Pith. sign in

REVIEW 2 major objections 1 minor 113 references

Robocalls affect every continent but occur at far higher rates and cause more damage in the United States than elsewhere.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.3

2026-07-01 04:26 UTC pith:IVGIFTG6

load-bearing objection New 65-country robocall dataset is the real output here, but the US-severity claim rests on unnormalized aggregates that the abstract does not defend. the 2 major comments →

arxiv 2606.31790 v1 pith:IVGIFTG6 submitted 2026-06-30 cs.CR

Robocalls: A Worldwide or US-only Problem? Analyzing Spam and Fraud in International Phone Calls

Kemal Altwlkany , Andro Mer\'cep , Tomislav {\DJ}uri\v{c}i\'c , Ante Kapetanovi\'c , Emanuel Lacic This is my paper
classification cs.CR
keywords robocallsspam callsphone fraudinternational callscall detail recordscampaign clusteringmultimodal dataset
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper gathers robocall reports and call records across 65 countries on all inhabited continents to test whether the problem is US-specific or global. It releases the first public multimodal dataset of 8.7 million call detail records plus transcripts and audio from 28 campaign clusters collected over nine months. Comparative analysis of calling patterns, co-targeting attacks, callback numbers, and language use shows robocalls operate internationally yet reach substantially greater volume and reported losses inside the US. The work supplies concrete steps for future research and remedies aimed at reducing robocall effectiveness.

Core claim

Although robocalls constitute an international problem, the severity of the threat is significantly higher in the US than in other countries.

What carries the argument

A multimodal international robocall dataset containing 8.7 million call detail records, 839 transcripts, and 677 recordings from 28 identified campaign clusters, used to compare patterns across 65 countries.

Load-bearing premise

The collected robocall reports and records from 65 countries are representative enough to support cross-country comparisons despite differences in legal systems, reporting mechanisms, and data availability.

What would settle it

Uniform measurement of robocall volume per capita in a new set of countries showing rates comparable to the United States would contradict the finding of markedly higher US severity.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Mitigation resources can be allocated most efficiently by treating the US as the primary target while still addressing shared international campaigns.
  • Common robocall campaigns identified across borders enable coordinated blocking of callback numbers.
  • Linguistic adaptations within the same language across regions indicate scammers tailor messages to local audiences.
  • Public release of the dataset supports development of detection tools that work beyond single-country data.

Where Pith is reading between the lines

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

  • The same data-collection approach could be applied to other scam vectors such as SMS or messaging apps to test whether US concentration holds there as well.
  • Differences in severity may trace to phone penetration rates or enforcement differences, suggesting targeted regulatory comparisons between the US and lower-severity nations.
  • Callback-number extraction methods described could feed into real-time reputation systems shared across carriers worldwide.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 1 minor

Summary. The manuscript aggregates robocall reports and call records from 65 countries, releases the first public multimodal international robocall dataset (8.7 million CDRs, 839 transcripts from 28 clusters, 677 recordings), describes 9-month collection methodology, and performs comparative analysis of calling patterns, co-targeting, campaigns, callback numbers, and linguistic features, concluding that robocalls are an international problem but with significantly higher severity in the US.

Significance. The public dataset release is a clear strength that can support reproducible follow-on work. If the severity comparison is shown to be robust after normalization, the findings would usefully inform US-centric and cross-border anti-robocall measures.

major comments (2)
  1. [Methodology and Comparative Analysis] Methodology and Comparative Analysis sections: the claim that severity is 'significantly higher in the US' is based on aggregated reports from 65 countries but supplies no description of normalization by population, telephone subscriptions, or reporting propensity; without these denominators the US excess cannot be distinguished from differences in data availability and incentives.
  2. [Results] Results section: no error bars, sampling methodology, or bias-correction steps are reported for the cross-country incidence metrics, leaving the statistical support for the headline severity claim unverifiable from the presented data.
minor comments (1)
  1. [Abstract] Abstract: the $1.1 billion loss figure is dated 'during 2025'; clarify whether this is a projection or a typographical reference to the collection window.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on the comparative analysis. We agree that the severity claim requires explicit normalization and statistical details to be robust. We will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [Methodology and Comparative Analysis] Methodology and Comparative Analysis sections: the claim that severity is 'significantly higher in the US' is based on aggregated reports from 65 countries but supplies no description of normalization by population, telephone subscriptions, or reporting propensity; without these denominators the US excess cannot be distinguished from differences in data availability and incentives.

    Authors: We agree this is a substantive gap. The current manuscript presents aggregated counts without normalization. In revision we will add per-country incidence rates normalized by population and telephone subscriptions (sourced from ITU and World Bank data). We will also include a limitations subsection discussing reporting propensity differences and how our 9-month collection methodology (honeypots and report aggregation) attempts to address them where possible. revision: yes

  2. Referee: [Results] Results section: no error bars, sampling methodology, or bias-correction steps are reported for the cross-country incidence metrics, leaving the statistical support for the headline severity claim unverifiable from the presented data.

