# Pump-and-Dump meets Honeypot Tokens: Detection and Analysis of Telegram Bait-and-Trap Schemes

Crypto.world

> Prior research on Pump-and-Dump schemes has focused almost entirely on centralized exchanges, leaving decentralized exchanges largely unexplored. The authors monitor Telegram channels that advertise…

**arXiv**: 2610.10149

**Published**: 2026-10-07

**Authors**: Federico Cernera, Massimo La Morgia, Alessandro Mei, Francesco Sassi

## AI summary

## Overview

**Research area:** Blockchain security and measurement — decentralized finance (DeFi) fraud, market manipulation, and Telegram-coordinated cryptocurrency schemes.

**Technical level:** Intermediate. The paper is readable without deep smart-contract expertise, but familiarity with tokens, liquidity pools, and automated market makers helps.

**Scope:** A large-scale measurement study of 83 Telegram channels and 3,677 advertised Pump-and-Dump events on BNB Smart Chain and Ethereum, showing that most are actually "Bait-and-Trap" honeypot schemes and introducing a simulation tool (HoneyCheck) to detect them.

## What This Paper Is About

Prior research on Pump-and-Dump schemes has focused almost entirely on centralized exchanges, leaving decentralized exchanges largely unexplored. The authors monitor Telegram channels that advertise Pump-and-Dump operations on DEXs and find that the tokens being promoted usually contain hidden code that lets victims buy but blocks them from selling, while a small set of insider addresses trade freely. The paper names this fraud "Bait-and-Trap," quantifies how much money it extracts, and builds a tool that lets a user test whether a token can actually be sold before committing funds.

## Key Contributions

1. **Discovery of Bait-and-Trap as a distinct fraud.** By jointly analyzing Telegram coordination messages and on-chain execution across 3,677 events, the authors show that 71.5% of operations advertised as Pump-and-Dump are actually honeypot-based traps, making Bait-and-Trap the dominant phenomenon in the dataset.
2. **Characterization of the scheme's lifecycle and economics.** The paper documents how targeted tokens are created, how colluding addresses generate fake trading activity, how organizers conceal evidence by deleting messages, and how profits from one operation finance later ones.
3. **HoneyCheck, an open-source detection tool.** HoneyCheck emulates a buy followed by an attempted sell against historical or current blockchain state, classifying tokens as honeypots based on transaction reversions and economically prohibitive fees.
4. **Public datasets.** The paper releases, to the authors' knowledge, the largest dataset of announced Pump-and-Dump operations on DEXs: 3,677 events, the source code of 1,377 associated token contracts, and real-world code snippets implementing the honeypot mechanisms uncovered.

## Main Findings

- **Honeypots dominate the dataset:** HoneyCheck classifies 2,629 of 3,677 operations (71.5%) as Bait-and-Trap, involving 1,492 distinct tokens (70.3% of all targeted tokens). The remaining 1,048 operations (28.5%), covering 629 tokens (29.7%), show no honeypot behavior under the tested conditions.
- **Selling is rare compared with buying:** Across operation windows there are 528,866 swap transactions from 171,755 unique addresses, totaling $74.1M in buy volume and $61.2M in sell volume. 159,831 addresses execute at least one buy, but only 73,693 execute at least one sell. On BSC, 35.8% of operations involve five or fewer sellers versus 7.9% with five or fewer buyers; on Ethereum the figures are 45.5% and 9.1%.
- **Colluding addresses create the illusion of a real market:** The small set of addresses allowed to trade freely makes up only 10.6% of addresses participating in Bait-and-Trap operations, yet accounts for more than 72% of total buy volume and 92% of total sell volume.
- **Organizers almost always profit:** Token creators and colluding addresses realize positive returns in 99.3% of analyzed cases, extracting more than $7 million in aggregate.
- **Detection mechanisms break down into five categories:** Of the Bait-and-Trap operations, 1,902 (72.3%) are sell reversions, 273 (10.4%) timed sell reversions, 120 (4.6%) approval reversions, 288 (11.0%) high buy-side fees above 90%, and 46 (1.7%) high sell-side fees above 90%. At zero delay HoneyCheck detects 1,095 tokens (983 sell reverts, 112 approval reverts); adding delays of 1, 5, and 10 seconds adds 62, 187, and 191 tokens cumulatively, for 1,286 detected via reversion. Fee-based criteria add 206 tokens (180 buy-side, 26 sell-side).
- **Targets are disposable, freshly created tokens:** Bait-and-Trap tokens have a median market lifetime of one day, with 79.6% living less than 72 hours and 91.8% less than one week. Pump-and-Dump targets have a median lifetime of 211 days. 91.1% of Bait-and-Trap targets are deployed within 48 hours of the corresponding operation.
- **The fraud mechanism shifted over time:** 94.5% of Pump-and-Dump operations occurred before June 2022; from mid-2022 onward, Bait-and-Trap dominates. Channels running Bait-and-Trap are also far more prolific, with the most active channel carrying out 296 operations.
- **Operations are financially linked:** Fund-flow tracing (limited to two hops and stopped at known services) identified 70 money-flow clusters containing 1,661 operations, roughly 63% of all Bait-and-Trap operations, with the top four clusters each containing more than 100 connected operations.
- **Detection generalizes to external data:** On the 1,859-token honeypot dataset from Huynh et al., evaluated at the state right after initial liquidity provision, 90.2% (1,676 tokens) met the honeypot criteria. On 967 tokens from PancakeSwap's official token list, evaluated in July 2026, all 967 were correctly classified as non-honeypots.

