pith. sign in

arxiv: 2605.01207 · v1 · submitted 2026-05-02 · 💻 cs.CR

Phishing Detection in Ethereum via Temporal Graph Contrastive Learning

Cong Wu , Jing Chen , Siqi Lin , Hongda Li , Ziming Zhao This is my paper

Pith reviewed 2026-05-09 14:48 UTC · model grok-4.3

classification 💻 cs.CR
keywords phishing detectionethereumtemporal graphscontrastive learningself-supervised learningblockchain securitygraph neural networkscryptocurrency
0
0 comments X

The pith

PhishEye detects Ethereum phishing by modeling transactions as heterogeneous temporal attributed multi-graphs and applying self-supervised temporal graph contrastive learning.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces PhishEye as a fully dynamic self-supervised system that monitors on-chain Ethereum transactions to identify phishing. It represents the data as a heterogeneous temporal attributed multi-graph to handle varied transaction types and their irregular timing, then trains a contrastive learning model to separate phishing patterns from normal activity without labeled examples. Prior methods fall short because static graphs miss evolution over time, homogeneous representations ignore transaction diversity, and supervised approaches require scarce labels that do not adapt to new threats. A reader would care because effective detection can flag attacks early and limit financial damage in decentralized finance.

Core claim

PhishEye formulates Ethereum transactions as a heterogeneous temporal attributed multi-graph and incorporates a novel temporal graph contrastive learning model that captures both temporal patterns and heterogeneous transaction types, enabling self-supervised detection that outperforms existing approaches on transaction and account tasks while identifying new phishing addresses in live deployment.

What carries the argument

The heterogeneous temporal attributed multi-graph representation of Ethereum transactions together with a temporal graph contrastive learning objective that learns distinguishing features across time and transaction types without supervision.

If this is right

  • Phishing transaction detection reaches an F1 score of 87.23% and AUC of 98.43%.
  • Phishing account detection reaches an F1 score of 94.19% and AUC of 98.03%.
  • The system identified 1,803 previously unknown phishing addresses during 15 months of real-world operation.
  • Early alerts from the system helped prevent losses exceeding 2 billion USD.
  • The self-supervised design supports ongoing monitoring without the need for extensive new labels as threats change.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same temporal multi-graph contrastive approach could be adapted to detect other on-chain anomalies such as money laundering or rug pulls.
  • Emphasizing bursty temporal features may improve anomaly detection in other high-volume event streams like network traffic or financial ledgers.
  • Long-term deployment success indicates the method may resist concept drift better than models that require periodic retraining on new labels.
  • Combining the graph model with wallet-level alerts could form part of a layered defense for DeFi users and exchanges.

Load-bearing premise

That phishing and legitimate Ethereum activity produce reliably separable patterns in the temporal and heterogeneous structure of the multi-graph, allowing contrastive learning to generalize without labeled data.

What would settle it

A measurable drop in F1 and AUC scores when the trained model is evaluated on fresh Ethereum transaction data after July 2024 that includes new phishing tactics designed to match the timing and type distributions of normal activity.

Figures

Figures reproduced from arXiv: 2605.01207 by Cong Wu, Hongda Li, Jing Chen, Siqi Lin, Ziming Zhao.

