This LUNES implementation is used to simulate the Bitcoin's protocol with options to run some known attacks.
- Mining
- Transactions
- Block and transaction propagation
- Blockchain concurrent behaviour
- Forking: it is possible to use the old version with no chain forks allowed or the new version, in which the blocks actually have an ID, a reference to the previous block, a certain position in a chain and the program keeps track of a given number (Default 10) of chain forks, forgetting gradually the shortest chains
- 51%
- DoS
- Selfish Mining
To reduce RAM usage all messages are pointers. This setup cannot be used with multiple LPs. The program must be run without parallelism.
run-blockchain
script is configured to execute the program and the simulator manager with 1 LP.
If your device has more than 16GB (at least, 36GB should be enough) you can change all function calls to pass by value all messages.
Run make
inside this folder to compile the binary: blockchain
.
First of all to create a corpus for the simulator run ./make-corpus <#NODES> <#EDGES> <#MAX_DIAMETER>.
For example ./make-corpus 1000 4 8 will create a graph with 1000 nodes, 4000 edges and a maximum diameter minor or equal than 8
To start the run use run-blockchain
Bash script: this will take care of sourcing and setting all global variables.
This script is used to start the SImulator MAnager (SIMA) and the binary blockchain
with all necessary arguments.
USAGE: ./run-blockchain --nodes|-n #SMH [--test|-t TESTNAME] [--debug|-d DEBUGCMD]
#SMH total number of nodes to simulate
[TESTNAME] name of the test to execute: 51, selfish or dos (optional)
[DEBUGCMD] used for injecting *trace commands (optional)
For example to run a simple simulation execute ./run-blockchain -n 1000
. This will prints all logs in stdout
. Logs are enabled or disabled using the #define
statements in sim-parameters.h
.
// stale blocks debug: prints all msgs with rejected blocks
#define STALEBLOCKDEBUG
//
// rejected transactions debug: prints all msgs
#define STALETXDEBUG
//
// transactions debug: prints all received and sent txs msgs
#define TXDEBUG
//
// request for specific blocks from peers: prints all msg to ask a peer for a specific block
#define ASKBLOCKDEBUG
//
// defining DOS it's possible to simulate the Denial of Service attack
#define DOS
Logs for mined and received blocks are enabled by default.
Warning: enabling logs will reduces performances and increments RAM usage.
To execute some attack scenarios use the --test
flag with 51
, selfish
or dos
. This flag initializes the simulator to run ALL tests with this configurations. Each run's output is saved in a txt
file inside the folder ./outputs
.
For example running ./run-run-blockid -n 10000 -t 51
will perform 100
tests with a single attacker with increasing hashrate (from 1 to 100). Outputs are collected in the ./outputs
folder.
Simulation output is composed of a tag
and a data
section (like a CSV file). Tags define what kind of output the program is printing:
BS
: aB
lock is mined and isS
ent to all the neighbors (BlockMsg
);BR
: aB
lock isR
eceived (BlockMsg
);TS
: aT
ransaction is created andS
ent to all the neighbors (TransMsg
);TR
: aT
ransaction isR
eceived (TransMsg
);BRS
: aB
lock is received but it'sS
tale (BlockMsg
);TRS
: aT
ransaction is received but it'sS
tale (TransMsg
);ABS
: aA
skB
lock message isS
ent (AskMsg
);ABR
: aA
skB
lock message isR
eceived (AskMsg
)
The format of each data section is:
BS
:clock - nodeid - blockminedid - hashrate
BR
:clock - nodeid - originalcreartor - timestamp - blockid
TS
:clock - nodeid - txid - from - to - blockid
TR
:clock - nodeid - originalcreartor - timestamp - txid - from - to - blockid
BRS
:clock - nodeid - originalcreartor - timestamp - blockid - latestblock
TRS
:clock - nodeid - originalcreartor - timestamp - txid - from - to
ABS
:clock - nodeid - blockid
ABR
:clock - nodeid - blockid - tonode
-- forking mode --
To evaluate the outcome of 51% attack run the script evaluate51.sh The output will be a file named 51-res.txt whose records are in the format:
attacker hashrate - number of blocks in the main chain mined by the attacker - length of the longest chain
Similarly to evaluate the outcome of the selfish mining attack run the script evaluateSelfish.sh The output will be a file named selfish-res.txt whose records are in the format:
attacker hashrate - number of succeded attacks
To evaluate the outcome of dos attack run the script evaluateDoS.sh
Documentation (Doxygen)
Run doxygen
in this folder to generate a HTML and a LaTeX report.
The command will generate a /doc
folder (ignored by the repository indexing) with two subdirectories:
html
: for the HTML report (open theindex.html
)latex
: runmake
inside this folder to generate the PDF report namedrefman.pdf
This execution mode will run 100 times the main simulation using an increasing value of the hashrate of an attacker (by default the attackerid
is 337).
Output of this tests can be found in ./outputs
directory with the name: test-51-H.txt
with H the value of the attacker's hashrate.
Also here the execution will run 100 times the main simulation using an increasing value of the hashrate of an attacker
Output of this tests can be found in ./outputs
directory with the name: test-selfish-H.txt
with H the value of the attacker's hashrate.
The execution is divided into epochs, each one last the same number of steps as the time to live for messages. In every epoch a victim node is chosen and it's evaluated the percentage of nodes among the honest ones that manage to receive a message.
Output of this tests can be found in ./outputs
directory with the name: test-dos-H.txt
with H the percentage of malicious nodes in the network.
51% attack has been tested both with 9 pools (whose overall hashrate is 82,7%) and without any pool. The pools are dealt as nodes with a very high hashrate, that represent what in the real life are groups of miners who put together their computing power in order to have a better chance to mine a node. The results between the two configuration slightly differ, in the no pools configuration the number of succeded attacks is higher. Then the attack can be tested with different network topologies and with a different number of nodes and edges.
Selfish mining has also been tested both with 9 pools and with no pools. The results are simular but in the no pools configuration the number of succeded attacks is slightly higher.
DoS attack has been tested with the proposed gossip algorithms. It turned out that Dandelion is much more prone to the attack with respect to the other protocols. However using the fail-safe machanism of Dandelion++ it's possible to reach the broadcast coverage levels.
Gabriele D'Angelo: [email protected]
Edoardo Rosa: [email protected]
Luca Serena [email protected]