    Authors: We acknowledge the absence of these elements. The revision will add error bars to the incidence metrics, explicitly describe the sampling approach (including the 9-month collection window, data sources for the 8.7M CDRs, and how the 65-country reports were obtained), and outline bias-correction steps or their limitations based on the available data. revision: yes

Circularity Check

0 steps flagged

No circularity: purely observational data aggregation and comparison

full rationale

The paper performs empirical collection of robocall reports and call records across 65 countries, followed by descriptive analysis and cross-country comparison. No equations, fitted parameters, predictions derived from inputs, or self-citations are invoked as load-bearing steps in the central claim. The severity comparison rests directly on the collected dataset rather than any self-referential definition or reduction. This is a standard observational study whose validity hinges on data quality and normalization (addressed by the skeptic), not on circular derivation.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The paper is an empirical measurement study; no free parameters, invented entities, or non-standard axioms are described in the abstract.

axioms (1)
  • domain assumption Robocall reports can be aggregated across countries despite legal and privacy constraints
    The study relies on this premise to compile data from 65 countries.

pith-pipeline@v0.9.1-grok · 5876 in / 1228 out tokens · 56619 ms · 2026-07-01T04:26:32.040827+00:00 · methodology

0 comments
read the original abstract

Unsolicited automated phone calls (robocalls) are a serious threat: in the US alone, these calls resulted in reported losses of 1.1$ billion during 2025. Phishing and spoofing consistently rank among the most reported crimes within the FBI's Internet Crime Complaint Center, with phone call scams having the highest reported median loss. Combating robocalls is difficult due to many legal and practical constraints: robocalls often encompass multiple legal jurisdictions of different countries/states, the large volume of robocalls, their multilingual nature, the lack of publicly available data, privacy concerns with obtaining data, etc. We present a study of international robocalls, aggregating robocall reports from countries across all inhabited continents and contribute by providing new findings on international robocalls from 65 different countries. We also present the first publicly available multimodal and international robocall dataset: 8.7 million call detail records, 839 robocall transcripts from 28 identified robocall campaign clusters, and 677 robocall recordings. We describe our methodology for collecting robocall data over a 9-month period and provide a detailed analysis comparing robocalls in the US with those in other countries. Our analysis covers several aspects, including uncovering calling patterns, identifying co-targeting attacks, discovering common robocall campaigns, extracting callback numbers, analyzing linguistic differences among robocalls in the same language but different regions, and other insights. Our results indicate that although robocalls are an international problem, the severity of the threat is significantly higher in the US than in other countries. We provide steps for future research and suggest remedies to reduce the effectiveness of robocalls based on our analysis.

Figures

Figures reproduced from arXiv: 2606.31790 by Andro Mer\'cep, Ante Kapetanovi\'c, Emanuel Lacic, Kemal Altwlkany, Tomislav {\DJ}uri\v{c}i\'c.