Figure 2
Figure 2. Figure 2: Workflow of PhishEye of each node (account) or edge (transaction) can be different. For example, the accounts could be EOA or CA, while the transaction types could be Ether transfers, FT transfers, NFT transfers, contract interactions, or internal contract transactions. Therefore, the graph 𝐻𝑇𝐴𝑀𝐺 is a heterogeneous temporal attributed multi-graph. Our goal is to effectively learn how the transactional beha… view at source ↗
Figure 3
Figure 3. Figure 3: Distribution of phishing addresses’ indegree (a), view at source ↗
Figure 4
Figure 4. Figure 4: PDF and density of prediction scores in detecting phishing transactions (a) and accounts (b), and detecting phishing view at source ↗
Figure 5
Figure 5. Figure 5: ROC curves under different sampled neighbors view at source ↗
Figure 6
Figure 6. Figure 6: Precision, recall, F1, AUC, and BAC in detecting phishing transactions (a) and accounts (b) under the model with or view at source ↗
Figure 8
Figure 8. Figure 8: FNR/FPR of different methods in detecting real view at source ↗
Figure 9
Figure 9. Figure 9: Case 1 with the phishing address, 0x69..55: fund view at source ↗
Figure 10
Figure 10. Figure 10: Case 2 with the phishing address, 0x1d..df: fund view at source ↗
Figure 11
Figure 11. Figure 11: Case 3 with the phishing address, 0x9f..26: fund view at source ↗
read the original abstract

Blockchain and decentralized finance have revolutionized the financial ecosystem while simultaneously exposing it to cryptocurrency phishing attacks. Existing phishing detection methods primarily rely on graph learning, but they face significant limitations. Static graph learning approaches fail to account for the temporal evolution of phishing patterns, while semi-dynamic methods, such as those combining static GNNs with LSTM, struggle to capture the irregular and bursty nature of blockchain transactions. Moreover, these methods overlook the diversity of Ethereum transactions, treating them as homogeneous graphs, and heavily rely on supervised learning, which requires extensive labeled data that is not readily available. These limitations reduce their adaptability to emerging phishing threats. In this paper, we present PhishEye, a fully dynamic self-supervised system that monitors on-chain transactions to detect phishing activities. PhishEye formulates Ethereum transactions as a heterogeneous temporal attributed multi-graph and incorporates a novel temporal graph contrastive learning model, which captures both temporal patterns and heterogeneous transaction types. The evaluation on a dataset of 161,658 addresses and 416,541 transactions shows that PhishEye outperforms existing methods, achieving an F1 score of 87.23% and an AUC of 98.43% for phishing transaction detection, and an F1 score of 94.19% and an AUC of 98.03% for phishing account detection. In real-world deployment from May 1, 2023 to July 31, 2024, PhishEye identified 1,803 previously unknown phishing addresses, providing early alerts that helped prevent losses exceeding 2 billion USD.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript introduces PhishEye, a self-supervised phishing detection system for Ethereum that models transactions as a heterogeneous temporal attributed multi-graph and employs temporal graph contrastive learning to capture temporal patterns and transaction heterogeneity. On a dataset comprising 161,658 addresses and 416,541 transactions, it achieves F1 scores of 87.23% and 94.19% along with AUCs of 98.43% and 98.03% for transaction and account detection, respectively. Real-world deployment over 15 months identified 1,803 unknown phishing addresses and is claimed to have prevented losses exceeding 2 billion USD.

Significance. If the central claims hold, the work is significant for advancing self-supervised methods in blockchain security, where labeled data is limited. The use of temporal heterogeneous graphs addresses key limitations of static and homogeneous approaches in prior work. The inclusion of real-world deployment results provides valuable evidence of practical utility beyond benchmark datasets. This could influence future research on adaptive, label-efficient detection systems for evolving threats like phishing in DeFi.

major comments (2)
  1. [Temporal graph contrastive learning model] The construction of positive and negative pairs for the contrastive objective is not sufficiently specified (model description section). Given that the self-supervised separation of phishing and legitimate behaviors relies entirely on these pairs (likely based on temporal proximity or attribute matching), the lack of explicit criteria makes it impossible to evaluate whether the embeddings capture robust, generalizable signals or merely dataset-specific correlations. This is load-bearing for the performance claims and generalization to future tactics.
  2. [Evaluation] The evaluation section lacks detailed ablation studies on the impact of the heterogeneous and temporal components or on negative sample construction strategies. Without this, the reported outperformance over baselines (F1 87.23%, AUC 98.43% for transactions) cannot be confidently attributed to the proposed model rather than data characteristics or evaluation setup.
minor comments (1)
  1. [Real-world deployment] Clarify in the deployment section how the 1,803 addresses were verified as previously unknown phishing addresses and how the prevented loss estimate of over 2 billion USD was calculated, to strengthen the real-world evidence.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript describing PhishEye. The comments highlight important areas for improving clarity and rigor, particularly around the contrastive learning details and evaluation. We address each major comment below and will incorporate revisions to strengthen the paper.

