IN THIS ARTICLE YOU WILL LEARN:
- Study Overview: NFTEvening and Storible collaborated on a study to test whether AI, specifically long-term and short-term memory (LSTM) networks, can recover missing cryptocurrency seed phrases by analyzing 85.7 million of combinations using the BIP39 word list.
- AI Capabilities: The neural network, trained for 30 days, achieved a prediction rate of 994,051 guesses per second. It could recover one missing word in 0.02 seconds, two words in 29 seconds, and three words in 2.28 hours. Retrieving up to four words took 178 hours, with retrieval times increasing exponentially for more missing words.
- LSTM functionality: LSTMs are a type of RNN that excels at processing sequential data, using memory cells and gates to predict and complete patterns in text-like data, making them suitable for seed sentence prediction .
- Security of cryptographic seed phrases: Despite advances in AI, the massive number of possible seed phrase combinations makes it virtually impossible for AI to retrieve enough words to hack cryptocurrency wallets. For example, deciphering eight missing words would take 174 times the age of the universe.
- Conclusion and disclaimer: The study shows that AI does not pose a current threat to the security of cryptocurrency seed phrases if the phrase remains secret. However, these results are based on current AI capabilities and computing limitations.
NFT-focused news site NFTEvening and NFT market data and analytics-driven platform Storible collaborated to conduct a study to determine whether powerful AI, specifically a type called long-term memory and Short Term (LSTM), can decipher cryptocurrency seed phrases.
For this research, they analysis 85,714,285 seed phrase combinations using 2,048 words from the BIP39 word list to determine the time required for AI to recover missing cryptographic seed phrases and the feasibility of LSTM networks to recover lost or incomplete cryptocurrency phrases .
For this, a neural network was trained on a cloud server for 30 days to predict seed sentences and estimate recovery time. The model used the LSTM architecture, achieving an average prediction rate of 994,051 guesses per second. Retrieval time was analyzed using scenarios ranging from one to twelve missing words.
The results showed that the AI could recover one missing word in 0.02 seconds, two words in 29 seconds, three words in 2.28 hours, and up to four words in a maximum retrieval time of 178. Additionally, AI can discover the correct sequence of 12-word seed phrases in just 8 minutes, while the time required to recover 8 missing words in a seed phrase is 174 times longer than the current age of the universe.
What is LSTM?
FYI, a seed phrase is like a secret password for your cryptocurrency. This is a list of 12 to 24 words that gives you access to your digital wallet.
LSTM networks are recurrent neural networks (RNNs) that excel at memorizing information over extended sequences. They integrate memory cells and gates to control the flow of information, allowing them to efficiently learn and predict patterns in sequential data such as text and time series.
How LSTMs Work in Seed Sentence Retrieval
LSTMs can predict the likely next word in a sequence or fill in missing words by observing patterns in known seed phrases. Memory cells store crucial information, with three gates – forget, input and output – regulating the flow of information. The forget gate discards irrelevant information, the input gate determines new information, and the output gate controls the prediction of the next word.
The researchers tried to use this AI to guess missing words in seed sentences. They tested it on millions of possible combinations. They discovered that even with this powerful AI, it would take an incredibly long time, longer than the age of the universe, to guess enough words to steal someone’s cryptocurrency.
Research indicates that AI does not pose a threat to your cryptocurrency if you keep your seed phrase secret. Although AI could eventually decipher seed phrases, it is currently highly secure. Although LSTMs provide powerful sequence learning capabilities, the large number of possible combinations and the inherent complexity of seed phrase structures make successful retrieval highly unlikely.
Disclaimer: This analysis provides an estimate based on current AI capabilities and computing power.
RELATED TOPICS
- AI-based model to predict extreme wildfire danger
- Are we about to say goodbye to passwords?
- Google’s Gemini AI chatbot keeps telling users to die
- Guess the password “just by the way we tilt our phone”
- unCAPTCHA Algorithm Crack Google’s AI system reCAPTCHA