Figure 1
Figure 1. Figure 1: The average number of monthly robocalls in the US over 2025 was 2.56 billion, which is the highest since 2019. to US consumers, with the volume of scam and telemarketing calls increasing by 15.6% in 2025 compared to 2024 [2]. 1.1. What exactly are robocalls? Definitions vary. According to the US Federal Communica￾tions Commission (FCC): “robocalls are calls made with an autodialer or that contain a prereco… view at source ↗
Figure 2
Figure 2. Figure 2: The Netherlands. NOS, part of the Dutch public broad￾casting system, released an article in which they state that the Dutch Fraud Help Desk received nearly 10,000 reports of scam calls that try to defraud people and extort money [58]. Scam￾mers are relying on AI-enabled technology to facilitate such calls. The news portal Telecoms reported that O2, the largest tele￾com in the United Kingdom, detected 150 m… view at source ↗
Figure 2
Figure 2. Figure 2: Info flyer published by the Croatian Police as part of Europol’s campaign relating to phone theft and fraud. If the callers present themselves as an organization the police in￾structs citizens to not call back numbers that the caller might have specified during the call. Citizens are instead advised to look up official phone numbers of those organizations. This info flyer was automatically translated using… view at source ↗
Figure 3
Figure 3. Figure 3: provides a simplified diagram of two different hon￾eypot configurations. We used the same honeypot consisting of the same phone numbers, but in two different configurations (setups/modes). The first setup is termed passive, which corresponds to Fig￾ure 3a. In this scenario, any incoming calls are simply re￾jected by the honeypot, and the honeypot does not interact with the calls. However, the honeypot regi… view at source ↗
Figure 4
Figure 4. Figure 4: Daily volume of phone calls registered by the honeypot. Days colored in blue represent calls registered while the honeypot was operating in passive mode, i.e. the honeypot does not interact with the calls, it simply rejects them. Days colored in orange indicate the interfering mode, i.e. the honeypot answers calls and plays warning messages to inform callers about the recording process. We noticed a patter… view at source ↗
Figure 5
Figure 5. Figure 5: Temporal distribution of robocalls by hour of day and day of week for US (left) and international (right) callees. Timestamps are in callee local time. Color intensity indicates the percentage of total calls in each region [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Distribution of calls per caller ID across the full dataset (US and international data combined) on log-log scale. The heavy-tailed pattern shows that most callers make few calls, while a small number of prolific callers generate substantial traffic. For US callees, robocall activity concentrates during busi￾ness hours. Calls to US numbers peak at 10:00 AM local time (9.7% of daily volume), with secondary … view at source ↗
Figure 7
Figure 7. Figure 7: Most frequent non-domestic caller country by destination country. The United States is the dominant external source for 16 destinations, followed by the United Kingdom for 5 destinations. dominant source for Denmark (3,553 of 13,253) and Sweden (47 of 126), and India is the top caller for Germany (72 of 73). Overall, the figure shows that international call traffic is not evenly distributed across countrie… view at source ↗
Figure 8
Figure 8. Figure 8: Word clouds of caller countries by US state, with word size proportional to the number of non-domestic calls. Nigeria is the dominant external source in 35 states, followed by Canada and Mexico, each leading in 4 states. 101 102 103 Co-targeting group size 10−4 10−3 10−2 10−1 100 CCDF US International [PITH_FULL_IMAGE:figures/full_fig_p011_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Complementary CDF of temporal co-targeting group sizes for US and international callees. Both distributions are heavy-tailed: a large majority of groups contain only a few caller IDs, while a small number of groups exceed 102 mem￾bers. Groups of size 1 are excluded. use a multilingual model, because the transcript set spans multi￾ple languages and we compare US and international calls in the same analysis.… view at source ↗
Figure 11
Figure 11. Figure 11: Duration of recorded US and international phone calls. The distributions for both regions follow a similar pat￾tern, with most calls being shorter than 20 seconds. 0 20 40 60 80 100 120 Vocal activity [%] 0 5000 10000 15000 20000 Number of calls US International [PITH_FULL_IMAGE:figures/full_fig_p012_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Percentage of vocal activity detected. In both re￾gions, a substantial amount of robocalls are silent, with very little vocal activity. This is a common phenomenon among robo￾calls and has been previously observed [4, 20]. “Blank” calls are usually made with the goal of determining whether the di￾aled number has voice capabilities (scouting calls). using voice activity detection features that trigger play… view at source ↗
Figure 13
Figure 13. Figure 13: Hallucinations caused by transformer-based automatic speech recognition can heavily impact the transcriptions. Hallu￾cinations are more prominent in recordings that contain a lower ratio of speech compared to silence. To the left, we show the most common words occurring in low speech ratio recordings, while the image on the right shows the most frequently occurring words in high speech ratio recordings. T… view at source ↗
Figure 14
Figure 14. Figure 14: Word clouds for English and Spanish, generated separately for US recordings and international recordings. Although robocalls were made in the same language, the word clouds differ, indicating that language is not the only predictor of robocall content - the region matters as well. 6.2.3. Regional Differences Between The Same Language We want to inspect whether there are differences between the most common… view at source ↗
Figure 15
Figure 15. Figure 15: Chord diagram showing the flow of languages towards callee countries. Country names are in ISO 3166-1-alpha-2 codes. 6.3.1. Campaign Discovery via Transcript Clustering Similar to [21], we apply Density-Based Spatial Clustering of Applications with Noise (DBSCAN) [100] to identify robocall campaigns based on their transcripts. Text normalization. Each transcript undergoes a text nor￾malization process in … view at source ↗
Figure 16
Figure 16. Figure 16: Robocall categorization per region (left - US, right - international). Similar to the results obtained with robocall cluster￾ization (Tables 4 and 5), scams related to financial services are more common in international robocalls, while tech scams are more common in US robocalls. Additionally, US robocalls contain a larger percentage of legitimate automated notifications. transcripts u and v is equal to t… view at source ↗

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

113 extracted references · 113 canonical work pages · 1 internal anchor

  1. [1]

    The record- ings are anonymized and assigned a unique, randomly gener- ated ID

    No metadata about recordings will be published. The record- ings are anonymized and assigned a unique, randomly gener- ated ID. They cannot be traced back to any caller ID or caller country

    work page

  2. [2]

    The number of robocall recordings that we make publicly avail- able (677) is significantly smaller than the number of CDRs (more than 8.7 million)

    Any published recording must be verified by a human. The number of robocall recordings that we make publicly avail- able (677) is significantly smaller than the number of CDRs (more than 8.7 million). If a recording is included in our dataset, it is confirmed by at least one human evaluator to be a robocall prior to publishing

    work page

  3. [3]

    All other (unverified) recordings were filtered in a process we describe later to minimize the chance that they are accidental Table 1:Quantity of phone numbers employed in the honeypot. Although the total amount of available phone numbers varied per day, this variance manifested steadily and on average resulted in a change of 2.27% for US phone numbers a...

    work page

  4. [4]

    Finally, any recordings shorter than 5 seconds were automat- ically deleted to further ensure that accidental misdials are excluded in our research. Overall, we approached the sensitive topic of collecting robocall data with extreme caution and have taken several steps to ensure no violation of caller privacy and to minimize the chance of accidental/live ...

    work page

  5. [5]

    A key consideration to keep in mind is caller ID spoofing

    Call Metadata Analysis This section presents the analysis of robocall metadata collected from our honeypot infrastructure. A key consideration to keep in mind is caller ID spoofing. Robocallers routinely falsify the displayed phone number, so a call may appear to originate from the callee’s country, to increase the level of trust. Because caller metadata ...

    work page 2025

  6. [6]

    thank you

    Call Content Analysis In this section, we analyze the audio content of the recorded calls. In total, we have recorded 220,301 calls in the US, and 42,798 calls in the rest of the world. Due to the aforementioned legal reasons, calls shorter than 5 seconds were discarded, after which 197,765 (89.77%) US calls were left and 33,253 (77.7%) international call...

