Pith. sign in

REVIEW 4 major objections 6 minor 51 references

BotHash: Efficient and Training-Free Bot Detection Through Approximate Nearest Neighbor

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read BotHash claims a training-free hash pipeline can detect social bots, including LLM-generated ones, by voting among approximate-nearest-neighbor reference users.

desk verdict BotHash is a genuinely new training-free bot-detection pipeline, but its headline numbers are tuned on the test set—worth reviewing, not worth trusting as reported. read the letter →

arxiv 2506.20503 v1 pith:FOGTZ25X submitted 2025-06-25 cs.SI

Edoardo Di Paolo , Fabio De Gaspari , Angelo Spognardi This is my paper
classification cs.SI
keywords socialbotdetectionDigitalDNAMinHashLocalitySensitiveHashingapproximatenearestneighborLLM-poweredbotstraining-freeclassificationTwitter/X
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

BotHash sets out to show that social-bot detection does not need a trained classifier. Its pipeline encodes each user's tweet stream as a string of symbols (tweet type, content category, and time gaps between posts), segments that string into shingles, compresses the shingle set with MinHash, and labels an unknown account by majority vote over the labeled users whose Locality-Sensitive-Hashing buckets collide with it. The paper claims this simple, training-free pipeline reliably separates humans from bots on standard benchmarks, including the Fox-8 corpus of bots whose posts were generated by large language models, and that it matches or exceeds the F1 of the trained machine- and deep-learning baselines considered. If true, the practical consequence is that an operator can detect bots almost immediately, with a small labeled seed set and seconds of preprocessing, instead of assembling large training corpora. The paper itself limits the claim by showing that performance degrades sharply when ground truth and test users come from behaviorally distant datasets.

What carries the argument

The central machinery is the combination of Digital-DNA encodings with MinHash signatures and Locality-Sensitive Hashing. Digital-DNA is a character encoding of a user's tweet history, and Multiple Digital-DNA interleaves several alphabets so a single sequence carries tweet type, content category, and timing information; the new $B^9_{\text{temporal}}$ alphabet records which of nine time-interval bins separates consecutive tweets. The argument's load-bearing identity is that MinHashing preserves Jaccard similarity between the $k$-shingle sets of two users, while LSH turns high signature similarity into bucket collisions; the majority-vote rule of Equation 7 labels an account as a bot when more than half of its colliding reference users are bots. This gives the classifier local sensitivity, meaning two accounts need only share one behavioral shingle to become neighbors.

What would settle it

Assemble a new, independently verified bot corpus from activity that postdates the training-style reference, freeze a reference set built only from Cresci-15, Cresci-17, Cresci-18, Twibot-22, and Fox-8, and measure BotHash's F1 on the new corpus; if the score stays at the paper's cross-dataset lows rather than the in-distribution highs, the claim that the local-signature mechanism can recognize genuinely new bot behaviors is falsified.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that a user's tweet stream can be compressed to a small behavioral signature whose local collisions already separate humans from bots. BotHash encodes each account as a Digital-DNA sequence (with alphabets for tweet type, content category, and time gaps), slices the sequence into $k$-shingles, compresses the shingle set with MinHash into a fixed-size signature, and then uses Locality-Sensitive Hashing to find labeled reference accounts whose signatures collide. Equation 7 makes the final decision by majority vote over those colliding labels. The paper reports that this training-free pipeline achieves an F1 above 91% on the Cresci benchmarks, 75.83% on Twibot-22, and 98.39% on Fox-8, matching or beating every considered ML and DL baseline on four of five fixed-split benchmarks; with as few as 20 tweets per user and with only 10% of the reference data, performance remains within a few points of the best result.

Load-bearing premise

The method labels an unknown account only on the basis of which already-labeled accounts it is nearest to, so the whole detector depends on that labeled reference set already containing the same kinds of bot behavior the detector will meet.

