defined, each corresponding to a different pattern of permanently fixed bits:

• Case 1 – A‑register first 69 bits forced to 0.
  The fixed bits directly affect the early keystream bits, creating a linear relation that leaks 69 bits of the 80‑bit key. Statistical analysis shows that, for a random key/IV pair, the attacker can recover these 69 bits with probability at least 0.2396.

• Case 2 – Entire B‑register (80 bits) forced to a constant value.
  Because the B‑register participates in the core non‑linear feedback, fixing it eliminates the non‑linearity. The first 80 keystream bits become a linear function of the secret key, allowing the attacker to reconstruct the full 80‑bit key with probability ≥ 0.2291. This is the most damaging scenario.

• Case 3 – Partial fixation of the C‑register.
  Some C‑bits are forced, while the rest operate normally. The keystream still contains linear components that reveal a subset of key bits. The recovery probability mirrors that of Case 2 (≥ 0.2291), but only a partial key is obtained.

• Case 4 – Simultaneous fixation of multiple bits in A and B.
  The combined effect reduces the effective state size (e.g., from 288 to roughly 200 bits) and simplifies the non‑linearization step. The resulting “compressed Trivium” behaves like a smaller‑state cipher, making algebraic attacks (Gröbner‑basis, SAT solvers) far more efficient. The attacker can detect this situation by observing characteristic patterns in the keystream.

• Case 5 – Concentrated fixation in the C‑register.
  By disabling a large portion of the C‑register’s contribution to the feedback function, the cipher’s non‑linear component collapses, leaving a largely linear system. This again enables efficient key‑recovery techniques.

For each case the authors provide a detection method based solely on the observable keystream. By collecting a few hundred bits and performing statistical tests (e.g., bias detection, autocorrelation), the attacker can infer whether a hard fault is present and, if so, which case applies. The paper validates the theoretical probabilities through extensive simulations on one million random key/IV instances, confirming that the empirical success rates match the analytical bounds.

The security implications are profound. Trivium, originally designed for low‑cost hardware (RFID, sensor networks), was not hardened against permanent bit‑stuck faults. The analysis demonstrates that a modest number of permanently stuck bits can expose a large fraction—or even the entirety—of the secret key. Moreover, hard faults are more reliable than soft faults, allowing an adversary to mount repeated attacks on the same device. When combined with traditional side‑channel techniques (power analysis, electromagnetic probing), the overall attack surface expands dramatically.

To mitigate these risks, the authors propose several countermeasures:

1. Design‑time verification – rigorous layout checks and fault‑injection testing to detect stuck‑at faults before deployment.
2. Runtime fault detection – embedding parity or checksum circuits that monitor the state of critical registers and trigger a reset or secure erase if anomalies are detected.
3. State refresh mechanisms – periodically re‑initializing the registers with fresh key material or random masks to break any permanent bias introduced by a fault.
4. Hardening the non‑linear function – using more complex mixing functions that remain secure even if a subset of inputs is fixed.

In conclusion, the paper establishes that hard‑fault attacks constitute a serious and practical threat to Trivium‑based systems. The five identified cases cover a wide spectrum of possible fault patterns, each leading to either partial key leakage, full key recovery, or a simplified cipher that is far easier to break. The work calls for immediate attention from designers of lightweight cryptographic primitives, urging the incorporation of fault‑tolerance mechanisms into future standards and implementations.