read point-by-point responses

  1. Referee: [Temporal graph contrastive learning model] The construction of positive and negative pairs for the contrastive objective is not sufficiently specified (model description section). Given that the self-supervised separation of phishing and legitimate behaviors relies entirely on these pairs (likely based on temporal proximity or attribute matching), the lack of explicit criteria makes it impossible to evaluate whether the embeddings capture robust, generalizable signals or merely dataset-specific correlations. This is load-bearing for the performance claims and generalization to future tactics.

    Authors: We appreciate the referee's emphasis on this critical component. The manuscript outlines the temporal graph contrastive learning framework on the heterogeneous temporal attributed multi-graph but does not provide sufficiently explicit rules or pseudocode for generating positive and negative pairs. We agree this limits assessment of whether the learned embeddings capture generalizable temporal and heterogeneous signals. In the revised manuscript, we will expand the model description section with a detailed specification of the pair construction process, including the criteria based on temporal proximity and attribute matching, along with an algorithm for sampling. revision: yes

  2. Referee: [Evaluation] The evaluation section lacks detailed ablation studies on the impact of the heterogeneous and temporal components or on negative sample construction strategies. Without this, the reported outperformance over baselines (F1 87.23%, AUC 98.43% for transactions) cannot be confidently attributed to the proposed model rather than data characteristics or evaluation setup.

    Authors: We concur that additional ablation experiments are needed to attribute performance gains specifically to the heterogeneous and temporal aspects of the model, as well as to the negative sampling approach. The current evaluation demonstrates outperformance but does not isolate these factors. In the revised version, we will add a new subsection with ablation studies: variants without the temporal modeling, without heterogeneity (homogeneous graph), and with alternative negative sampling strategies. These will be reported with corresponding F1 and AUC metrics to clarify the contributions. revision: yes

Circularity Check

0 steps flagged

No circularity: self-supervised contrastive objective and held-out/live evaluation are independent of fitted parameters.

full rationale

The paper defines a heterogeneous temporal attributed multi-graph from Ethereum transactions and trains a temporal graph contrastive learning model whose objective operates directly on the observed graph structure and attributes. Performance is measured via F1/AUC on a held-out portion of the 416k-transaction collection plus a separate 2023-2024 live deployment that identified 1,803 new addresses. Neither the contrastive loss nor the reported metrics reduce by construction to the model's own fitted values; the pair-construction and embedding separation are data-driven rather than tautological. No self-citation is shown to be load-bearing for the central claim, and no uniqueness theorem or ansatz is imported from prior author work to force the result.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the design choice that a heterogeneous temporal multi-graph plus contrastive learning can capture phishing signals without labels; this is an engineering assumption rather than a derivation from first principles.

free parameters (1)
  • model hyperparameters (embedding size, contrastive temperature, temporal window, etc.)
    Standard in graph contrastive learning; values are chosen or tuned on data and directly affect learned representations.
axioms (1)
  • domain assumption Ethereum transaction data can be faithfully represented as a heterogeneous temporal attributed multi-graph whose structure encodes phishing-relevant patterns.
    Invoked when the paper formulates the input to the contrastive model.

pith-pipeline@v0.9.0 · 5583 in / 1529 out tokens · 62290 ms · 2026-05-09T14:48:06.458354+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

78 extracted references · 5 canonical work pages

  1. [1]

    [n. d.]. Annual Scams and Cybercrime Brief 2024. https://www.scamshield.gov.s g/files/Scams%20and%20Cybercrime%20Briefs/2024_annual_scams_and_cyber crime_brief.pdf

