One of the reasons blockchain has gained such a following in a relatively short time is its natural immutability.
By hosting multiple versions of a chain on hundreds, if not thousands, of hardened servers around the world, blockchain allows users to see, but not alter, past records – providing a level of trust that cannot be matched by single-version digital ledgers no matter how good their security is.
This is why many users are starting to look past blockchain’s ability to simply record transactions to more far-reaching applications like automated contract execution, scientific discovery and logistics management.
But there is a big difference between trustworthiness and reliability. While the technology does seem to have a lock on preventing records from being tampered with, are there vulnerabilities elsewhere that might cause some users to put too much faith in the technology’s ability to keep everything on the up and up?
Faults in the Blockchain
In a technology as complex as blockchain, the odds are high that things can go wrong, either by intent or happenstance. Recent tests by Portugal’s University of Coimbra have shown that a wide range of faults can enter a chain to disrupt the execution of smart contracts and cause other problems. The team, led by the university’s Centre for Informatics and Systems and the Department of Informatics Engineering, injected faults into 400 duplicate smart contracts and then compared them against the originals to see what, if any, discrepancies arose.
Using state-of-the-art detection tools, the team was able to determine that the original faults propagated out to more than 15,000 contracts, primarily by evading standard verification tools.
From there, they were able to introduce runtime errors into the chain and even cause severe latency, integrity and correctness issues. In the end, the group detected failures of various types in nearly 75 percent of all transactions, including full halts in execution and reversions back to original states.
Too Much Intelligence
Blockchains can also falter when exposed to too much artificial intelligence-generated data. A recent study by the US National Institute of Health found that misplaced trust in AI’s conclusions, coupled with a lack of explainability in the decision-making process, can jeopardize health outcomes. Most AI platforms are vulnerable to adversarial attacks and biased output, which can enter blockchains without the right safeguards.
In many cases, proposals from deep learning systems are accepted into blockchains tied to healthcare systems, as well as those related to security and financing. The NIH says this situation is urgent in that it has the potential to directly affect substantial assets in the healthcare industry, including human lives.
Fortunately, AI has the ability to effectively analyze blockchain’s performance when exposed to other AI platforms, providing a ready means to correct any deficiencies before them become critical. But these measures must be implemented proactively by each blockchains’ stakeholders.
With or without AI, however, many automated processes within blockchains cannot function with just the trusted data inside the chain, so they must turn to outside sources that may or may not be accurate. The most common way of doing this is through third-party oracles entrusted to bring in external data without compromising the integrity of the chain. But that leads to the question, can any given oracle be trusted?
In many cases, the answer is no. Research from Concordia University’s Institute for Information System Engineering shows that oracles can be “deviant and commit ill-intentioned behaviors, or be selfish and hide their actual available resources to gain optimal profit.” To counter this, the team has proposed a vetting system called BLOR (Beyesian Bandit Learning Oracles Reliability) that it says can identify oracles that are both reputable and cost-efficient. In essence, the system works by sharing the results of incorporated oracles while algorithmic processes are still running, providing an automated clearinghouse of sorts where good oracles are promoted and bad ones are weeded out.
The integrity of data sources and other elements that surround blockchains will always pose risks, but are there any ingrained aspects of blockchain that can prove unreliable as well? Perhaps the most fundamental is the anonymity of participants. While virtually everything else on a chain is open and transparent, the identity of wallet-holders is not. This means you never know who else is adding or deducting assets, or whether this activity is on the level.
With modern analytics, however, it is possible to identify these hidden identities by their transactions both inside and outside a chain, says Tung Li Lim, of blockchain solutions provide Elliptic. This, in fact, is how many chains abide by national and international laws. But this has also produced an escalating battle with those seeking anonymity through the use of mixers, tumblers, privacy wallets and other forms of obfuscation to hide their digital trails. Regulations are underway to block or limit the use of these tools, but new ones are appearing at a rapid pace.
No system, digital or otherwise, can be 100 percent reliable, and the fact is that blockchain deployments on the whole so far are proving to be highly effective at fulfilling their mandates. But as the use cases grow and the technologies surrounding digital ledgers become more varied and sophisticated, the world economy is likely to become more dependent on blockchain as time goes by.
This means that as the consequences of blockchain failure steadily mount, so too should the efforts to ensure ever-greater levels of reliability.