Are cryptocurrencies the future of financial fraud?
Bennett Kleinberg
Department of Security and Crime Science Dawes Centre for Future Crime University College London
Anatomy of a P&D
The three phases of a pump-and-dump (Kamps & Kleinberg, 2018)
2 things happened
- The Internet
- Cryptocurrencies
What could go wrong?
The recent explosion of nearly 2,000 cryptocurrencies in a largely unregulated environment has greatly expanded the scope for abuse.
The story of OfficialMcAfee and OfficiallMcAfee.
Coin of the day …
When the tweet was first broadcast at around 3 p.m. ET, GVT was bought and sold on the market at $30.
When the tweet was first broadcast at around 3 p.m. ET, GVT was bought and sold on the market at $30.
By 3:04, it was at $45, and trading volume had doubled.
When the tweet was first broadcast at around 3 p.m. ET, GVT was bought and sold on the market at $30.
By 3:04, it was at $45, and trading volume had doubled.
But by 3:19, GVT’s price had fallen back to $30.
Morale of the story
The smart money — the early money — had gotten out, leaving the late money holding a bag of now-diminished value.
Crypto P&Ds
| Old challenges | New challenges |
|---|---|
| low market cap | faster time scale (near real-time) |
| lack of reliable information | broader spread of misinformation |
| low price | new vehicles for “rumours” |
Yes, yes, yes… but what do I care?
“It’s just another problem”
A problem with \($\) involved
- $50m (Onecoin)
- $825m over six months
- $222m for a single group
The “grey zone” problem
Can we detect P&Ds?
Data
- scraped cryptocurrency exchange data
- trading data of 1-h granularity
- 20 days of trading, 5 exchanges, 977 trading pairs
Data + code available at https://osf.io/827wd/
Criteria for P&Ds
- sudden price increase (PUMP)
- sudden volume increase (PUMP)
- marked price drop (DUMP)
Conditional local point anomaly detection
Anomalie parameters
| Balanced parameters | |
|---|---|
| Estimation window | 12h |
| Volume increase | 300% |
| Price increase | 5% |
| Price drop | 1.00 SD |
| Alleged pumps | 2150 |
| Pump-and-dumps | 1617 |
| Crypto/crypto | 97.0% |
| Low market cap | 81.8% |
P&D detection
Findings
- potential for detectability
- mainly crypto/crypto pairs
- mostly low market cap
Intermezzo: situational crime prevention
- away from the criminological-sociological model
- away from the “individual-centric” model
- fraud/crime as a complex problem
- … embedded in a situation
- robust effect: clustering
What does this mean for cryptocurrency?
- understanding the P&D situation
- Does clustering occur?
Zooming in
- Exchange-level
- % of traded coins \(\neq\) % of P&Ds
- Some exchanges are used more often than others
- e.g. “Kraken”: 6% of traded coins, < 1% of P&Ds
- among most regulated marketplaces
Zooming in
- Coin-level
- Most are never targeted
- Some are targeted again and again
- 30% of coins \(\sim\) 80% of P&Ds
Repeat victimisation of coins
Core findings
- potential for detectability
- mainly crypto/crypto pairs
- mostly low market cap
- repeat exchange victimisation
- repeat coin victimisation
- no evidence for “Bitcoin is not immune from the pump-and-dump phenomenon” (Hamrick et al., 2018)
In context
- P&D groups successful in pumping the price
- (very) short timeframe
- Telegram and Discord used for organising and coordination
see Kamps & Kleinberg, 2018; Xu & Livshits, 2018; Hamrick et al., 2018; Li et al., 2018
Maybe there’s more to the coins…
High-potential coins and coin malleability
Arguably, the decision to choose one coin for a pump-and-dump scheme and not another are not random but rather subject to cost-benefit processes.
[…] if one is targeting a coin for a P&D, it is useful to be able to manipulate a coin’s price as much as possible, while using the least amount of capital
Coin malleability
“the percentage price increase that a coin would experience, with respect to a certain amount of volume injected”
- data
- 239 x/BTC pairs
- scraped in April 2019
- levels of BTC injection
- 0.25 BTC (~ $1,300)
- 1.00 BTC (~ $5,100)
- 5.00 BTC (~ $25,500)
- 10.00 BTC (~ $51,000)
Key findings on coin malleability
Not all coins are equal!
- some coins are more malleable than others
- trading volume highly clustered
- malleability not identical to low volume
- possible explanation: too low volume doesn’t do the trick!
Vision on research & bigger picture
Old-school approach
- detection
- prevention
Circular problem.
Ongoing research
- Detection:
- (Weak) signals in the accumulation phase
- (Weak) signals in the pump phase
- Anticipation:
- Changes in MO
- Transition to new targets (coins)
Interdisciplinary challenge!
New problems
- Twitter impersonations
- New means to spread information
- Transactions in milliseconds
New questions
- trust
- how do groups maintain internal trust?
- how do they punish?
New questions
- trust
- decision-making processes
- how are targets chosen?
- what is the cost-benefit process?
- what do we not see?
Situational prevention
- does displacement happen?
- how?
- where to?
Vision on research
- collaboration between universities, fintech, government, banks, “citizen science”
Vision on research
(unrealistic but cool)
- infiltration of P&D groups
- own P&D
Are cryptocurrencies the future of financial fraud?
Yes Maybe Probably No
What do we not know?
What are the challenges for the future?
10th + 11th June, Crypto-fraud “sandpit”
End.
Credits to the team: Josh Kamps, Florian Hetzel.