Editorial extensions

If this is right

  • An operator with only a small labeled seed set can classify a new account in constant lookup time after a few seconds of preprocessing; the paper measures roughly 20 seconds end-to-end on Twibot-22.
  • With just 20 tweets per user, the F1 stays above 85% on the Cresci datasets, so a bot can be flagged almost as soon as it starts posting.
  • Because classification is a hash lookup, refreshing the ground truth for a new bot wave does not require retraining; the operator only rehashes and adds the new labeled accounts.
  • The paper's cross-dataset table shows that this in-distribution strength is conditional: when the reference set is temporally or behaviorally far from the test set, F1 can fall to about 45-54%, so deployment needs ongoing ground-truth refresh.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial: Because the classifier is just a hash lookup, a deployed system could continuously add newly confirmed bots to the reference set and improve over time without retraining; the paper does not test this continual-refresh regime.
  • Editorial: The same sequence-encoding pipeline could be applied to other chronological behavior logs (browser traces, API logs, payment sequences) where bot-like repetition is a signal, though BotHash itself is only evaluated on X data.
  • Editorial: The Fox-8 results suggest the detector may be reading timing and syntactic structure rather than lexical meaning; an ablation study that shuffles words while preserving the temporal/alphabet pattern would identify which component carries the signal.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The manuscript introduces BotHash, a training-free social bot detector. User activity is encoded as Digital-DNA sequences using three alphabets (B3_type, B5_content, and a newly proposed B9_temporal), optionally combined into Multiple-Digital DNA. Each user's sequence is shingled and MinHashed, and classification is performed by LSH-based approximate nearest neighbor search with majority voting (Eq. 7). The paper reports F1 and accuracy on five benchmark datasets, compares against ML/DL baselines, and analyzes cross-dataset generalization, early detection with limited tweets, small ground-truth sizes, and resource usage.

Significance. If the reported results hold, BotHash would be a valuable contribution: a training-free, lightweight detector that achieves competitive or superior F1 to trained baselines while needing very few labeled users and few tweets per user. The proposal of the temporal alphabet and MD-DNA combination is a reasonable extension of prior Digital-DNA work, and the efficiency measurements are useful. However, the current evaluation does not firmly establish the headline superiority claim because hyperparameters appear to be selected on the test sets, and the results lack repeated-run variability estimates. The paper's own cross-dataset results (e.g., Table 5, Cresci-15 to Twibot-22 F1=53.19%) show that the general claim needs careful qualification.

major comments (4)
  1. [5.1.2/Table 7] The hyperparameters in Table 7 (k-shingle size, Jaccard threshold, alphabet combination) are reported as found through grid search, but no validation split or nested selection procedure is described; Table 2 similarly selects the best alphabet per dataset by comparing test-set F1. Because the configuration is chosen using the same test data that produces the headline numbers, the comparison in Table 3 is not an out-of-sample evaluation, and the baselines are not given an equivalent test-set tuning opportunity. Please re-run with hyperparameters selected on a separate validation split (or via nested cross-validation) and report test results for the fixed configuration.
  2. [5.1.2/Table 3] No error bars, confidence intervals, or repeated runs are reported, and the Twibot-22 subset is described as randomly sampled without a stated seed. The margin over the closest baseline on Twibot-22 is only about 2.5 F1 points (75.83 vs. 73.29, Table 3), so with a single unseeded run it is unclear whether the difference is beyond selection noise. Please provide multiple runs with different seeds (or at least error bars) and state the subsampling seed.
  3. [5.3.1/Table 5] The parameters reported in Table 5 do not always match the statement that the best alphabet from Table 2 was used (e.g., Cresci-15 uses B3_type+B9_temporal in Table 5, whereas Table 2's best for Cresci-15 is B3_type+B5_content). It appears that each cross-dataset pair may have been tuned on the test set, but this is not stated. Please clarify the selection protocol for Table 5. In addition, the large drops in this table (e.g., Cresci-15 to Twibot-22 F1=53.19%) should be reflected in the abstract, which currently states that BotHash effectively differentiates bot and human accounts even for LLM-generated content without noting that this holds mainly for in-distribution test users.
  4. [3.3-3.4/Table 7] The manuscript lists signature length (number of MinHash permutations) and LSH band/row configuration as influencing the accuracy/speed tradeoff, but Table 7 does not report these values. Without them, the method is not fully reproducible and it is unclear whether these additional degrees of freedom were also tuned. Please report all hyperparameters, including signature length and any band/row parameters, and state which were fixed a priori.
minor comments (6)
  1. [3.2.1] The reference 'Appendix??' appears unresolved; the detailed alphabet definitions mentioned there are missing.
  2. [Equation 7] The variables y_v and y_u are not defined; specify that labels are, for example, 1 for bot and 0 for human.
  3. [5.1.2/Takeaway] The sentence 'BotHash shows superior performance compared to the considered approaches in terms of F1 across all datasets, with the exception of Cresci-15' is internally inconsistent; rephrase to 'across all datasets except Cresci-15, where SGBot is best.'
  4. [5.2/Table 4] The text refers to 'Bothash' instead of 'BotHash' in the resource-usage paragraph.
  5. [Table 3] The formatting of Table 3 is dense and difficult to read; consider splitting into per-dataset panels or using clearer column grouping.
  6. [5.1.2] The explanation that Table 2 and Table 3 differ because of different user sets is useful; please quantify how many users are in each split.