    2024

  2. [2]

    [n. d.]. chainalysis. https://www.chainalysis.com/blog/approval-phishing- cryptocurrency-scams-2023/

    2023

  3. [3]

    [n. d.]. Introducing a Cutting-Edge ML-Powered Detection Bot for Phishing Scams. https://forta.org/blog/introducing-a-cutting-edge-ml-powered- detection-bot-for-phishing-scams/

  4. [4]

    [n. d.]. What Is a Cross-Chain Bridge? https://chain.link/education-hub/cross- chain-bridge

  5. [5]

    blocksec

    2024. blocksec. https://blocksec.com/

    2024

  6. [6]

    2024. certik. https://www.certik.com/

    2024

  7. [7]

    Crypto Mixers and AML Compliance

    2024. Crypto Mixers and AML Compliance. https://www.chainalysis.com/blog/c rypto-mixers/

    2024

  8. [8]

    Ethereum TVL

    2024. Ethereum TVL. https://defillama.com/chain/Ethereum

    2024

  9. [9]

    Four Crypto And Blockchain Phishing Scams To Be Aware Of Before You Get Excited About Web3

    2024. Four Crypto And Blockchain Phishing Scams To Be Aware Of Before You Get Excited About Web3. https://www.forbes.com/sites/forbestechcouncil/202 2/04/27/four-crypto-and-blockchain-phishing-scams-to-be-aware-of-before- you-get-excited-about-web3/?sh=2aca988d25c5

    2024

  10. [10]

    2024. goplus. https://gopluseco.io/

    2024

  11. [11]

    Know Your Customer (KYC)

    2024. Know Your Customer (KYC). https://academy.binance.com/en/glossary/kn ow-your-customer

    2024

  12. [12]

    Label Word Cloud

    2024. Label Word Cloud. https://etherscan.io/labelcloud

    2024

  13. [13]

    LightGBM

    2024. LightGBM. https://github.com/microsoft/LightGBM

    2024

  14. [14]

    Reverse Engineering Inferno Drainer

    2024. Reverse Engineering Inferno Drainer. https://muellerberndt.medium.com/a- peek-inside-inferno-drainer-0a69647b85ca

    2024

  15. [15]

    Scamsniffer

    2024. Scamsniffer. https://scamsniffer.io/

    2024

  16. [16]

    slowmist

    2024. slowmist. https://www.slowmist.com/

    2024

  17. [17]

    Tornado.cash

    2024. Tornado.cash. https://tornado.ws/

    2024

  18. [18]

    Ahod Alghuried and David Mohaisen. 2026. Phishing in Wonderland: Evaluat- ing Learning-Based Ethereum Phishing Transaction Detection and Pitfalls. In Proceedings of the NDSS Symposium

    2026

  19. [19]

    Shaked Brody, Uri Alon, and Eran Yahav. 2021. How Attentive are Graph Atten- tion Networks?. InInternational Conference on Learning Representations (ICLR)

    2021

  20. [20]

    Federico Cernera, Massimo La Morgia, Alessandro Mei, and Francesco Sassi. 2023. Token Spammers, Rug Pulls, and Sniper Bots: An Analysis of the Ecosystem of Tokens in Ethereum and in the Binance Smart Chain (BNB). InUSENIX Security Symposium

    2023

  21. [21]

    Ting Chen, Yufei Zhang, Zihao Li, Xiapu Luo, Ting Wang, Rong Cao, Xiuzhuo Xiao, and Xiaosong Zhang. 2019. Tokenscope: Automatically detecting inconsis- tent behaviors of cryptocurrency tokens in ethereum. InACM SIGSAC Conference on Computer and Communications Security (CCS)

    2019

  22. [22]

