SNOW-SCA: ML-assisted Side-Channel Attack on SNOW-V

Harshit Saurabh,Anupam Golder,Samarth Shivakumar Titti,Suparna Kundu,Chaoyun Li,Angshuman Karmakar,Debayan Das

DOI: https://doi.org/10.48550/arXiv.2403.08267

2024-03-13

Abstract:This paper presents SNOW-SCA, the first power side-channel analysis (SCA) attack of a 5G mobile communication security standard candidate, SNOW-V, running on a 32-bit ARM Cortex-M4 microcontroller. First, we perform a generic known-key correlation (KKC) analysis to identify the leakage points. Next, a correlation power analysis (CPA) attack is performed, which reduces the attack complexity to two key guesses for each key byte. The correct secret key is then uniquely identified utilizing linear discriminant analysis (LDA). The profiled SCA attack with LDA achieves 100% accuracy after training with $<200$ traces, which means the attack succeeds with just a single trace. Overall, using the \textit{combined CPA and LDA attack} model, the correct secret key byte is recovered with <50 traces collected using the ChipWhisperer platform. The entire 256-bit secret key of SNOW-V can be recovered incrementally using the proposed SCA attack. Finally, we suggest low-overhead countermeasures that can be used to prevent these SCA attacks.

Cryptography and Security,Machine Learning,Networking and Internet Architecture

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is the power side - channel attack (SCA) problem of SNOW - V, a candidate algorithm for 5G mobile communication security standards. Specifically, for the first time, researchers demonstrated a method of power side - channel analysis (SCA) attack on the SNOW - V stream cipher running on a 32 - bit ARM Cortex - M4 microcontroller. The following are the specific problems that this paper attempts to solve:

1. **Identify leakage points**: Determine the leakage points in the SNOW - V algorithm through general known - key correlation (KKC) analysis.

2. **Reduce attack complexity**: Use correlation power analysis (CPA) to reduce the complexity of each attack to two key guesses per key byte.

3. **Accurately recover keys**: Combine linear discriminant analysis (LDA) to uniquely identify the correct secret key byte from the ghost peaks obtained in the CPA attack.

4. **Efficient attack model**: The proposed combined CPA and LDA attack model can achieve 100% accuracy after training with less than 200 traces and can successfully attack with only one trace.

5. **Fully recover keys**: Recover the entire 256 - bit secret key step by step through incremental attacks.

6. **Propose countermeasures**: Suggest and evaluate several low - cost countermeasures to prevent these side - channel attacks, such as Boolean masking, constant - time implementation of branch operations, and obfuscation methods.

### Key contributions of the paper

- **First attack on SNOW - V**: This is the first reported power side - channel attack on the SNOW - V algorithm.
- **Successful verification of key recovery**: Using the proposed CPA and LDA attack model based on the known initialization vector (IV), the key recovery was successfully verified on a 32 - bit ARM microcontroller.
- **Efficient attack means**: The LDA model can uniquely identify the correct key byte with only a single trace.
- **Effective defense measures**: The proposed software countermeasures, such as Boolean masking, show the highest SCA resistance. Even with 50,000 traces, the correct key cannot be uniquely recovered, showing an improvement of more than 1,000 times in minimum trace disclosure (MTD).

### Method overview

1. **Architecture analysis**: Determine that the most vulnerable part in SNOW - V is the LFSR because the key and IV are loaded at this stage.
2. **TVLA test**: Use Welch's t - test hypothesis to check the time points where the |t| value is greater than 4.5, indicating the existence of data - dependent side - channel leakage.
3. **KKC analysis**: Verify and confirm the attack model and identify the leakage patterns related to the selected attack points.
4. **CPA attack**: Recover the correct key byte by byte, but ghost peaks appear.
5. **LDA model**: Used to uniquely identify the correct key byte, especially to deal with the LSB loss problem caused by the `mul_xinv()` function.
6. **Incremental attack**: Recover all key bytes step by step.

Through these steps, researchers not only demonstrated how to effectively attack SNOW - V but also proposed countermeasures to enhance its security, providing important security guarantees for future 5G systems.