Blockchain technology is marketed as the Web 3.0 and because of it’s distributed structure it wipes out single points of failure. But does that mean there are no points of failures at all? Let’s look at some important blockchain hacks / failures from the tech perspective.
[Remark: This is not about $$$ Bitcoin hacks, where lousy DB implementations, web applications, key handling or simply social engineering let to hacked bitcoin exchanges or wallets.]
51% attack
From a technical point of view, the 51% attack is probably the most famous one. It’s as simple as obvious – in theory you need 50% of the networks hash rate to control the blockchain in the long run. The estimated cost of such an attack against the Bitcoin network today are 1.3 billion $.
But for smaller proof-of-work blockchains with less hashrate this is a potential threat. And this attack got successful executed on an Ethereum like blockchain called Krypton. The attacker rented mining power at a cloudmining provider (NiceHash) and at the same time started to DDoS existing Krypton ming nodes. Now he was able to double spend Krypton Tokens by simple redoing the proof-of-work.
Krypton’s Founder stated:
“This attack may be a “dry run” intended as proof of concept before targeting other Ethereum based blockchains. […] Ethereum based blockchains are being targeted predominantly because they’re easy to fork and manipulate offline, while being used in conjunction with DDoS attacks.”
After the hack Krypton changed to a “bitcoin-based proof-of-work”, before the core devs dropped the project.
Smart contract bug
The infamous multimillion DAO hack was enabled due to a smart contract bug. The DAO (decentralized autonomous organization) was a smart contract construct written in solidity, a JavaScript similar language. The DAO was basically an investment fund where the business logic was written in code. So the investors aka token holders can vote for business projects and get their return on invest.
One of the features as a token holder was to propose a split of the DAO. This is the way the hacker started his attack. After the voting period for the proposal expired, the split executed and a new DAO got created, nothing special. While the splitDAO function runs, the tokens of the curator (attacker) get sent towards the new DAO. The func withdrawRewardFor gets called during the func splitDAO as well and tries to pay out any available investment returns. Now the critical part! The hacker called the splitDAO func again before the balances got updated, meaning before the function terminated. Like this the old DAO again sent tokens towards the new DAO. The hacker repeats this step over and over again.
This way he drained the DAO and captured 3.6 million ether an equivalent of $50 million at this time. The hack led to a hard fork and a split of Ethereum into Ethereum Classic and Ethereum
Phil Daian did a real good technical analysis of the hack.
Blockchain Protocol Attack
The Bitcoin network is seen as the original blockchain. But since quite a while the community is struggling to get consensus over the scaling problem. One solution is Bitcoin Unlimited (BU), which is a fork of the Bitcoin Core protocol, to increase the block size.
Recently the Bitcoin Unlimited network suffered from a critical bug that got exploited. Besides increasing the block size, BU introduced Xtreme Thinblocks, which helps reduce block size by filtering transactions, which initially got into the mempool. However this feature has had at least two critical bugs, which got detected and fixed. With the development happening publicly on github, somebody took notice of the commit and exploited this bug. The bug was a reachable assertion in C++, see the simplified code below.
void SendXThinBlock(CBlock &block, CNode* pfrom, const CInv &inv)
{
if (inv.type == MSG_XTHINBLOCK)
{
// code
}
else if (inv.type == MSG_THINBLOCK)
{
// code
}
else
{
assert(0);
}
// more code
}
The exploit was just to send a GET_XTHIN with an invalid message type (see python code). At the time of the attack 774 of 6415 nodes where running BU, the attack took around 500 down. The xthin block feature is probably to blame for even more downtime in BU nodes, see picture below.
This low code quality inside BU gives the scaling discussion new fire and could possible decide it.
Bitcoin Unlimited nodes online. Attack occured on 14.March – first drop in the chart.
Antbleed: Hardware Backdoor
In Bitcoin mining the use of ASICS is the only economically reasonable way to go. Bitmain is the biggest producer of mining ASICS. It is estimated that around 70% of the global Bitcoin hashrate comes from Bitmain’s Antminers. At the same time they are operating the biggest mining pool AntPool with around 20% of the global hashrate.
Recently a backdoor in the Antminer firmware has been discovered. The firmware regularly connects to a server controlled by Bitmain and transmits information including the device’s serial number, MAC address and IP address. Nothing special, since this could just track the distribution of Bitmain’s sales, but the response of the server can disable the requesting miner, see the code below or on pastebin (since it got deleted on github). Easiest solution is to edit your etc/hosts file and dissolve auth.minerlink.com to localhost.
Bitmain stated, they built a “remote shutdown backdoor” only for test purposes and won’t use it. More info on antbleed.
#define AUTH_URL "auth.minerlink.com"
//code
void send_mac() {
//code
get_mac("eth0",&mac);
while(need_send)
{ // code
stop_mining = setup_send_mac_socket(s);
if(stop_mining)
{
applog(LOG_NOTICE,"Stop mining!!!");
break;
} // code
} }
Webapplication Attack
Despite claiming not to mention webapplication flaws, this one led to a hard fork, so it’s maybe worth mentioning.
Steemit is a social media platform based on a blockchain with it’s one currency called steem, to encourage high-quality content. They want to avoid censorship and any “facebook-like” middle-man, by being decentralized. Nevertheless there was a central vulnerability regarding the facebook and reddit login integration on the frontend, said CEO Ned Scott. This way the hacker got access to around 260 accounts and drained $85,000. All the money got refunded, as mentioned, due to a hard fork, which is quite easy to implement in steemit because of their witnesses consensus mechanism.
Conclusion
Blockchain technology is an ongoing beta test and you shouldn’t believe the promise of salvation that blockchain can’t be hacked. There hasn’t been any major flaw / hack that dismisses the core technology as a whole. But usually nobody sees the black swan coming.
- The Future of Bitcoin - 6. July 2017
- Security of Things – World Conference - 21. June 2017
- Immutable, reliable, secure – A brief history of blockchain security - 15. May 2017