Circularity Check

0 steps flagged · score 2.0 of 10

No derivation-level circularity; the BotHash pipeline is self-contained, with only a test-set hyperparameter-selection concern that is an evaluation-protocol issue rather than a circular step.

full rationale

The core method does not reduce to its inputs by construction. User activity is encoded into Digital-DNA sequences, shingled, MinHashed, and classified by LSH-collision majority voting (Equations 1-7), with no learned parameters; the ground-truth labels enter only through the neighbor labels in Equation 7, which is the standard operation of a nearest-neighbor classifier rather than a hidden restatement of the output. The building blocks (Digital-DNA, MinHash, LSH) are external, standard results, and the self-citations in the paper are not load-bearing: Di Paolo et al. 2023 and Allegrini et al. 2024a,b are mentioned only in related work, and De Gaspari et al. appear only as motivation about adversarial attacks. The only circularity-adjacent concern is that Section 5.1.2 reports hyperparameters "found through grid search" without describing a validation split, and Table 2's best alphabet is chosen on the random-split test portion, so some headline F1 values may be optimistically biased. This is a selection/overfitting issue in the evaluation protocol, not a case where the prediction is forced by the construction; it should be weighed as a correctness risk, and a validation-based tuning protocol would be needed to confirm the comparison against baselines.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

The central claim rests on a behavioral-similarity assumption (test bots resemble labeled bots), standard MinHash/LSH theory, and per-dataset hyperparameters selected via grid search. No new physical or conceptual entities are introduced; the B9_temporal alphabet and MD-DNA are encoding schemes, not independent entities.

