Unit:
Department of Informatics and TelecommunicationsΠληροφορική
Supervisors info:
Κωνσταντίνος Χατζηκοκολάκης, Αναπληρωτής Καθηγητής, Τμήμα Πληροφορικής & Τηλεπικοινωνιών, Σχολή Θετικών Επιστημών
Original Title:
Εντοπισμός ύποπτων συναλλαγών στο Blockchain, με την χρήση Μοντέλων Μηχανικής Μάθησης
Translated title:
Detection of suspicious transactions in the Blockchain, using Machine Learning Models
Summary:
Blockchain was created by Satoshi Nakamoto in 2008. Its initial application was the
digital currency Bitcoin, aiming to establish a network of computers for the execution
of financial transactions with mathematically proven security, without the presence of a
central authority. Each blockchain consists of smart contracts, which are programs stored
on the blockchain that are activated when the programmed conditions are met.
From 2008 until today, the blockchain has evolved and managed to create a market
worth of 19.36 billion dollars . However, over time, malicious users have exploited
various techniques to make a profit. In 2023, 1.8 billion dollars were stolen from blockchain users . Notable incidents like the FTX hack, for example, impacted the daily lives of many users, resulting in a loss of 600 million dollars on the day the company declared bankruptcy.
Various malicious practices, such as fraud, money laundering and hacking, are used daily by users and organizations to exploit vulnerabilities in blockchain smart contracts. Often, these practices go unnoticed or are not addressed promptly due to the vast amount of information present on the blockchain or the use of techniques aimed at concealing these illicit activities.
This work, presents a program that addresses this issue. The program creates a dataset
of transactions that have occurred within a smart contract, using it to build and train
multiple machine learning models. These models perform unsupervised anomaly detection on that dataset. Thus, new transactions done with that smart contract, can be split to anomalies and normal,shortly after their publication within a block on the blockchain.
Main subject category:
Technology - Computer science
Keywords:
blockchain, hack, attack, machine learning, cyber security, IDS, intrusion detection system, anomally detection, unsupervised learning, hack
