North Korea has been confirmed to be researching methods to predict Bitcoin prices using artificial intelligence (AI) technology. While there have been many observations that North Korea holds and trades virtual assets (digital assets) like Bitcoin, this is the first time specific details have been confirmed through internal North Korean documents.
According to sources familiar with North Korea, researchers from Kim Il Sung University submitted a paper titled “A Method for Predicting Bitcoin Prices Using Deep Learning Models” to the university's journal Economics, Issue 3, 2025.
According to the paper obtained by NK Economy, the North Korean researchers argued, “One of the crucial issues in proactively entering the international financial market and aggressively conducting financial activities is establishing modern computational tools and scientific methodologies to accurately grasp the price fluctuations of various financial products, such as Bitcoin, which are widely traded in the current international financial market.”
The North Korean researchers pointed out, “Bitcoin, due to its characteristic of having no material collateral, exhibits significantly larger price fluctuations compared to other commodities traded in the market.” They added, “Therefore, efforts to widely utilize virtual currencies like Bitcoin in external financial transactions are becoming increasingly active on a global scale, aiming to enhance the effectiveness of virtual currency use through accurate price prediction.”
They introduced that traditional statistical methods like ARIMA models and deep learning models such as artificial neural networks have been proposed for predicting the prices of virtual currencies, including Bitcoin, in the past.
Examples include ‘Bitcoin Price Forecasting Using Machine Learning, McNally, 2016’, ‘Real-Time Bitcoin Price Prediction Using Machine Learning Techniques and Social Sentiment Analysis, Raju & Tarif, 2020’, ‘Application of Recurrent Neural Networks for Multivariate Time Series Forecasting of Volatile Financial Data, Torres & Qiu, 2018’, ‘Application of Deep Learning Algorithms for Bitcoin Price Prediction, Maharetal, 2021’, ‘Comparison of Machine Learning Models for Bitcoin Price Prediction, Phaladisailoed & Numnonda, 2018’, ‘Deep Learning-Based Bitcoin Price Prediction, Jiang, 2020’, ‘Bitcoin Price Time Series Prediction Using High-Dimensional Features: Application of Machine Learning Models, 2020’, among others. This indicates that North Korean researchers are continuously reviewing overseas research materials related to Bitcoin price prediction.
The North Korean research team stated in their paper that they propose a new hybrid prediction model combining Long Short-Term Memory (LSTM) and Particle Swarm Optimization (PSO) to enhance the accuracy of Bitcoin price prediction.
Long Short-Term Memory (LSTM) is a deep learning technique within the Recurrent Neural Network (RNN) methodology. Particle Swarm Optimization (PSO) is a swarm-based optimization algorithm used in the field of artificial intelligence. The North Korean researchers combined these two technologies to predict Bitcoin prices.
The North Korean researchers also introduced Bitcoin. They stated, “In 2008, an anonymous programmer (or group) named Satoshi Nakamoto published a paper related to cryptography on the mailing list of the website www.metzdowd.com이라는. This paper introduced a digital currency called Bitcoin.” “In 2009, Nakamoto released the Bitcoin source code in binary format on the website www.sourceforge.net에. In June 2009, he launched a peer-to-peer network and began developing Bitcoin.”
It also noted, “Bitcoin is the world's first completely decentralized virtual currency, or cryptocurrency, created and circulated on a computer network based on specific cryptographic programs, free from the centralized control of banks and other financial institutions.” It pointed out, “Bitcoin is the most representative virtual currency and cryptocurrency, maintaining the highest global usage share from its early development to the present.”
It further explained that Bitcoin is recognized as the first virtual currency to realistically escape centralized control and achieve widespread use. Unlike traditional fiat currencies issued and circulated based on gold or commodity collateral, Bitcoin is a new form of cryptocurrency issued through the process of solving cryptographic puzzles on the Bitcoin network.
The most crucial distinction between Bitcoin and traditional fiat currencies lies in its issuance method: mining, a cryptographic calculation process. Each Bitcoin issued through mining exists as an encrypted file that the miner or purchaser can store in their own wallet or deposit into a Bitcoin exchange.
The North Korean research team argued that the defining characteristic of Bitcoin price fluctuations is its exceptionally large price swings and chaotic ups and downs compared to other financial products traded in past international financial markets. They compared Bitcoin prices from September 30, 2013, to April 6, 2020, with the S&P 500 stock index, the dollar and euro exchange rates, gold prices, and crude oil prices.
The North Korean research team explained that they utilized deep learning technology for this Bitcoin price prediction.
The North Korean research team pointed out that although long short-term memory neural networks demonstrate excellent performance in processing various time series or sequence data, they have a flaw: the parameter settings for neural network learning are typically set based on experience, leading to strong subjectivity and reduced prediction accuracy.
Therefore, to solve this problem, they combined it with the particle swarm optimization method. The North Korean research team explained that particle swarm optimization (PSO), developed in 1995, is an optimization technique simulating the process of birds searching for food.
The North Korean research team compared and analyzed their developed PSO-LSTM model against standard LSTM and genetic algorithm-long short-term memory neural networks (GA-LSTM). They used the Coindesk Bitcoin Price Index as the foundational data for price prediction. The North Korean research team reportedly used daily closing Bitcoin prices from September 30, 2013, to May 23, 2021.
In their conclusion, the North Korean researchers stated that through comparative analysis, they verified that the results of the prediction model using the proposed particle swarm optimization method and long short-term memory neural network were relatively superior.
Further verification is needed to determine how accurately the North Korean researchers actually developed their Bitcoin price prediction technology. However, the North Korean researchers claimed in their paper that their newly developed model showed 84.08% accuracy in predicting the direction of Bitcoin price changes.
The content of this paper aims to generate greater profits from Bitcoin trading. It is interpreted as an intent to predict Bitcoin price trends—both upward and downward—to execute sales and purchases accordingly.
There have been numerous claims that North Korea utilizes virtual assets to evade financial sanctions and engage in money laundering. However, North Korea has never publicly disclosed matters related to its own virtual asset transactions. This North Korean paper now offers a glimpse into their virtual asset trading activities. It appears possible that North Korea is conducting virtual asset transactions not only for the purposes of sanctions evasion and money laundering, but also with the primary goal of generating foreign currency.
* 이 내용은 인공지능 서비스 DeepL에 의해 번역된 것입니다.
This article was translated by the artificial intelligence service DeepL.