free parameters (6)
  • k-shingle size = Cresci-15: 2, Cresci-17: 4, Cresci-18: 11, Twibot-22: 4, Fox-8: 7 (Table 7)
    Grid-searched per dataset; controls the granularity of behavior subsequences and directly affects set similarity and classification results.
  • LSH Jaccard threshold = 0.6 (Cresci-15), 0.4 (Cresci-17), 0.4 (Cresci-18), 0.1 (Twibot-22), 0.3 (Fox-8); Table 7
    Grid-searched per dataset; determines which neighbor collisions are counted in the majority-vote classification.
  • Alphabet combination = B5 for Cresci-15, B3 for Cresci-17, B3 for Cresci-18, B5+B9 for Twibot-22, B3+B9 for Fox-8; Table 7
    The best alphabets are selected per dataset from the random-split experiments, a form of model selection performed on the evaluation data.
  • Number of MinHash permutations (signature length) = not reported
    The paper mentions a trade-off between accuracy and speed but never reports the signature length, leaving a key implementation detail unspecified.
  • LSH band/row configuration = not reported
    Only the similarity threshold is given; the mapping from threshold to bands and rows is not stated, which is necessary to reproduce the collision behavior.
  • Temporal alphabet bin boundaries = 1h, 5h, 10h, 15h, 20h, 1 day, 1 week, 1 month (Equation 4)
    Hand-chosen discretization of inter-tweet intervals; no justification or sensitivity analysis is provided for these specific cutoffs.
assumptions (6)
  • domain assumption Bots share recurring sequences of actions that distinguish them from humans (Cresci et al., 2016).
    This is the core premise for local sensitivity and nearest-neighbor matching, invoked throughout Section 3.4 and inherited from prior D-DNA work.
  • standard math MinHash signatures preserve Jaccard similarity between shingle sets.
    Used in Section 3.3 to compress user sets into signatures; standard result from Broder 1997.
  • domain assumption LSH collisions at the chosen threshold identify the most similar users in the ground truth.
    Section 3.4; the classification rule assumes that colliding users are genuinely similar and that majority vote over them is reliable.
  • domain assumption Ground-truth labels in the public datasets (Cresci-15/17/18, Twibot-22, Fox-8) are correct.
    Section 4.1; the evaluation treats all dataset labels as true, including extrapolated labels in Twibot-22 and heuristic identifications in Fox-8.
  • domain assumption A random 70/30 split of a dataset provides a ground truth representative of the test users.
    Section 5.1.1; this is what makes nearest-neighbor matching succeed in the main experiments, while Table 5 shows it fails when this assumption is violated across datasets.
  • ad hoc to paper Fox-8 LLM-powered bot accounts are representative of current state-of-the-art LLM-generated bot content.
    Used to support the abstract claim about SOTA LLM bots; the paper evaluates only a single 2023 botnet and does not generate or test newer LLM-based bots.

how reviews work

0 comments
Cite this review

Pith. "Pith review of BotHash: Efficient and Training-Free Bot Detection Through Approximate Nearest Neighbor." pith.science (2026). https://pith.science/paper/FOGTZ25X

@misc{pith2026250620503,
  author       = {Pith},
  title        = {Pith review of: BotHash: Efficient and Training-Free Bot Detection Through Approximate Nearest Neighbor},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FOGTZ25X}},
  note         = {Machine review of arXiv:2506.20503}
}
read the original abstract

Online Social Networks (OSNs) are a cornerstone in modern society, serving as platforms for diverse content consumption by millions of users each day. However, the challenge of ensuring the accuracy of information shared on these platforms remains significant, especially with the widespread dissemination of disinformation. Social bots -- automated accounts designed to mimic human behavior, frequently spreading misinformation -- represent one of the critical problems of OSNs. The advent of Large Language Models (LLMs) has further complicated bot behaviors, making detection increasingly difficult. This paper presents BotHash, an innovative, training-free approach to social bot detection. BotHash leverages a simplified user representation that enables approximate nearest-neighbor search to detect bots, avoiding the complexities of Deep-Learning model training and large dataset creation. We demonstrate that BotHash effectively differentiates between human and bot accounts, even when state-of-the-art LLMs are employed to generate posts' content. BotHash offers several advantages over existing methods, including its independence from a training phase, robust performance with minimal ground-truth data, and early detection capabilities, showing promising results across various datasets.

Figures

Figures reproduced from arXiv: 2506.20503 by the authors.