    work page

  7. [7]

    The median number of phone num- bers in our US honeypot was 11,751, compared to 101,913 in- ternational phone numbers

    Discussion Even though robocalls are an international problem, our find- ings indicate that US citizens are substantially more affected than the rest of the world. The median number of phone num- bers in our US honeypot was 11,751, compared to 101,913 in- ternational phone numbers. Despite an almost tenfold differ- ence in the number of available phone li...

    work page

  8. [8]

    Conclusion In this paper, we presented a detailed overview of the state of international robocalls. Our honeypot registered 9.6 mil- lion calls, of which we make publicly available a dataset of 8.7 million call detail records, 677 robocall recordings, and 839 clustered and categorized robocall transcripts. Our anal- ysis showed that although robocalls are...

    work page

  9. [9]

    Acknowledgments This research was supported in part by the project Infobip Global Communication Platform (PK.1.1.07.0001), part of the Important Project of Common European Interest on Next Gen- eration Cloud Infrastructure and Services (IPCEI-CIS) consor- tium. The authors also want to thank Patricio Marcos Petri´c for his assistance during the writing of...

    work page

  10. [10]

    Ringing in our fears 2025: Robocalls hit 6-year high,

    U.S. Public Interest Research Group Education Fund, “Ringing in our fears 2025: Robocalls hit 6-year high,” https://pirg.org/e dfund/resources/ringing-in-our-fears-2025-robocalls-hit-6-yea r-high/, 2025, [Online] Accessed: 2025-12-11

    work page 2025

  11. [11]

    Phone scams flourish as robocalls increase by 16%,

    U.S. Public Interest Research Group Education Fund, “Phone scams flourish as robocalls increase by 16%,” https://pirg.org/e dfund/media-center/phone-scams-flourish-as-robocalls-incre ase-by-16-consumer-protection-week-2026/, 2026, [Online] Accessed: 2026-04-30

    work page 2026

  12. [12]

    Stop unwanted robo- calls and texts,

    Federal Communications Commission, “Stop unwanted robo- calls and texts,” https://www.fcc.gov/consumers/guides/sto p- unwanted- robocalls- and- texts, 2025, [Online] Accessed: 2026-05-02

    work page 2025

  13. [13]

    Who’s calling? characterizing robocalls through audio and metadata analysis,

    S. Prasad, E. Bouma-Sims, A. K. Mylappan, and B. Reaves, “Who’s calling? characterizing robocalls through audio and metadata analysis,” in29th USENIX Security Symposium (USENIX Security 20), 2020, pp. 397–414

    work page 2020

  14. [14]

    Detection of robocall and spam calls using acoustic features of incoming voicemails,

    B. Elizalde and D. Emmanouilidou, “Detection of robocall and spam calls using acoustic features of incoming voicemails,” in Proceedings of Meetings on Acoustics, vol. 45, no. 1. AIP Pub- lishing, 2021

    work page 2021

  15. [15]

    Fraud reports,

    Federal Trade Commission Public Tableau, “Fraud reports,” ht tps://public.tableau.com/app/profile/federal.trade.commis sion/viz/FraudReports/FraudFacts, 2025, [Online] Accessed: 2025-12-11

    work page 2025

  16. [16]

    Facts and figures 2024,

    International Telecommunications Union Telecommunication Development Sector, “Facts and figures 2024,” https://www.it u.int/itu-d/reports/statistics/facts-figures-2024/, 2024, [Online] Accessed: 2025-12-11

    work page 2024

  17. [17]

    Facts and figures 2025,

    International Telecommunications Union Telecommunication Development Sector, “Facts and figures 2025,” https://www.it u.int/itu-d/reports/statistics/facts-figures-2025/, 2026, [Online] Accessed: 2026-05-02

    work page 2025

  18. [18]

    Recent advances in speech language models: A survey,

    W. Cui, D. Yu, X. Jiao, Z. Meng, G. Zhang, Q. Wang, S. Y . Guo, and I. King, “Recent advances in speech language models: A survey,” inProceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), W. Che, J. Nabende, E. Shutova, and M. T. Pilehvar, Eds. Vienna, Austria: Association for Computational Lingui...

    work page 2025

  19. [19]

    A survey on large language models for code generation,

    J. Jiang, F. Wang, J. Shen, S. Kim, and S. Kim, “A survey on large language models for code generation,”ACM Transactions on Software Engineering and Methodology, vol. 35, no. 2, pp. 1–72, 2026

    work page 2026

  20. [20]

    Robocall Audio from the FTC’s Project Point of No Entry,

    Sathvik Prasad and Bradley Reaves, “Robocall Audio from the FTC’s Project Point of No Entry,” North Carolina State Univer- sity, Tech. Rep. TR-2023-1, Nov 2023

    work page 2023

  21. [22]

    Introduction to CCITT Sig- nalling System No. 7,

    International Telecommunication Union Telecommunication Standardization Sector (ITU-T), “Introduction to CCITT Sig- nalling System No. 7,” https://www.itu.int/rec/T-REC-Q.700, 1988, [Online] Accessed: 2025-12-11

    work page 1988

  22. [23]