    Weili Chen, Xiongfeng Guo, Zhiguang Chen, Zibin Zheng, and Yutong Lu. 2020. Phishing Scam Detection on Ethereum: Towards Financial Security for Blockchain Ecosystem.. InInternational Joint Conference on Artificial Intelligence (IJCAI)

    2020

  23. [23]

    Weili Chen, Tuo Zhang, Zhiguang Chen, Zibin Zheng, and Yutong Lu. 2020. Traveling the token world: A graph analysis of ethereum erc20 token ecosystem. InACM Web conference (WWW)

    2020

  24. [24]

    Dipanjan Das, Priyanka Bose, Nicola Ruaro, Christopher Kruegel, and Giovanni Vigna. 2022. Understanding security issues in the NFT ecosystem. InACM SIGSAC Conference on Computer and Communications Security (CCS)

    2022

  25. [25]

    Bingxue Fu, Yixuan Wang, and Tao Feng. 2024. CT-GCN+: a high-performance cryptocurrency transaction graph convolutional model for phishing node classi- fication.Cybersecurity(2024)

    2024

  26. [26]

    Medhasree Ghosh, Dyuti Ghosh, Raju Halder, and Joydeep Chandra. 2023. Inves- tigating the impact of structural and temporal behaviors in Ethereum phishing users detection.Blockchain: Research and Applications(2023)

    2023

  27. [27]

    Medhasree Ghosh, Swapnil Srivastava, Apoorva Upadhyaya, Raju Halder, and Joydeep Chandra. 2025. Catalog: Exploiting joint temporal dependencies for enhanced phishing detection on ethereum. InProceedings of the ACM on Web Conference (WWW)

    2025

  28. [28]

    Justin Gilmer, Samuel S Schoenholz, Patrick F Riley, Oriol Vinyals, and George E Dahl. 2017. Neural message passing for quantum chemistry. InInternational Conference on Machine Learning (ICML)

    2017

  29. [29]

    Shixuan Guan and Kai Li. 2024. Characterizing Ethereum address poisoning attack. InACM SIGSAC Conference on Computer and Communications Security (CCS)

    2024

  30. [30]

    Will Hamilton, Zhitao Ying, and Jure Leskovec. 2017. Inductive representation learning on large graphs. InAdvances in Neural Information Processing Systems (NeurIPS)

    2017

  31. [31]

    Bowen He, Yuan Chen, Zhuo Chen, Xiaohui Hu, Yufeng Hu, Lei Wu, Rui Chang, Haoyu Wang, and Yajin Zhou. 2023. TxPhishScope: Towards Detecting and Under- standing Transaction-based Phishing on Ethereum. InACM SIGSAC Conference on Computer and Communications Security (CCS)

    2023

  32. [32]

    Bowen He, Xiaohui Hu, Yufeng Hu, Ting Yu, Rui Chang, Lei Wu, and Yajin Zhou

  33. [33]

    Phishing tactics are evolving: an empirical study of phishing contracts on ethereum.Proceedings of the ACM on Measurement and Analysis of Computing Systems (POMACS)(2025)

    2025

  34. [34]

    Zheyuan He, Shuwei Song, Yang Bai, Xiapu Luo, Ting Chen, Wensheng Zhang, Peng He, Hongwei Li, Xiaodong Lin, and Xiaosong Zhang. 2023. Tokenaware: Accurate and efficient bookkeeping recognition for token smart contracts.ACM Transactions on Software Engineering and Methodology (TOSEM)(2023). 13 CCS ’26, November 15–19, 2026, Hague, Netherlands Cong Wu et al

    2023

  35. [35]

    Baoying Huang, Jieli Liu, Jiajing Wu, Quanzhong Li, and Dan Lin. 2023. Ethereum Phishing Fraud Detection Based on Heterogeneous Transaction Subnets. InIEEE International Symposium on Circuits and Systems (ISCAS)

    2023

  36. [36]

