Challenges for Providing Processing Integrity in Grid Computing - - PDF document
Challenges for Providing Processing Integrity in Grid Computing Felipe Martins 1 , Mrcio Maia 1 , Rossana M. de C. Andrade 2 Aldri L. dos Santos 2 , Jos Neuman de Souza 1,2 1 Teleinformatics Engineering Department - Federal University of Cear
1 Teleinformatics Engineering Department - Federal University of Ceará 2 Computer Science Department – Federal University of Ceará
{felipe,marcio}@cenapadne.br, {rossana,aldri,neuman}@lia.ufc.br
2
3
Heterogeneity Geographic dispersion Transparent access to the resources
User and servers masquerading Abusive usage of the resources Non-authorized access to the services Subversion of the resources
4
Defense access control
Inefficient against internal attacks
Defense limitation of the resources usage
5
6
To ensure the environment is
not “ “contaminated” with malicious codes
To encourage a greater
participation and availability
Viruses, worms, trojans Defense virtualization
To ensure the environment is
not “contaminated” with malicious hosts
Applications endangered by
incorrect results
Non-trivial task
Data Transmission Job Processing
7
Do not cooperate to the network Avoid forwarding packets Refuse to process the jobs Omit information about available resources
Neglect help to other nodes OurGrid
Free-rider Consume resources from the grid without providing its own resources once
requested
Subvert the grid resources Provide an invalid result Spread viruses and worms
8
Majority Voting
Jobs replicas are distributed among the
nodes
Majority of results matching is taken as
valid
Spot-Checking
Test jobs whose results are previously
known
Blacklist
9
Nodes do not need to be tested so frequently It reduces the processing overhead
File sharing Minimize the presence of peers interested in diffusing false or
incomplete files, and also viruses and worms
10
Which units are faulty and which are fully functional Syndrome = set of obtained results
PMC, ADSD, Hi-ADSD Comparison-based
MM, Broadcast, and others
Task Result Comparison Result
11
Considers the heterogeneous and dynamic nature of
such environments
Public and private grids
Proposed Solution
Diagnosis combined to spot checking and reputation
Tests Format
to a same task
Time to answer a test
times
Test Job Test job Result
12
13
GridSim 3.3 New introduced methods Without reputation scheme
10.000 jobs 200 worker nodes Percentage of malicious nodes
1/6, 1/3 and 2/3 of the grid nodes providing bad results
Amount of test rounds
3, 5, 8, 10, 15 and 20
14
Amount of necessary test rounds Overhead Impact of the blacklist
Probability of 25% chances of returning an invalid
15
String “abcde” Set of primes {3,5,7,11} Result: 97 x 3 + 98 x 5 + 99 x 7 + 100 x 11 + 101 x 3
16
17
In the best case, 26 detected nodes In the worst, only 12 detected nodes
18
19
High overhead From 10.000 jobs, over
Acceptable trade-off With 1/6 of malicious
20
1/6 of Malicious Nodes 200 400 600 800 1000 1200 1400 1600 1800 3 5 8 10 15 20 Nº of Test Rounds Manipulated Results Sem Blacklist Com Blacklist 1/3 of Malicious Nodes 200 400 600 800 1000 1200 1400 1600 1800 3 5 8 10 15 20 Nº of Test Rounds Manipulated Results Sem Blacklist Com Blacklist 2/3 of Malicious Nodes 200 400 600 800 1000 1200 1400 1600 1800 3 5 8 10 15 20 Nº of Test Rounds Manipulated Results Sem Blacklist Com Blacklist
Without blacklist
Number of manipulated results
remains the same
Double the number of malicious
nodes, double the manipulated results
With blacklist
Manipulated results decrease with
more test rounds
Less efficiency with a higher
number of malicious nodes
Example: Manipulated results with
5 test rounds
20% 24% 28%
21
22
Apply these concepts in a diagnosis model Even with different quota of malicious nodes, practically
A further study to use a reputation scheme Scrutinize other possible metrics and scenarios
Treat other kinds of misbehavior nodes
Investigate the usage of this solution in real grids
OurGrid and Globus
23
Felipe Sampaio Martins felipe@cenapadne.br