Anatomy of a Scam Call: What 10,000 real scam and spam calls reveal about how phone scammers operate

Anatomy of a vishing call: A theoretically grounded review of mobile money fraud and AI voice cloning in Ghana

Mobile money has become the financial backbone of Ghana, with transaction values reaching several trillion cedis annually and active accounts numbering in the tens of millions. This scale has attracted a parallel surge in fraud, much of it perpetrated through voice phishing (vishing): unsolicited telephone calls in which fraudsters impersonate telecommunications providers to manipulate subscribers into surrendering credentials or authorizing transfers. This paper reviews the state of vishing-enabled mobile money fraud in Ghana, synthesizing regulator data, industry reporting, documented case typologies, and the academic literature on social engineering and mobile money security. It organizes the dominant attack scripts - security-verification, loyalty-reward, misdirected-payment, and SIM-swap pretexts - into a common four-stage pattern of impersonation, pretext, manipulation, and extraction, and interprets this pattern through routine activity theory and persuasion theory. The paper then examines an escalating threat: the maturation of AI voice cloning, which can now replicate a target voice from seconds of audio, and which controlled studies show human listeners cannot reliably detect. Because Ghanaian vishing already succeeds using ordinary human voices, the paper argues that voice cloning constitutes a force multiplier that will erode voice familiarity, the last perceptual defense available to targets, and expand the pretext space from institutional to personal impersonation. The review contributes a consolidated threat taxonomy, a theoretically grounded attack model, and a research and policy agenda for providers, regulators, and awareness programs in Ghana and comparable mobile-money-dependent economies.

TENET: Telegram Mini App (in)security

Telegram, with over 450 million daily active users, has introduced Mini Apps---web-based applications running directly within its client. However, this integration introduces notable security risks. As we demonstrate, many Mini Apps store authentication materials---such as session tokens and wallet mnemonic phrases---in plaintext on client devices, exposing users to unauthorized access, impersonation, and financial exploitation. While insecure client-side storage is a known risk in web applications, the Telegram Mini App ecosystem presents a uniquely dangerous combination of factors absent from prior work: no platform-level security review, no storage access restrictions, a financially motivated user base handling live cryptocurrency assets, and a WebView environment that offers weaker protections than standalone browsers. To investigate this threat, we present TENET, a purpose-built auditing tool whose design decisions---pattern selection, entropy thresholds, and charset validation---are grounded in the structural properties of the secrets targeted and empirically validated against a ground-truth dataset. We screened 61 Mini Apps using a stratified, popularity-weighted sampling strategy based on popularity. Of the 37 applications that met our processing criteria and were analyzed, 30 exhibited security flaws, which we classify into three severity tiers: plaintext storage, recoverable encryption, and replayable tokens. Notably, even Telegram's official Wallet exhibits a severe vulnerability that may lead to full account compromise. Following our responsible disclosure, Telegram implemented two new secure-storage APIs, and our post-remediation verification confirmed that its official Wallet no longer exposes the recovery mnemonic in plaintext. Finally, we propose mitigation measures and best practices for both Telegram platform developers and third-party Mini App creators.

CallScreenBench: Benchmarking On-Device Models as Phone Secretaries

THE STRANGER ON THE VIDEO CALL: TRUTH-DEFAULT, TRANSPARENCY AND INDIA’S DIGITAL ARREST SCAMS

On 28 August 2024, the eighty-two-year-old chairman of one of India’s largest textile groups answered a telephone call that kept him under fake police custody, on camera, for two days and cost him seven crore rupees. This article reads India’s digital arrest epidemic through two connected bodies of work: Timothy Levine’s truth-default theory and Malcolm Gladwell’s popularisation of it in Talking to Strangers (2019). It argues that these frauds succeed not because their victims are foolish but because the scam weaponises three ordinary features of human communication: our default to belief, our faith in appearances, and the coupling of behaviour to engineered circumstance. It ends with what the psychology of deception cannot explain, which is the entirely rational unease Indian citizens feel when the state appears to come calling.