    Hexiang Huang, Xuan Zhang, Jishu Wang, Chen Gao, Xue Li, Rui Zhu, and Qiuying Ma. 2024. PEAE-GNN: Phishing detection on Ethereum via augmentation ego-graph based on graph neural network.IEEE Transactions on Computational Social Systems(2024)

    2024

  37. [37]

    Lisa JY Tan. 2024. NFT security. InWeb3 Applications Security and New Security Landscape: Theories and Practices

    2024

  38. [38]

    Hanna Kim, Jian Cui, Eugene Jang, Chanhee Lee, Yongjae Lee, Jin-Woo Chung, and Seungwon Shin. 2023. DRAINCLoG: Detecting Rogue Accounts with Illegally-obtained NFTs using Classifiers Learned on Graphs.arXiv preprint arXiv:2301.13577(2023)

    work page arXiv 2023

  39. [39]

    Thomas N Kipf and Max Welling. 2016. Semi-Supervised Classification with Graph Convolutional Networks. InInternational Conference on Learning Repre- sentations

    2016

  40. [40]

    Anastasiya Lavrova and Oleg Iakushkin. 2022. NFT performance and security review. InInternational Conference on Computational Science and Its Applications

    2022

  41. [41]

    Sijia Li, Gaopeng Gou, Chang Liu, Chengshang Hou, Zhenzhen Li, and Gang Xiong. 2022. TTAGN: Temporal transaction aggregation graph network for ethereum phishing scams detection. InACM Web Conference (WWW)

    2022

  42. [42]

    Sijia Li, Gaopeng Gou, Chang Liu, Gang Xiong, Zhen Li, Junchao Xiao, and Xinyu Xing. 2023. TGC: Transaction Graph Contrast Network for Ethereum Phishing Scam Detection. InAnnual Computer Security Applications Conference (ACSAC)

    2023

  43. [43]

    Shucheng Li, Runchuan Wang, Hao Wu, Sheng Zhong, and Fengyuan Xu. 2023. SIEGE: Self-Supervised Incremental Deep Graph Learning for Ethereum Phishing Scam Detection. InACM International Conference on Multimedia (MM)

    2023

  44. [44]

    Wenkai Li, Jiuyang Bu, Xiaoqi Li, and Xianyi Chen. 2022. Security analysis of DeFi: Vulnerabilities, attacks and advances. InIEEE International Conference on Blockchain (Blockchain)

    2022

  45. [45]

    Xigao Li, Anurag Yepuri, and Nick Nikiforakis. 2023. Double and Nothing: Understanding and Detecting Cryptocurrency Giveaway Scams. InNetwork and Distributed Systems Security (NDSS) Symposium

    2023

  46. [46]

    Ruichao Liang, Jing Chen, Kun He, Yueming Wu, Gelei Deng, Ruiying Du, and Cong Wu. 2024. Ponziguard: Detecting ponzi schemes on ethereum with contract runtime behavior graph (crbg). InIEEE/ACM International Conference on Software Engineering (ICSE)

    2024

  47. [47]

    Ruichao Liang, Jing Chen, Cong Wu, Kun He, Yueming Wu, Ruochen Cao, Ruiying Du, Ziming Zhao, and Yang Liu. 2025. Vulseye: Detect smart contract vulnera- bilities via stateful directed graybox fuzzing.IEEE Transactions on Information Forensics and Security(2025)

    2025

  48. [48]

    Jieli Liu, Jinze Chen, Jiajing Wu, Zhiying Wu, Junyuan Fang, and Zibin Zheng

  49. [49]

    Fishing for Fraudsters: Uncovering Ethereum Phishing Gangs with Blockchain Data.IEEE Transactions on Information Forensics and Security (TIFS) (2024)

    2024

  50. [50]

    Tao Liu, Aimin Jiang, Jia Zhou, Min Li, and Hon Keung Kwan. 2023. GraphSAGE- based dynamic spatial–temporal graph convolutional network for traffic predic- tion.IEEE Transactions on Intelligent Transportation Systems(2023)

    2023

  51. [51]

