Exploiting MMORPG Log Data toward Efficient RMT Player...

1
0 5 10 15 20 25 29 0 200 400 600 800 1000 Number of identified RMT Players Top N RMT Suspects TCH TAC TCC AT TCH/TAC TCH/TCC TCH/AT 10 1 10 2 0 3 6 9 12 15 18 21 24 Total actions counts / Activity time (log scale) Average activity hours per day Seller Collector Earner Other 10 3 10 4 10 5 10 2 10 3 10 4 10 5 10 6 Total currency handled (log scale) Log count (log scale) Seller Collector Earner Other RMT Real currency Property in virtual world Buyer Earner Collector Seller Exploiting MMORPG Log Data toward Efficient RMT Player Detection Hiroshi Itsuki, Asuka Takeuchi, Atsushi Fujita, Hitoshi Matsubara Solution Future University Hakodate, JAPAN Issues of Operating MMORPG Table 1: Mean Values of Statistics Type Seller Collector Earner Others 10 4 15 971 n 254.4 745.5 106,137.8 75,008.8 TAC 55.8 138.3 16,772.5 3,718.0 AT 1.3 46.8 11.0 951.0 TCC 12,209.2 23,247.0 13,438.7 3,111.0 TCH Table 2: The Number of Extracted RMT Suspects Seller 10 639 998 1,000 990 72 226 81 Collector 4 452 889 1,000 891 133 347 123 Earner 15 80 360 1,000 48 419 349 643 29 639 998 1,000 990 419 349 643 All Statistics n TCH TAC TCC AT TCH/TAC TCH/TCC TCH/AT Currency Items Currency Items Preliminary Investigation RMT players deal with a huge volume of virtual currency Most of the RMT players are silent Earners spend long time at a low frequency of action Sellers and collectors take actions less frequently MMORPG: Multi-player online role-playing game Countermeasure against unfavarable behavior RMT: Real-money trading Harassment (PK, robbing) and it’s automation Manual investigation requires human effort and time Effective use of huge volume of log data Computes “suspecthood” using statistics derived from log data Provided data Log data (Action log, Chat log) List of manually identified RMT players Target Commercial MMORPG “Uncharted Waters Online” TECMO KOEI GAMES CO., LTD. http://global.netmarble.com/uwo/ Target period 2009/8/30~2009/9/13 (in which the operators had identified RMT players) Acknowledgemnt Our greatest thanks to TECMO KOEI GAMES CO., LTD. Derived four types of statistics from both types of data TCC: The total number of utterances recorded TAC: The total number of action records AT: The amount of minutes in which at least one action is taken TCH: The amount of virtual currency handled in the period (The absolute values of currency increase and decrease) Automatic detection of RMT players using log data Ranking all players with regard to their “suspiciousness” Manual verification starting from rank 1 Summary 1. Statistical Analysis 2. RMT Player Detection Problems of RMT in MMORPGs Imbalances virtual economy Direct attacks to general players Unauthorized behavior Prevents new players Discourages existing players RMT player network 0. Assumption RMT players handle huge amount of currency 15,250 players top 1,000 players based on TCH Significant differences between RMT players and others Different statistics achieved the best results for each type of RMT players Earners (15) no effect sellers (10) collectors (4) Future Work Explore a more effective statistics especially from short-term log data Apply to other game titles

Transcript of Exploiting MMORPG Log Data toward Efficient RMT Player...

Page 1: Exploiting MMORPG Log Data toward Efficient RMT Player ...paraphrasing.org/~fujita/publications/coauthor/... · Exploiting MMORPG Log Data toward Efficient RMT Player Detection Hiroshi

0

5

10

15

20

25

29

0 200 400 600 800 1000

Num

ber o

f ide

nti�

ed R

MT

Play

ers

Top N RMT Suspects

TCHTACTCCATTCH/TACTCH/TCCTCH/AT

101

102

0 3 6 9 12 15 18 21 24

Tota

l act

ions

cou

nts

/ Act

ivity

tim

e (lo

g sc

ale)

Average activity hours per day

SellerCollectorEarnerOther

103

104

105

102 103 104 105 106

Tota

l cur

renc

y ha

ndle

d (lo

g sc

ale)

Log count (log scale)

SellerCollectorEarnerOther

RMT

Real currency

Property in virtual world

Buyer

EarnerCollectorSeller

Exploiting MMORPG Log Data toward Efficient RMT Player DetectionHiroshi Itsuki, Asuka Takeuchi, Atsushi Fujita, Hitoshi Matsubara

Solution

Future University Hakodate, JAPAN

Issues of Operating MMORPG

Table 1: Mean Values of StatisticsTypeSellerCollectorEarnerOthers

104

15971

n254.4745.5

106,137.875,008.8

TAC55.8

138.316,772.5

3,718.0

AT1.3

46.811.0

951.0

TCC12,209.223,247.013,438.7

3,111.0

TCH

Table 2: The Number of Extracted RMT SuspectsSeller

10639998

1,000990

72226

81

Collector4

452889

1,000891133347123

Earner1580

3601,000

48419349643

29639998

1,000990419349643

AllStatisticsnTCHTACTCCATTCH/TACTCH/TCCTCH/AT

CurrencyItems

CurrencyItems

Preliminary Investigation

RMT players deal with a huge volume of virtual currency

Most of the RMT players are silent

Earners spend long time at a low frequency of action

Sellers and collectors take actions less frequently

MMORPG: Multi-player online role-playing gameCountermeasure against unfavarable behavior RMT: Real-money trading Harassment (PK, robbing)and it’s automation Manual investigation requires human effort and time Effective use of huge volume of log data

Computes “suspecthood” using statistics derived from log dataProvided data Log data (Action log, Chat log) List of manually identified RMT players

Target Commercial MMORPG “Uncharted Waters Online” TECMO KOEI GAMES CO., LTD. http://global.netmarble.com/uwo/Target period 2009/8/30~2009/9/13 (in which the operators had identified RMT players)

AcknowledgemntOur greatest thanks to TECMO KOEI GAMES CO., LTD.

Derived four types of statistics from both types of data TCC: The total number of utterances recorded TAC: The total number of action records AT: The amount of minutes in which at least one action is taken TCH: The amount of virtual currency handled in the period (The absolute values of currency increase and decrease)

Automatic detection of RMT players using log data Ranking all players with regard to their “suspiciousness” Manual verification starting from rank 1

Summary

1. Statistical Analysis

2. RMT Player Detection

Problems of RMT in MMORPGs Imbalances virtual economy Direct attacks to general players Unauthorized behavior ↓ Prevents new players Discourages existing players

RMT player network

0. Assumption RMT players handle huge amount of currency 15,250 players ⇒ top 1,000 players based on TCH

Significant differences between RMT players and others

Different statistics achieved the best results for each type of RMT players

Earners (15)

no effect

sellers (10)collectors (4)

Future WorkExplore a more effective statistics especially from short-term log dataApply to other game titles