CallScreenBench: Benchmarking Small Language Models as Phone Secretaries

Language models small enough to run on a handset, quantized to a few bits, are increasingly capable of acting on their user's behalf -- which makes on-device task automation newly plausible. One such task is answering the phone. A phone secretary takes an unknown inbound call on its owner's behalf, and unlike the agents most benchmarks evaluate, it has no cooperative caller-assigned task to complete: the caller holds the goal and may be an adversary, while the secretary must begin deciding how to respond without an oracle. What matters is not task success but whether the owner would endorse how their proxy handled the call. We evaluate only the text-domain conversational decision layer; speech recognition, audio interaction, end-to-end latency, and handset execution are outside scope. We present CallScreenBench, which reports five automated call-and-note measure groups motivated by owner endorsement. Each is paired, where available, with a counter-metric and an uncertainty estimate; no benchmark-wide Q1-Q5 composite or leaderboard score is defined. Three guardedness diagnostics identify candidate cases for a toolless proxy that holds no credentials and calls no tools. Across three model families represented by paired 4-bit checkpoints (0.6-4B), the primary scoring snapshot gives the larger checkpoint higher point estimates on several service, recall, and plausibility measures, while triage discrimination follows a different ordering. Bare scam-side TPR rewards universal suspicion, and pairwise separation changes when legitimate-side false positives are included and across judge snapshots. Scripted degenerate agents expose further floors, including a hangup-and-echo policy with entity recall 1.000. We report quality measures and guardedness channels separately so that a single pass/fail score does not hide their trade-offs.

Profiling the Invisible Insider: A UEBA-Based Machine Learning Framework for Low-and-Slow Data Exfiltration Detection

Low-and-slow data exfiltration by malicious insiders remains among the most operationally difficult threat classes to detect: the behavior is unremarkable in any individual session and becomes recognizable only across weeks of otherwise routine activity. This paper presents a UEBA-based machine learning framework that constructs per-user behavioral profiles from enterprise proxy and access log data, scoring sessions against a 30-feature behavioral representation spanning temporal patterns, data-transfer anomalies, domain interactions, HTTP characteristics, and session-device signals. The contribution is an operationally integrated and empirically audited UEBA pipeline that combines a pre-specified behavioral representation, isolated evaluation regimes, and session-level analyst explanations. Training used a hybrid corpus of approximately 8.96 million sessions drawn from 160 GB of real and behaviorally parameterized simulated logs. On the full hybrid held-out partition, the LightGBM classifier achieved 96.84% overall accuracy, 95.38% balanced accuracy, 91.80% malicious-class precision, 92.90% recall, 92.35% F1-score, and 98.2% ROC-AUC. Because the hybrid test set uses an enriched 20.5% malicious-session evaluation prevalence, balanced accuracy and malicious-class F1 are emphasized alongside overall accuracy. In the strict real-only evaluation, the independently trained and calibrated LightGBM model achieved 95.99% overall accuracy, 93.80% balanced accuracy, 90.30% malicious-class precision, 90.10% recall, 90.20% F1-score, and 95.8% ROC-AUC, while hybrid-to-real transfer achieved 90.1% F1-score, 89.8% recall, and 96.1% ROC-AUC. These settings are reported separately to distinguish full-corpus benchmark performance from real-log generalization. Each scored session is accompanied by a SHAP-based decomposition that identifies which behavioral signals drove the alert, supporting analyst triage in operational SOC environments. The framework was further validated on a confirmed 17-day insider exfiltration incident that existing organizational controls had not detected. BiLSTM and a Tabular Transformer evaluated under the same regime-specific partitions produced closely matched performance, with malicious-class F1 differences of at most 0.4 percentage points; LightGBM retained the strongest measured performance–deployment trade-off for the engineered tabular pipeline.