    Zhaowei Liu, Haitao Yang, Zhizhong Liu, Xiangfu Zhao, JinDong Zhao, Dong Yang, and Yao Shan. 2025. Phishing Detection on Ethereum via Graph Neural Architecture Search of Transaction Subgraph.IEEE Transactions on Computational Social Systems(2025)

    2025

  52. [52]

    Haifeng Lv and Yong Ding. 2023. Phishing detection on Ethereum via transaction subgraphs embedding.IET Blockchain(2023)

    2023

  53. [53]

    Xingyu Lyu, Mengya Zhang, Xiaokuan Zhang, Jianyu Niu, Yinqian Zhang, and Zhiqiang Lin. 2022. An Empirical Study on Ethereum Private Transactions and the Security Implications.arXiv preprint arXiv:2208.02858(2022)

    work page arXiv 2022

  54. [54]

    Fuchen Ma, Meng Ren, Lerong Ouyang, Yuanliang Chen, Juan Zhu, Ting Chen, Yingli Zheng, Xiao Dai, Yu Jiang, and Jiaguang Sun. 2023. Pied-piper: Revealing the backdoor threats in ethereum erc token contracts.ACM Transactions on Software Engineering and Methodology (TOSEM)(2023)

    2023

  55. [55]

    Kai Ma, Jintao Huang, Ningyu He, Zhuo Wang, and Haoyu Wang. 2025. SoK: On the Security of Non-Fungible Tokens.Blockchain: Research and Applications (2025)

    2025

  56. [56]

    Tomas Mikolov, Ilya Sutskever, Kai Chen, Greg S Corrado, and Jeff Dean. 2013. Distributed representations of words and phrases and their compositionality. (2013)

    2013

  57. [57]

    Mritunjay Shall Peelam, Vinay Chamola, Aditya Kumar Sharma, and Brijesh Ku- mar Chaurasia. 2025. Decentralized Trust: NFT and Blockchain-Enabled Evidence System using Fog Computing.Blockchain: Research and Applications(2025)

    2025

  58. [58]

    Bryan Perozzi, Rami Al-Rfou, and Steven Skiena. 2014. Deepwalk: Online learning of social representations. InACM SIGKDD International Conference on Knowledge Discovery and Data Mining

    2014

  59. [59]

    Ali Rahimi and Benjamin Recht. 2007. Random features for large-scale kernel machines. InAdvances in Neural Information Processing Systems (NeurIPS)

    2007

  60. [60]

    Sayak Saha Roy, Dipanjan Das, Priyanka Bose, Christopher Kruegel, Giovanni Vigna, and Shirin Nilizadeh. 2023. Demystifying NFT Promotion and Phishing Scams.arXiv preprint arXiv:2301.09806(2023)

    work page arXiv 2023

  61. [61]

    Jason Scharfman. 2024. Crypto phishing and spoofing scams. InThe Cryptocur- rency and Digital Asset Fraud Casebook, Volume II: DeFi, NFTs, DAOs, Meme Coins, and Other Digital Asset Hacks

    2024

  62. [62]

    Shuwei Song, Ting Chen, Ao Qiao, Xiapu Luo, Leqing Wang, Zheyuan He, Ting Wang, Xiaodong Lin, Peng He, Wensheng Zhang, et al. 2025. Automated and Accurate Token Transfer Identification and Its Applications in Cryptocurrency Security.ACM on Software Engineering (ASE)(2025)

    2025

  63. [63]

    Gaurav Varshney, Rahul Kumawat, Vijay Varadharajan, Uday Tupakula, and Chandranshu Gupta. 2024. Anti-phishing: A comprehensive perspective.Expert Systems with Applications(2024)

    2024

  64. [64]

    Haixian Wen, Junyuan Fang, Jiajing Wu, and Zibin Zheng. 2022. Hide and seek: An adversarial hiding approach against phishing detection on ethereum.IEEE Transactions on Computational Social Systems(2022)