    SIP: Session Initiation Protocol,

    E. Schooler, J. Rosenberg, H. Schulzrinne, A. Johnston, G. Camarillo, J. Peterson, R. Sparks, and M. J. Handley, “SIP: Session Initiation Protocol,” RFC 3261, Jul. 2002. [Online]. Available: https://www.rfc-editor.org/info/rfc3261

    work page 2002

  23. [24]

    Secure Telephone Identity Problem Statement and Requirements,

    J. Peterson, H. Schulzrinne, and H. Tschofenig, “Secure Telephone Identity Problem Statement and Requirements,” RFC 7340, Sep. 2014. [Online]. Available: https://www.rfc-editor.or g/info/rfc7340

    work page 2014

  24. [25]

    Secure Telephone Identity Credentials: Certificates,

    J. Peterson and S. Turner, “Secure Telephone Identity Credentials: Certificates,” RFC 8226, Feb. 2018. [Online]. Available: https://www.rfc-editor.org/info/rfc8226

    work page 2018

  25. [26]

    Secure Telephone Identity Revisited (STIR) Out-of-Band Architecture and Use Cases,

    E. Rescorla and J. Peterson, “Secure Telephone Identity Revisited (STIR) Out-of-Band Architecture and Use Cases,” RFC 8816, Feb. 2021. [Online]. Available: https://www.rfc-edi tor.org/info/rfc8816

    work page 2021

  26. [27]

    Personal Assertion Token (PaSSporT) Extension for Signature-based Handling of Asserted information using toKENs (SHAKEN),

    C. Wendt and M. Barnes, “Personal Assertion Token (PaSSporT) Extension for Signature-based Handling of Asserted information using toKENs (SHAKEN),” RFC 8588, May 2019. [Online]. Available: https://www.rfc-editor.org/info/rfc8588

    work page 2019

  27. [28]

    Compliance and enforcement and telecom decision crtc 2021-123,

    Canadian Radio-television and Telecommunications Commis- sion, “Compliance and enforcement and telecom decision crtc 2021-123,” https://crtc.gc.ca/eng/archive/2021/2021-123.htm, 2021, [Online] Accessed: 2025-12-11

    work page 2021

  28. [29]

    Diving into robo- call content with{SnorCall},

    S. Prasad, T. Dunlap, A. Ross, and B. Reaves, “Diving into robo- call content with{SnorCall},” in32nd USENIX Security Sympo- sium (USENIX Security 23), 2023, pp. 427–444

    work page 2023

  29. [30]

    Characterizing robo- calls with multiple vantage points,

    S. Prasad, A. Nahapetyan, and B. Reaves, “Characterizing robo- calls with multiple vantage points,” in2025 IEEE Symposium on Security and Privacy (SP). IEEE, 2025, pp. 19–36

    work page 2025

  30. [31]

    A conformer-based classifier for variable-length utterance process- ing in anti-spoofing,

    E. Rosello, A. Gomez-Alanis, A. M. Gomez, and A. Peinado, “A conformer-based classifier for variable-length utterance process- ing in anti-spoofing,” inInterspeech 2023, 2023, pp. 5281–5285

    work page 2023

  31. [32]

    Spoofed Speech Detection with a Focus on Speaker Embedding,

    H. M. Tran, D. Guennec, P. Martin, A. Sini, D. Lolive, A. Del- hay, and P.-F. Marteau, “Spoofed Speech Detection with a Focus on Speaker Embedding,” inInterspeech 2024, 2024, pp. 2080– 2084

    work page 2024

  32. [33]

    Can Emotion Fool Anti-spoofing?

    A. Mahapatra, I. R. Ulgen, A. Reddy Naini, C. Busso, and B. Sis- man, “Can Emotion Fool Anti-spoofing?” inInterspeech 2025, 2025, pp. 5628–5632

    work page 2025

  33. [34]

    Generalizable Audio Spoofing Detection using Non-Semantic Representations,

    A. Das, Y . El Kheir, C. Franzreb, T. Herzig, T. Polzehl, and S. Möller, “Generalizable Audio Spoofing Detection using Non-Semantic Representations,” inInterspeech 2025, 2025, pp. 4553–4557

    work page 2025

  34. [35]

    Real-time korean voice phishing detection based on machine learning approaches,

    M. Lee and E. Park, “Real-time korean voice phishing detection based on machine learning approaches,”Journal of Ambient In- telligence and Humanized Computing, vol. 14, no. 7, pp. 8173– 8184, 2023

    work page 2023

  35. [36]

    Automatically detecting voice phishing: A large audio model approach,

    B. M. Ampel, S. Samtani, and H. Chen, “Automatically detecting voice phishing: A large audio model approach,”MIS Quarterly, pp. 1–21, 2026

    work page 2026

  36. [37]

    New ftc data show consumers re- ported losing nearly $8.8 billion to scams in 2022,

    Federal Trade Commission, “New ftc data show consumers re- ported losing nearly $8.8 billion to scams in 2022,” https: //www.ftc.gov/news-events/news/press-releases/2023/02/ne w-ftc-data-show-consumers-reported-losing-nearly-88-billion -scams-2022, 2023, [Online] Accessed: 2025-12-11

    work page 2022

  37. [38]

