A blockchain anomaly detection system was suggested by Matteo Signorini et al. [1]. It recognizes unusual transactions and prevents them from spreading. It also prevents malicious attackers from repeating assaults and creates a threat database. As a result, it's a tamper-proof solution.
The Bitcoin network has been investigated[3]. Users and transactions are the first two focuses to portray the data. Then, three key social network strategies employed to identify probable abnormal transactions, which are anomalies. In which they used K-means and the one class unsupervised svm for the identification of the anomalies in the blockchain networks. Out of the 30 known examples, they are able to identify two known cases of theft and 1 known case of loss despite the low agreement metrics.
Sirine Sayadi et al. [5] created a strategy that uses blockchain to detect abnormalities in electronic transactions. Based on the responses of these assaults on specific attributes, a dataset holding a set of regular bitcoin transactions was used to construct a dataset including a set of attacked data. The OSVM approach was employed to detect outliers, and it performed effectively.
Using the isolation forest technique, which is powered by blockchain, Xiong Yang et al.[6] provided a solution for identifying abnormalities in a wireless sensor network. The approach divides the wireless sensor network into layers, with anomaly detection conducted in separate network domains for each layer. Other superior approaches were also compared to the method.
Pham, T. et al[7] focuses on Bitcoin hacks and introduces a security method that can spot irregularities in the underlying BN traffic statistics. With the implemented prototype system that detects BT traffic and periodically generates multi-dimensional data streams, revealed data gathering strategy. We introduced our detection engine based on AE as a realisation of semi-supervised learning, and it is based on the semi-supervised learning technique that is practical and resistant to unknown threats through the profile of background valid patterns. With a set of simulated attack traces, experiments utilising the real dataset gathered from the Mainnet of the public BT.
Reid, F. et al[8] create a supervised ML model that can quickly classify anomalies in systems[8]. A low-latency anomaly detection model is built using time series that are created by processing unbounded streams of data in this model. Also expand on initial objective of only anomaly detection to include simultaneous anomaly prediction. And build a Bayesian Network framework to tackle this extremely difficult problem, capturing details about the lagged regressor parameters calibrated in the first stage of our method, and employed this structure to learn local conditional probability distributions.
The framework based on the adjustable blockchain described in this study may be used to understand the voting process issues, the best hash algorithm to use, the choice of blockchain adjustments, the process of maintaining voting data, and the security and authentication of the voting process[9]. The flexibility of blockchain technology has allowed it to be utilized to adapt to the dynamics of the electronic voting process.
Farren, D. et al[10] suggested a blockchain-based plan to address privacy concerns in 5G content-centric mobile networks. The authors put into practice the mutual trust between users and content producers. Additionally, the provider's access control and privacy are guaranteed by the blockchain ledger's openness and tamper-resistance. The writers can efficiently maintain the public ledger with the aid of a miner chosen among users. Additionally, the authors exchange valuable information while minimizing network delays, overhead, and traffic to achieve green communication.
Shahzad, B. et al[11] provided a thorough analysis of the security and privacy features of Bitcoin in this article. And provided an introduction to Bitcoin system and its key parts, including how they function and interact with one another. Also examined the flaws in Bitcoin's primary supporting technologies, including the blockchain and the PoW-based consensus algorithm.
A model made up of six distinctive properties, the primary contribution to the challenge of recognising rogue users[12]. Although not entirely accurate, they believe that this model is a reasonable place to start when examining bitcoin users to identify potential rogue users. They were able to distinguish between trustworthy users and bad users in their testing that were based on actual reported robberies and able to attain 76.5 percent accuracy for artificially created rogue users.
Conti, M. et al[13] focused on important security issues by giving a general overview of the Bitcoin protocol and outlining its main parts. Then gone into detail into the threats to and flaws in the Bitcoin system, as well as its key components including the blockchain protocol. Finally, review open research issues and trends for future research in Bitcoin security and analyze current security studies and solutions.
Zambre, D. et al[14] used the graphs of Bitcoin transactions to illustrate community patterns where, demonstrated that our feature extraction method outperforms the heuristic method for identifying the community by using a DL method to recognize Bitcoin mixing services and testing it in a dataset containing data from genuine BT and manufactured data.
Kim, J. et al[17] focuses on Bitcoin hacks and introduces a security method that can spot irregularities in the underlying BN traffic statistics[15]. They debuted their data gathering strategy with a functional prototype system that monitors BN activity and periodically generates multi-dimensional data streams using semi-supervised approaches. In comparison to using all available features, the experimental results demonstrated a significant reduction in time complexity (up to 66.8% for training and 85.7% for testing), without any performance impairment.
Hirshman, J. et al[18] used unsupervised learning algorithms as Kmeans and RolX ultimately helped them to accomplish their goal of identifying unusual activity in the network. This paper have a far better understanding of what to look for in a suspicious transaction, or more specifically, a string of suspicious transactions, according to clustering and role identification. It should be done more work to measure and classify these oddities.
This study identifies the circumstances in which various blockchain systems are unable to guarantee consensus and presents a reproducible execution on their Ethereum private chain[19]. In order to do this, defined the Blockchain Anomaly as the blockchain's inability to ensure that a committed transaction cannot be undone. The user of proof-of-work blockchains may experience drastic consequences as a result of this phenomenon.