    2022

  65. [65]

    Cong Wu, Jing Chen, Ziming Zhao, Kun He, Guowen Xu, Yueming Wu, Haijun Wang, Hongwei Li, Yang Liu, and Yang Xiang. 2024. Tokenscout: Early detection of ethereum scam tokens via temporal graph learning. InACM SIGSAC Conference on Computer and Communications Security (CCS)

    2024

  66. [66]

    Jiajing Wu, Qi Yuan, Dan Lin, Wei You, Weili Chen, Chuan Chen, and Zibin Zheng. 2020. Who are the phishers? phishing scam detection on ethereum via network embedding.IEEE Transactions on Systems, Man, and Cybernetics: Systems (2020)

    2020

  67. [67]

    Pengcheng Xia, Haoyu Wang, Bingyu Gao, Weihang Su, Zhou Yu, Xiapu Luo, Chao Zhang, Xusheng Xiao, and Guoai Xu. 2021. Trade or trick? detecting and characterizing scam tokens on uniswap decentralized exchange.ACM on Measurement and Analysis of Computing Systems (SIGMETRICS)(2021)

    2021

  68. [68]

    Siyi Xiao, Lejun Zhang, Zhihong Tian, Shen Su, Jing Qiu, and Ran Guo. 2025. Pheromone-based graph embedding algorithm for Ethereum phishing detection. Computer Networks(2025)

    2025

  69. [69]

    Da Xu, Chuanwei Ruan, Evren Korpeoglu, Sushant Kumar, and Kannan Achan

  70. [70]

    InInternational Conference on Learning Representations (ICLR)

    Inductive representation learning on temporal graphs. InInternational Conference on Learning Representations (ICLR)

  71. [71]

    Jiahua Xu and Benjamin Livshits. 2019. The anatomy of a cryptocurrency Pump- and-Dump scheme. InUSENIX Security Symposium

    2019

  72. [72]

    Jingjing Yang, Jieli Liu, and Jiajing Wu. 2023. With Trail to Follow: Measurements of Real-world Non-fungible Token Phishing Attacks on Ethereum.arXiv preprint arXiv:2307.01579(2023)

    work page arXiv 2023

  73. [73]

    Lin Zhao, Sourav Sen Gupta, Arijit Khan, and Robby Luo. 2021. Temporal analysis of the entire ethereum blockchain network. InProceedings of the Web Conference (WWW). 2258–2269

    2021

  74. [74]

    Weilin Zheng, Bo Liu, Hong-Ning Dai, Zigui Jiang, Zibin Zheng, and Muhammad Imran. 2022. Unravelling token ecosystem of eosio blockchain.arXiv preprint arXiv:2202.11201(2022)

    work page arXiv 2022

  75. [75]

    Liyi Zhou, Kaihua Qin, Christof Ferreira Torres, Duc V Le, and Arthur Gervais

  76. [76]

    InIEEE Symposium on Security and Privacy (SP)

    High-frequency trading on decentralized on-chain exchanges. InIEEE Symposium on Security and Privacy (SP)

  77. [77]

    Liyi Zhou, Xihan Xiong, Jens Ernstberger, Stefanos Chaliasos, Zhipeng Wang, Ye Wang, Kaihua Qin, Roger Wattenhofer, Dawn Song, and Arthur Gervais. 2023. Sok: Decentralized finance (defi) attacks. InIEEE Symposium on Security and Privacy (Oakland)

    2023

  78. [78]

    Yuanhang Zhou, Jingxuan Sun, Fuchen Ma, Yuanliang Chen, Zhen Yan, and Yu Jiang. 2024. Stop pulling my rug: Exposing rug pull risks in crypto token to in- vestors. InInternational Conference on Software Engineering: Software Engineering in Practice. 14 Phishing Detection in Ethereum via Temporal Graph Contrastive Learning CCS ’26, November 15–19, 2026, Hag...

    2024