Figure 1
Figure 1. High-Level overview of the BotHash classification pipeline. BotHash comprises three main steps: account encoding, signature generation from the [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. BotHash F1 performance across all considered datasets while limiting [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. BotHash F1 performance across all considered datasets when using a [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

51 extracted references · 47 canonical work pages

  1. [1]

    Deciphering Social Behaviour: a Novel Biological Approach For Social Users Classification

    author Allegrini, E. , author Di Paolo, E. , author Petrocchi, M. , author Spognardi, A. , year 2024 a. title Deciphering social behaviour: a novel biological approach for social users classification . journal arXiv preprint arXiv:2412.15410

    work page Pith review arXiv 2024

  2. [2]

    A Proposal for Uncovering Hidden Social Bots via Genetic Similarity

    author Allegrini, E. , author Di Paolo, E. , author Petrocchi, M. , author Spognardi, A. , year 2024 b. title A proposal for uncovering hidden social bots via genetic similarity . journal arXiv preprint arXiv:2410.13512

    work page Pith review arXiv 2024

  3. [3]

    , year 2017

    author Arnaudo, D. , year 2017 . title Computational propaganda in brazil: Social bots during elections

    work page 2017

  4. [4]

    , author Carley, K.M

    author Bellutta, D. , author Carley, K.M. , year 2023 . title Investigating coordinated account creation using burst detection and network analysis . journal Journal of big Data volume 10 , pages 20

    work page 2023

  5. [5]

    , author Ferrara, E

    author Bessi, A. , author Ferrara, E. , year 2016 . title Social bots distort the 2016 us presidential election online discussion . journal First monday volume 21

    work page 2016

  6. [6]

    , year 1997

    author Broder, A.Z. , year 1997 . title On the resemblance and containment of documents , in: booktitle Proceedings. Compression and Complexity of SEQUENCES 1997 (Cat. No. 97TB100171) , organization IEEE . pp. pages 21--29

    work page 1997

  7. [7]

    , year 2000

    author Broder, A.Z. , year 2000 . title Identifying and filtering near-duplicate documents , in: booktitle Annual symposium on combinatorial pattern matching , organization Springer . pp. pages 1--10

    work page 2000

  8. [8]

    , author Kapoor, Y

    author Chawla, V. , author Kapoor, Y. , year 2023 . title A hybrid framework for bot detection on twitter: Fusing digital dna with bert . journal Multimedia Tools and Applications volume 82 , pages 30831--30854

    work page 2023

Show all 51 references
  1. [9]

    , author Di Pietro, R

    author Cresci, S. , author Di Pietro, R. , author Petrocchi, M. , author Spognardi, A. , author Tesconi, M. , year 2015 . title Fame for sale: Efficient detection of fake twitter followers . journal Decision Support Systems volume 80 , pages 56--71

    2015

  2. [10]

    , author Di Pietro, R

    author Cresci, S. , author Di Pietro, R. , author Petrocchi, M. , author Spognardi, A. , author Tesconi, M. , year 2016 . title Dna-inspired online behavioral modeling and its application to spambot detection . journal IEEE Intelligent Systems volume 31 , pages 58--64

    2016

  3. [11]

    , author Di Pietro, R

    author Cresci, S. , author Di Pietro, R. , author Petrocchi, M. , author Spognardi, A. , author Tesconi, M. , year 2017 a. title The paradigm-shift of social spambots: Evidence, theories, and tools for the arms race , in: booktitle Proceedings of the 26th international confere...

    2017

  4. [12]

    , author Di Pietro, R

    author Cresci, S. , author Di Pietro, R. , author Petrocchi, M. , author Spognardi, A. , author Tesconi, M. , year 2017 b. title Social fingerprinting: detection of spambot groups through dna-inspired behavioral modeling . journal IEEE Transactions on Dependable and Secure Com...