    Internet crime report 2024,

    Federal Bureau of Investigation IC3 Division, “Internet crime report 2024,” https://www.ic3.gov/AnnualReport/Reports/202 4_IC3Report.pdf, 2025, [Online] Accessed: 2025-12-11

    work page 2024

  38. [39]

    Internet crime report 2025,

    Federal Bureau of Investigation IC3 Division, “Internet crime report 2025,” https://www.ic3.gov/AnnualReport/Reports/202 5_IC3Report.pdf, 2026, [Online] Accessed: 2026-05-02

    work page 2025

  39. [40]

    National do not call registry,

    Federal Trade Commission Public Tableau, “National do not call registry,” https://public.tableau.com/app/profile/federal.trade.co mmission/viz/DoNotCall/Infographic, 2025, [Online] Accessed: 2025-12-11

    work page 2025

  40. [41]

    2024 report of the federal communications com- mission consumer advisory committee,

    Federal Communications Commission Consumer Advisory Committee, “2024 report of the federal communications com- mission consumer advisory committee,” https://www.fcc.gov/si tes/default/files/CAC_Final_Report_September2024.pdf, 2024, [Online] Accessed: 2025-12-11

    work page 2024

  41. [42]

    Africa’s cybercrime crisis: A deep dive into GASA’s 2025 state of scams in Africa report,

    Centre for Multilateral Affairs, “Africa’s cybercrime crisis: A deep dive into GASA’s 2025 state of scams in Africa report,” https://thecfma.org/africas-cybercrime-crisis-a-deep-dive-int o-gasas-2025-state-of-scams-in-africa-report/, 2025, [Online] Accessed: 2026-05-02

    work page 2025

  42. [43]

    Understanding the computer misuse and cyber- crimes (amendment) act, 2025,

    Vellum Kenya, “Understanding the computer misuse and cyber- crimes (amendment) act, 2025,” https://vellum.co.ke/understan ding-the-computer-misuse-and-cybercrimes-amendment-act-2 025/, 2025, [Online] Accessed: 2026-05-02

    work page 2025

  43. [44]

    More than 300 arrests as African countries clamp down on cyber threats,

    INTERPOL, “More than 300 arrests as African countries clamp down on cyber threats,” https://www.interpol.int/en/News-and -Events/News/2025/More-than-300-arrests-as-African-count ries-clamp-down-on-cyber-threats, 2025, [Online] Accessed: 2026-05-02

    work page 2025

  44. [45]

    Telecoms fraud costing SA $300M annually – COMRiC,

    Connecting Africa, “Telecoms fraud costing SA $300M annually – COMRiC,” https://www.connectingafrica.com/cybersecurity /telecoms-fraud-costing-sa-300m-annually-comric, 2025, [Online] Accessed: 2026-05-02

    work page 2025

  45. [46]

    EFCC arrests 105 suspected internet fraudsters in Abuja,

    Economic and Financial Crimes Commission, “EFCC arrests 105 suspected internet fraudsters in Abuja,” https://www.efcc .gov.ng/news/efcc-arrests-four-chinese-101-others-for-suspect ed-internet-fraud-in-abuja, 2025, [Online] Accessed: 2026-05- 02

    work page 2025

  46. [47]

    NPA: Fraud cases at record for first half of 2025,

    Tokyo Reporter, “NPA: Fraud cases at record for first half of 2025,” https://www.tokyoreporter.com/japan-news/npa-fra ud-cases-at-record-for-first-half-of-2025/, 2025, [Online] Accessed: 2026-05-02

    work page 2025

  47. [48]

    Tokyo police app blocks incoming interna- tional calls to help combat fraud,

    The Japan Times, “Tokyo police app blocks incoming interna- tional calls to help combat fraud,” https://www.japantimes.co.jp /news/2025/12/01/japan/tokyo-police-app-fraud-block/, 2025, [Online] Accessed: 2026-05-02

    work page 2025

  48. [49]

    Fraud soars to record high in Japan as scam tactics grow more sophisticated,

    The Japan Times, “Fraud soars to record high in Japan as scam tactics grow more sophisticated,” https://www.japantimes.co.jp /news/2026/02/12/japan/crime-legal/japan-2025-crime-figures /, 2026, [Online] Accessed: 2026-05-02

    work page 2026

  49. [50]

    V oice phishing losses hit 5-year high as per-case damage doubles,

    Seoul Economic Daily, “V oice phishing losses hit 5-year high as per-case damage doubles,” https://en.sedaily.com/finance/2026/ 03/24/voice-phishing-losses-hit-5-year-high-as-per-case-dam age, 2026, [Online] Accessed: 2026-05-02

    work page 2026

  50. [51]

    South Korea’s biggest telco blocked over a billion calls and SMS messages in 2025,

    Commsrisk, “South Korea’s biggest telco blocked over a billion calls and SMS messages in 2025,” https://commsrisk.com/sout h-koreas-biggest-telco-blocked-over-1-billion-calls-and-sms -messages-in-2025/, 2025, [Online] Accessed: 2026-05-02

    work page 2025

  51. [52]

    INTERPOL financial crime operation makes record 5,500 arrests, seizures worth over USD 400 million,

    INTERPOL, “INTERPOL financial crime operation makes record 5,500 arrests, seizures worth over USD 400 million,” https://www.interpol.int/en/News-and-Events/News/2024/ INTERPOL-financial-crime-operation-makes-record-5-500-a rrests-seizures-worth-over-USD-400-million, 2024, [Online] Accessed: 2026-05-02

    work page 2024

  52. [53]

