Lockstep protocol Lockstep protocol
a1 = Rock p1 p2 a2 = Scissors
1. Announce a commitment to an action.
commitment can be easily calculated from the action but the action
action
c1 = H(a1) = 4736 c2 = H(a2) = 1832
cannot be inferred from the commitment
formed with a one-
one-way function (e.g., hash)
c2 = 1832 c1 = 4736
2. When everybody has announced their commitments for the
turn, announce the action. a1 = Rock
everybody knows what everybody else has promised to do
3. Verify that the actions correspond to the commitments. a1 = Rock
if not, then somebody is cheating…
a2 = Paper
a2 = Paper
H(a2) = 5383 ≠ c2
Loosening the synchronization 1(2) Loosening the synchronization 2(2)
the slowest player dictates the speed pipelined lockstep protocol
short turns player can send several commitments which are pipelined
time limits for the announcements drawback: look-
look-ahead cheating if a player announces
asynchronous lockstep protocol
action earlier than required
sphere of influence: synchronization is needed only when adaptive pipeline protocol
the players can affect each other in the next turn(s) measure the actual latencies between the players
otherwise, the players can proceed asynchronously grow or shrink the pipeline size accordingly
Drawbacks of the lockstep protocol Idea #1: Let’s get rid of the repeat!
requires two separate message transmissions send only a single message
commitment and action are sent separately but how can we be sure that the opponent cannot learn the action
before annoucing its own action?
slows down the communication
the message is an active object, a delegate
requires a synchronization step
program code to be run by the receiver (host)
the slowest player dictates the pace delegate’s behaviour cannot be worked out by analytical methods alone
~ improvements: asynchronous lockstep, pipelined lockstep, adaptive
adaptive guarantees the message exchange on a possibly hostile environment
environment
pipeline lockstep
delegate provides the action once the host has sent its own
does not solve the inconsistency problem! action using the delegate
1
Example with two players Threats
what if the host delays or prevents the delegate’s message
from getting to its originator?
Ap Ar the host will not receive the next delegate until the message is sent
what if the originator is malicious and the delegate spies or
wastes the host’s resources?
cr(ccpp(a(apapp))) cr(ccpp(a(apapp))) ddpp sandbox: the host restricts the resources available to the delegate
delegate
how can the delegate be sure that it is sending messages to its
D
Drr originator?
communication check-
check-up
D
Dpp
ddrr cp(crc(ar(arar)r)) cp(crc(ar(arar)r))
Communication check-
check-up Idea #2: Peer pressure
the delegate sends a unique players gossip the other players’ actions from the previous
identification to its originator turn(s)
static and dynamic information
compare gossip and recorded actions; if there are
the delegate waits until the originator Ap Ar inconsistencies, ban the player
has responded correctly
cheating is detected only afterwards
check-
check-ups are done randomly D
Drr gossiping imposes a threat of getting caught
probability can be quite low D
Dpp
host cannot know whether the gossip is piggybacked in the ordinary messages
transmission is the actual message or just no extra transmissions are required
a check-
check-up
how to be sure that the gossip is not forged?
rechecking with randomly selected players
How much is enough? Message
example: 10 players, 60 turns, 1 cheater who forges 10% of action for the current turn t
messages, gossip from one previous turn
1% gossip: P(cheater gets caught) = 0.44 delegate for the next turn t + 1
5% gossip: P(cheater gets caught) = 0.91 set of actions (i.e., gossip) from the previous turn t − 1
10% gossip: P(cheater gets caught) = 0.98
example: 100 players, 60 turns, 1 cheater who forges 10% of
messages
1% gossip: P(cheater gets caught) = 0.98
example: 10 players, 360 turns, 1 cheater who forges 10% of
messages
1% gossip: P(cheater gets caught) = 0.97
Dpptt++11 Gpt − 1 aaiitt−−11
mpt aapptt D aajjtt−− 11
2
Collusion Analysing collusion
imperfect information games tracking
infer the hidden information determine who the players are
outwit the opponents but physical identity does not reflect who is actually playing the
the game
collusion = two or more players play together without styling
analyse how the players play the game
informing the other participants
requires a sufficient amount of game data
how to detect collusion in online game? collusion can be detected only afterwards
players can communicate through other media → no pre-
pre-emptive nor real-
real-time counter-
counter-measures
one player can have several avatars
Collusion types Offending other players
active collusion acting against the ‘spirit’ of the game
cheaters play more aggressively than they normally would problematic: is camping in a first-
first-person shooter cheating or just a
good tactic?
can be detected with styling
some rules are ‘gentlemen’s agreements’
passive collusion examples
cheaters play more cautiously than they normally would killing and stealing from inexperiened and ill-
ill-equipped players
practically undetectable gangs and ghettoization of the game world
blocking exits, interfering fights, verbal abuse
Upholding justice Recapitulation: Outline of the course
players handle the policing themselves
theory: players take the law into their own hands (e.g., militia)
militia)
reality: gangs shall inherit the game world 8. Communication layers dead reckoning
systems records misconducts and brands offenders as physical platform local perception filters
criminals logical platform synchronized simulation
theory: bounties and penalties prevent crimes area-
area-of-
of-interest filtering
networked application
reality: throw-
throw-away avatars commit the crimes
players decide whether they can offend/be offended 9. Compensating resourse 10.Cheating
10.Cheating prevention
theory: players know what kind of game world they want limitations technical exploitations
reality: how to offend you? let me count the ways… aspects of compensation rule violations
protocol optimization
3
Examinations 1 (2) Examinations 2 (2)
examination dates questions
1. January 16, 2006 based on both lectures and lecture notes
2. February 13, 2006 two questions, à 5 points
3. March 2, 2006 to pass the examination, at least 5 points (50%) are required
check the exact times and places at grade: g = ⎡p − 5⎤
5⎤
http://www.it.utu.fi/opetus/tentit/ questions are in English, but you can answer in English or in Finnish
Finnish
if you are not a student of University of Turku, you must remember to enrol in time!
register to receive the credits
further instructions are available at
http://http://www.tucs.fi/education/
courses/participating_courses.php