    2017

  5. [13]

    , author Lillo, F

    author Cresci, S. , author Lillo, F. , author Regoli, D. , author Tardelli, S. , author Tesconi, M. , year 2018 . title \ fake: Evidence of spam and bot activity in stock microblogs on twitter , in: booktitle Proceedings of the International AAAI Conference on Web and Social Media

    2018

  6. [14]

    , author Lawler, R

    author Davis, W. , author Lawler, R. , year 2023 . title Elon musk blames data scraping by ai startups for his new paywalls on reading tweets . howpublished https://www.theverge.com/2023/7/1/23781198/twitter-daily-reading-limit-elon-musk-verified-paywall

    2023

  7. [15]

    , author Cosuti, L

    author De Faveri, F.L. , author Cosuti, L. , author Tricomi, P.P. , author Conti, M. , year 2023 . title Twitter bots influence on the russo-ukrainian war during the 2022 italian general elections , in: booktitle International Symposium on Security and Privacy in Social Networ...

    2023

  8. [16]

    , author Hitaj, D

    author De Gaspari, F. , author Hitaj, D. , author Mancini, L. , year 2024 . title Have you poisoned my data? defending neural networks against data poisoning , in: booktitle Proceedings of the 29th European Symposium on Research in Computer Security

    2024

  9. [17]

    , author Hitaj, D

    author De Gaspari, F. , author Hitaj, D. , author Pagnotta, G. , author De Carli, L. , author Mancini, L.V. , year 2022 . title Evading behavioral classifiers: a comprehensive analysis on evading ransomware detection techniques . journal Neural Computing and Applications volum...

    2022

  10. [18]

    , author Chang, M

    author Devlin, J. , author Chang, M. , author Lee, K. , author Toutanova, K. , year 2018 . title BERT: pre-training of deep bidirectional transformers for language understanding . journal CoRR volume abs/1810.04805 . http://arxiv.org/abs/1810.04805, http://arxiv.org/abs/1810.0...

    2018 arXiv

  11. [19]

    , author Petrocchi, M

    author Di Paolo, E. , author Petrocchi, M. , author Spognardi, A. , year 2023 . title From online behaviours to images: A novel approach to social bot detection , in: booktitle Computational Science -- ICCS 2023 , publisher Springer Nature Switzerland , address Cham . pp. page...

    2023

  12. [20]

    , author Besel, C

    author Echeverria, J. , author Besel, C. , author Zhou, S. , year 2019 . title Discovery of the twitter bursty botnet , in: booktitle Data Science for Cyber-Security . publisher World Scientific , pp. pages 145--159

    2019

  13. [21]

    , author Payne, S

    author Efthimion, P.G. , author Payne, S. , author Proferes, N. , year 2018 . title Supervised machine learning bot detection techniques to identify social twitter bots . journal SMU Data Science Review volume 1 , pages 5

    2018

  14. [22]

    , author Tan, Z

    author Feng, S. , author Tan, Z. , author Li, R. , author Luo, M. , year 2022 a. title Heterogeneity-aware twitter bot detection with relational graph transformers , in: booktitle Proceedings of the AAAI Conference on Artificial Intelligence , pp. pages 3977--3985

    2022

  15. [23]

    , author Tan, Z

    author Feng, S. , author Tan, Z. , author Wan, H. , author Wang, N. , author Chen, Z. , author Zhang, B. , author Zheng, Q. , author Zhang, W. , author Lei, Z. , author Yang, S. , et al., year 2022 b. title Twibot-22: Towards graph-based twitter bot detection . journal Advance...

    2022

  16. [24]

    , author Wan, H

    author Feng, S. , author Wan, H. , author Wang, N. , author Li, J. , author Luo, M. , year 2021 a. title Twibot-20: A comprehensive twitter bot detection benchmark , in: booktitle Proceedings of the 30th ACM international conference on information & knowledge management , pp. ...

    2021

  17. [25]

    , author Wan, H

    author Feng, S. , author Wan, H. , author Wang, N. , author Luo, M. , year 2021 b. title Botrgcn: Twitter bot detection with relational graph convolutional networks , in: booktitle Proceedings of the 2021 IEEE/ACM International Conference on Advances in Social Networks Analysi...