    Ministry releases details of successful crackdown on telecom and online fraud,

    China Daily, “Ministry releases details of successful crackdown on telecom and online fraud,” https://www.chinadaily.com.c n/a/202401/05/WS6597ad76a3105f21a507acc4.html, 2024, [Online] Accessed: 2026-05-02

    work page 2024

  53. [54]

    Prosecutors to inten- sify fight against telecom fraud,

    Supreme People’s Procuratorate of China, “Prosecutors to inten- sify fight against telecom fraud,” https://en.spp.gov.cn/2025-02/ 14/c_1070430.htm, 2025, [Online] Accessed: 2026-05-02

    work page 2025

  54. [55]

    China adopts rival caller identification standard to STIR/SHAKEN,

    Commsrisk, “China adopts rival caller identification standard to STIR/SHAKEN,” https://commsrisk.com/china-adopts-rival -caller-identification-standard-to-stir-shaken/, 2024, [Online] Accessed: 2026-05-02

    work page 2024

  55. [56]

    Scam calls up by 78% in 3rd quarter of 2025,

    ABS-CBN News, “Scam calls up by 78% in 3rd quarter of 2025,” https://www.abs-cbn.com/news/business/2025/10/1 4/scam-calls-up-by-78-in-3rd-quarter-of-2025-1841, 2025, [Online] Accessed: 2026-06-02

    work page 2025

  56. [57]

    Telecom spam and scams in the 5g and artificial intelligence era: analyzing eco- nomic implications, technical challenges and global regulatory efforts,

    M. Pietri, M. Mamei, and M. Colajanni, “Telecom spam and scams in the 5g and artificial intelligence era: analyzing eco- nomic implications, technical challenges and global regulatory efforts,”International Journal of Information Security, vol. 24, no. 3, p. 139, 2025

    work page 2025

  57. [58]

    Singapore and malaysia sign mou to strengthen cooperation to tackle scams across telecommunications chan- nels,

    Indiplomacy, “Singapore and malaysia sign mou to strengthen cooperation to tackle scams across telecommunications chan- nels,” https://indiplomacy.com/2024/02/27/singapore-and -malaysia-sign-mou-to-strengthen-cooperation-to-tackle-s cams-across-telecommunications-channels/, 2026, [Online] Accessed: 2026-05-02

    work page 2024

  58. [59]

    Impersonation scheme tar- geting Middle Eastern students in the United States,

    Federal Bureau of Investigation, “Impersonation scheme tar- geting Middle Eastern students in the United States,” https: //www.ic3.gov/PSA/2025/PSA250513, 2025, [Online] Ac- cessed: 2026-05-02

    work page 2025

  59. [60]

    e& and mavenir roll out some ai-powered voice security,

    Andrew Wooden, “e& and mavenir roll out some ai-powered voice security,” https://www.telecoms.com/ai/e-and-maven ir-roll-out-some-ai-powered-voice-security, 2025, [Online] Accessed: 2025-12-11

    work page 2025

  60. [61]

    Phone bank fraud warning issued by Saudi data agency,

    Arab News, “Phone bank fraud warning issued by Saudi data agency,” https://www.arabnews.com/node/2047176/saudi-arabi a, 2022, [Online] Accessed: 2026-05-02

    work page arXiv 2022

  61. [62]

    Desde este miércoles las llamadas comerciales y masivas deberán llevar los prefijos 600 y 809 para su identificación,

    Subsecretaría de Telecomunicaciones de Chile, “Desde este miércoles las llamadas comerciales y masivas deberán llevar los prefijos 600 y 809 para su identificación,” https://www.subtel.g ob.cl/desde-este-miercoles-las-llamadas-comerciales-y-masiv as-deberan-llevar-los-prefijos-600-y-809-para-su-identificacio n/, 2025, [Online] Accessed: 2026-05-02

    work page 2025

  62. [63]

    Chile implements mandatory 600 and 809 pre- fixes for all commercial calls to combat spam,

    IDT Express, “Chile implements mandatory 600 and 809 pre- fixes for all commercial calls to combat spam,” https://www.id texpress.com/news/chile-implements-mandatory-600-and-8 09-prefixes-for-all-commercial-calls-to-combat-spam/, 2025, [Online] Accessed: 2026-05-02

    work page 2025

  63. [64]

    Brazilians lead the world with 28 spam calls per month,

    Hiya, “Brazilians lead the world with 28 spam calls per month,” https://blog.hiya.com/brazilians-lead-the-world-with-28-spa m-calls-per-month/, 2025, [Online] Accessed: 2026-05-02

    work page 2025

  64. [65]

    The machine era of spam calls: The ten most spammed countries in the world,

    Truecaller, “The machine era of spam calls: The ten most spammed countries in the world,” https://www.truecaller.com/b log/insights/the-ten-most-spammed-countries-in-the-world, 2026, [Online] Accessed: 2026-05-02

    work page 2026

  65. [66]

    Internet fraud,

    Republic of Croatia, Ministry of the Interior, “Internet fraud,” https://policija.gov.hr/prevencija/racunalna-sigurnost/internet-p rijevare/456, [Online] Accessed: 2025-12-11

    work page 2025

  66. [67]