## Methodology in Plain English

The authors first find the fraudsters. They search the Telegram Desktop client and the channel index TGStat.com using generic terms like "pump" and "crypto pump" between November 2024 and January 2025, then follow forwards and mentions to reach channels that avoid obvious keywords. From 217 candidate channels, they keep the 83 that organized at least one DEX operation. Two researchers manually read the messages, which follow near-identical templates, and extract the start time, token, liquidity pool, exchange, and blockchain for each operation.

Because many channels delete the token address shortly after an operation, the authors reconstruct missing details from blockchain explorers, DEX pair-lookup functions, and a Web3 platform that maps token symbols to addresses. They then run dedicated nodes for BSC and Ethereum to pull every event from each target pool — 2,121 pool creations, 166,847 liquidity mints, 30,054 burns, and 4,746,720 swaps — and define the end of each operation as the point where trading volume falls back to its pre-announcement baseline.

To test whether a token was a trap, they built HoneyCheck. It uses a local development node to fork the blockchain at the exact block where the operation was announced, funds a fresh account, calculates how many tokens a buy should yield, executes the buy, then attempts to approve and sell. It records whether the approval or the sell reverts, and it compares expected versus actual amounts to estimate fees. Crucially, it can advance time before selling, which catches contracts that permit selling only within the same block or shortly after purchase.

Finally, they trace where the money goes from token creators, grouping operations that share funding sources into clusters.

## Why This Matters

**Impact on research:** The paper reframes what researchers thought they were measuring. Studies of Telegram "pump" channels on DEXs may largely have been observing honeypot fraud rather than price manipulation, which affects how trading data from those channels should be interpreted. It also provides a labeled-scale dataset and a reusable detection method.

**Real-world applications:**

- **Wallet and DEX front-ends** could run a pre-trade simulation and warn users before they buy a token they cannot sell.
- **Retail investors** can use an open-source tool to check sellability in seconds instead of relying on channel administrators' charts.
- **Blockchain analytics and compliance teams** can use fund-flow clustering to link apparently unrelated operations to common controllers.
- **Exchange and token-list curators** can screen submitted tokens with behavioral simulation rather than source-code review alone.

**Industry relevance:** The finding that organizers profit in 99.3% of cases and extract more than $7 million makes this a measurable, economically significant fraud pattern. Because the scheme relies on a token contract behaving differently for insiders than for ordinary users, it is directly relevant to token-listing policies, anti-fraud tooling, and user-protection design across EVM-compatible chains.

## Future Directions

- **Explaining missed detections:** The paper notes that 183 labeled honeypot tokens from the external dataset were not detected, and analyzes them in a later section; improving coverage of these cases is an open problem.
- **Hardening against anti-simulation defenses:** HoneyCheck already varies delay (0, 1, 5, 10, 30 seconds, 1 and 10 minutes) to defeat contracts that allow selling only immediately after purchase, but contracts could evolve further countermeasures.
- **Generalizing beyond BSC and Ethereum:** The dataset concentrates heavily on BSC (3,649 of 3,677 operations), and the authors discuss this concentration later in the paper; extending monitoring to other EVM chains and to non-EVM ecosystems is a natural next step.
- **Attribution and disruption:** The money-flow clustering links roughly 63% of Bait-and-Trap operations into 70 clusters, raising the question of how such cluster-level evidence could support takedowns or platform-level intervention rather than per-token warnings.

## Target Audience

Security and measurement researchers working on blockchain and DeFi; blockchain analytics and compliance engineers; developers of wallets, DEX interfaces, and token-listing pipelines; and technically literate crypto users or journalists who want to understand how Telegram "pump" channels actually operate. Readers new to smart contracts will need to pick up basic concepts (tokens, liquidity pools, buy/sell swaps), but the paper's structure and case study make the core argument accessible.

## Authors’ abstract

While Pump-and-Dump schemes have been extensively studied on centralized exchanges (CEXs), how they operate in decentralized exchanges (DEXs) remains largely unexplored. We monitor 83 Telegram channels used to coordinate Pump-and-Dump campaigns and collect 3,677 events across the BNB Smart Chain and Ethereum. Our analysis reveals that, although these operations superficially resemble CEX-based Pump-and-Dump schemes, their underlying mechanism is fundamentally different.…

[The full abstract on arXiv](https://arxiv.org/abs/2610.10149)

[Read the original paper](https://arxiv.org/abs/2610.10149)

---

Canonical: https://crypto.world/paper/pump-and-dump-meets-honeypot-tokens-detection-and-analysis-of-telegram-2610.10149