    2021

  18. [26]

    , author Wan, H

    author Feng, S. , author Wan, H. , author Wang, N. , author Tan, Z. , author Luo, M. , author Tsvetkov, Y. , year 2024 . title What does the bot say? opportunities and risks of large language models in social media bot detection . journal arXiv preprint arXiv:2402.00371

    2024 arXiv

  19. [27]

    , author Nakajima, K

    author Fukuda, M. , author Nakajima, K. , author Shudo, K. , year 2022 . title Estimating the bot population on twitter via random walk based sampling . journal IEEE Access volume 10 , pages 17201--17211

    2022

  20. [28]

    , author Venkatesan, A

    author Gilmary, R. , author Venkatesan, A. , year 2023 . title Entropy-based automation detection on twitter using dna profiling . journal SN Computer Science volume 4 , pages 847

    2023

  21. [29]

    , author Mathew, S

    author Hayawi, K. , author Mathew, S. , author Venugopal, N. , author Masud, M.M. , author Ho, P.H. , year 2022 . title Deeprobot: a hybrid deep neural network model for social bot detection based on user profile data . journal Social Network Analysis and Mining volume 12 , pages 43

    2022

  22. [30]

    , author Motwani, R

    author Indyk, P. , author Motwani, R. , year 1998 . title Approximate nearest neighbors: towards removing the curse of dimensionality , in: booktitle Proceedings of the thirtieth annual ACM symposium on Theory of computing , pp. pages 604--613

    1998

  23. [31]

    , author Hinders, M.K

    author Kirn, S.L. , author Hinders, M.K. , year 2021 . title Bayesian identification of bots using temporal analysis of tweet storms . journal Social Network Analysis and Mining volume 11 , pages 74

    2021

  24. [32]

    , author Ferrara, E

    author Kudugunta, S. , author Ferrara, E. , year 2018 . title Deep neural networks for bot detection . journal Information Sciences volume 467 , pages 312--322

    2018

  25. [33]

    , author Yang, J

    author Li, S. , author Yang, J. , author Zhao, K. , year 2023 . title Are you in a masquerade? exploring the behavior and impact of large language model driven social bots in online social networks . journal arXiv preprint arXiv:2307.10337

    2023 arXiv

  26. [34]

    , author Xu, B

    author Lyu, N. , author Xu, B. , author Guo, F. , author Shen, H. , year 2023 . title Dcgnn: Dual-channel graph neural network for social bot detection , in: booktitle Proceedings of the 32nd ACM International Conference on Information and Knowledge Management , pp. pages 4155--4159

    2023

  27. [35]

    , author Khan, M.U

    author Mohammad, S. , author Khan, M.U. , author Ali, M. , author Liu, L. , author Shardlow, M. , author Nawaz, R. , year 2019 . title Bot detection using a single post on social media , in: booktitle 2019 third world conference on smart trends in systems security and sustaina...

    2019

  28. [36]

    , author Salehi, M

    author Najari, S. , author Salehi, M. , author Farahbakhsh, R. , year 2022 . title Ganbot: A gan-based framework for social bot detection . journal Social Network Analysis and Mining volume 12 , pages 4

    2022

  29. [37]

    , author Hayes, C

    author Pasricha, N. , author Hayes, C. , year 2019 . title Detecting bot behaviour in social media using digital dna compression , in: booktitle 27th AIAI Irish Conference on Artificial Intelligence and Cognitive Science , organization AICS (Artificial Intelligence and Cogniti...