    Scammers are gratefully using ai tools: ’now even a lay- man can do it’,

    NOS, “Scammers are gratefully using ai tools: ’now even a lay- man can do it’,” https://nos.nl/artikel/2563317-oplichters-mak en-dankbaar-gebruik-van-ai-tools-een-leek-kan-het-nu-ook, 2025, [Online] Accessed: 2025-12-11

    work page arXiv 2025

  67. [68]

    O2 flags 50 million nuisance calls a month,

    Nick Wood, “O2 flags 50 million nuisance calls a month,” https: //www.telecoms.com/security/o2-flags-50-million-nuisance-cal ls-a-month, 2025, [Online] Accessed: 2025-12-11

    work page 2025

  68. [69]

    Combating fraud: collaboration, consumer educa- tion and law enforcement,

    Telefonica, “Combating fraud: collaboration, consumer educa- tion and law enforcement,” https://www.telefonica.com/en/co mmunication-room/blog/combating-fraud-collaboration-con sumer-education-law-enforcement/, 2026, [Online] Accessed: 2026-05-02

    work page 2026

  69. [70]

    Harassing phone calls, 7.5 million blocked every day with new agcom filte,

    Il sole 24 Ore, “Harassing phone calls, 7.5 million blocked every day with new agcom filte,” https://en.ilsole24ore.com/art/haras sing-phone-calls-every-day-the-new-agcom-filter-blocked-7 5-million-AHFmS8vD, 2025, [Online] Accessed: 2026-05-02

    work page 2025

  70. [71]

    Sending out an sms: Characterizing the security of the sms ecosystem with public gateways,

    B. Reaves, N. Scaife, D. Tian, L. Blue, P. Traynor, and K. R. Butler, “Sending out an sms: Characterizing the security of the sms ecosystem with public gateways,” in2016 IEEE Symposium on Security and Privacy (SP). IEEE, 2016, pp. 339–356

    work page 2016

  71. [72]

    {AuthLoop}:{End-to-End} cryptographic authentication for telephony over voice channels,

    B. Reaves, L. Blue, and P. Traynor, “{AuthLoop}:{End-to-End} cryptographic authentication for telephony over voice channels,” in25th USENIX Security Symposium (USENIX Security 16), 2016, pp. 963–978

    work page 2016

  72. [73]

    {AuthentiCall}: Efficient identity and content authentication for phone calls,

    B. Reaves, L. Blue, H. Abdullah, L. Vargas, P. Traynor, and T. Shrimpton, “{AuthentiCall}: Efficient identity and content authentication for phone calls,” in26th USENIX Security Sym- posium (USENIX Security 17), 2017, pp. 575–592

    work page 2017

  73. [74]

    On sms phishing tactics and in- frastructure,

    A. Nahapetyan, S. Prasad, K. Childs, A. Oest, Y . Ladwig, A. Kapravelos, and B. Reaves, “On sms phishing tactics and in- frastructure,” in2024 IEEE Symposium on Security and Privacy (SP). IEEE, 2024, pp. 1–16

    work page 2024

  74. [75]

    Jäger: Automated telephone call traceback,

    D. Adei, V . Madathil, S. Prasad, B. Reaves, and A. Scafuro, “Jäger: Automated telephone call traceback,” inProceedings of the 2024 on ACM SIGSAC Conference on Computer and Com- munications Security, 2024, pp. 2042–2056

    work page 2024

  75. [76]

    Sok: Everyone hates robocalls: A survey of techniques against telephone spam,

    H. Tu, A. Doupé, Z. Zhao, and G.-J. Ahn, “Sok: Everyone hates robocalls: A survey of techniques against telephone spam,” in 2016 IEEE Symposium on Security and Privacy (SP). IEEE, 2016, pp. 320–338

    work page 2016

  76. [77]

    A virtual honeypot framework

    N. Provoset al., “A virtual honeypot framework.” inUSENIX Security Symposium, vol. 173, no. 2004, 2004, pp. 1–14

    work page 2004

  77. [78]

    The honeynet project: Trapping the hackers,

    L. Spitzner, “The honeynet project: Trapping the hackers,”IEEE Security & Privacy, vol. 1, no. 2, pp. 15–23, 2003

    work page 2003

  78. [79]

    Phoneypot: Data-driven understanding of telephony threats

    P. Gupta, B. Srinivasan, V . Balasubramaniyan, and M. Ahamad, “Phoneypot: Data-driven understanding of telephony threats.” in NDSS, vol. 107, 2015, p. 108

    work page 2015

  79. [80]

    V oip honeypot architecture,

    M. Nassar, R. State, and O. Festor, “V oip honeypot architecture,” in2007 10th IFIP/IEEE International Symposium on Integrated Network Management, 2007, pp. 109–118

    work page 2007

  80. [81]

    Greystar: Fast and accurate detection of{SMS}spam numbers in large cellu- lar networks using gray phone space,

    N. Jiang, Y . Jin, A. Skudlark, and Z.-L. Zhang, “Greystar: Fast and accurate detection of{SMS}spam numbers in large cellu- lar networks using gray phone space,” in22nd USENIX Security Symposium (USENIX Security 13), 2013, pp. 1–16

    work page 2013

Showing first 80 references.