    2019

  30. [38]

    title Musk's twitter sues four texas entities for data scraping, seeks damages

    author Reuters , year 2023 . title Musk's twitter sues four texas entities for data scraping, seeks damages . howpublished https://www.reuters.com/business/media-telecom/musk-reaffirms-tweet-reading-limits-put-due-data-scraping-2023-07-13/

    2023

  31. [39]

    , author Alvarez-Galvez, J

    author Suarez-Lledo, V. , author Alvarez-Galvez, J. , year 2022 . title Assessing the role of social bots during the covid-19 pandemic: Infodemic, disagreement, and criticism . journal J Med Internet Res volume 24 , pages e36085 . https://www.jmir.org/2022/8/e36085, :10.2196/36085

    2022 doi

  32. [40]

    title Twitter just closed the book on academic research

    author theverge.com , year 2023 . title Twitter just closed the book on academic research . howpublished https://www.theverge.com/2023/5/31/23739084/twitter-elon-musk-api-policy-chilling-academic-research . note [Online; accessed 16-04-2024]

    2023

  33. [41]

    , author Martin, L

    author Touvron, H. , author Martin, L. , author Stone, K. , author Albert, P. , author Almahairi, A. , author Babaei, Y. , author Bashlykov, N. , author Batra, S. , author Bhargava, P. , author Bhosale, S. , et al., year 2023 . title Llama 2: Open foundation and fine-tuned cha...

    2023 arXiv

  34. [42]

    , author Harkreader, R

    author Yang, C. , author Harkreader, R. , author Gu, G. , year 2013 . title Empirical evaluation and new design for fighting evolving twitter spammers . journal IEEE Transactions on Information Forensics and Security volume 8 , pages 1280--1293

    2013

  35. [43]

    , author Ferrara, E

    author Yang, K.C. , author Ferrara, E. , author Menczer, F. , year 2022 . title Botometer 101: Social bot practicum for computational social scientists . journal Journal of computational social science volume 5 , pages 1511--1528

    2022

  36. [44]

    , author Hui, P.M

    author Yang, K.C. , author Hui, P.M. , author Menczer, F. , year 2019 . title Bot electioneering volume: Visualizing social bot activity during elections , in: booktitle Companion Proceedings of The 2019 World Wide Web Conference , pp. pages 214--217

    2019

  37. [45]

    , author Menczer, F

    author Yang, K.C. , author Menczer, F. , year 2023 . title Anatomy of an ai-powered malicious social botnet . journal arXiv preprint arXiv:2307.16336

    2023 arXiv

  38. [46]

    , author Varol, O

    author Yang, K.C. , author Varol, O. , author Hui, P.M. , author Menczer, F. , year 2020 . title Scalable and generalizable social bot detection through data selection , in: booktitle Proceedings of the AAAI conference on artificial intelligence , pp. pages 1096--1103

    2020

  39. [47]

    , author Yang, R

    author Yang, Y. , author Yang, R. , author Li, Y. , author Cui, K. , author Yang, Z. , author Wang, Y. , author Xu, J. , author Xie, H. , year 2023 . title Rosgas: Adaptive social bot detection with reinforced self-supervised gnn architecture search . journal ACM Transactions ...

    2023

  40. [48]

    , author Choi, C

    author Yao, H. , author Choi, C. , author Cao, B. , author Lee, Y. , author Koh, P.W.W. , author Finn, C. , year 2022 . title Wild-time: A benchmark of in-the-wild distribution shift over time . journal Advances in Neural Information Processing Systems volume 35 , pages 10309--10324

    2022

  41. [49]

    , author Romero, D

    author Yardi, S. , author Romero, D. , author Schoenebeck, G. , et al., year 2010 . title Detecting spam in a twitter network . journal First monday

    2010

  42. [50]

    , author Song, W

    author Zhang, Y. , author Song, W. , author Shao, J. , author Abbas, M. , author Zhang, J. , author Koura, Y.H. , author Su, Y. , year 2023 . title Social bots’ role in the covid-19 pandemic discussion on twitter . journal International Journal of Environmental Research and Pu...

    2023 doi

  43. [51]

    , author Zhang, D

    author Zhou, M. , author Zhang, D. , author Wang, Y. , author Geng, Y.A. , author Tang, J. , year 2023 . title Detecting social bot on the fly using contrastive learning , in: booktitle Proceedings of the 32nd ACM International Conference on Information and Knowledge Managemen...

    2023

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.