*3.3. Processing Capability*

The processing capability limitation is extremely relevant for blockchain distributed ledgers that allow the implementation of complex functionality and computations through smart contracts. For example, in Ethereum, the concept of gas was introduced to measure the amount of computational effort. When running smart contracts, each executed operation and processed byte of data is paid for with gas. This mechanism prevents an attacker from running extremely long tasks or infinite loops since the attacker would need to provide enough reward to incentivize the miner to execute each operation. When the reward provided runs out, the computation stops, and the transaction is dropped, thus avoiding situations where the nodes become unavailable due to attacks or complex computations. By integrating the Protocol and Network Tier system with external services, this shortcoming can be solved.

The nowadays solutions for addressing complex computational problems in blockchain ledgers are Oracles [128] and Proof of Computation mechanisms [129].

The Oracles are mechanisms that provide a secure connection between the blockchain and the outside world. They act as a trusted third-party entity, or a network of entities, for the Protocol and Network Tier. The Oracles can be used to offer results from different URLs, [130], IPFS, or units responsible for running more complex algorithms. The problem with interacting with the Oracles, directly from the chain, is that the response must be the same across any number of requests issued by the nodes during mining. This proves to be almost impossible when accessing dynamic changing data regarding weather, stock prices, etc. One problem that can appear in the Oracle-based system is data tampering or man-inthe-middle attack. In this sense, the Oracles are responsible to ensure the authenticity of data through authenticity proofs. One problem that persists is the centralized nature of the Oracles. The mechanism is presented in Figure 6 for a blockchain ledger [131]. An event containing details about the request is issued from the blockchain and intercepted by the Oracle. The necessary information is retrieved from external services and published back on the chain through a callback transaction.

Other implementations aim to outsource computing-intensive problems to off-chain nodes by implementing a Proof of Computation mechanism. Compared to the Oracles the Proof of Computation is a better choice. It is implemented without relying on any trusted party and provides validation of the result implemented directly on-chain. The proposed solutions show grea<sup>t</sup> potential for use cases requiring security and correctness validation for more complex computations than the ones that can be handled on-chain.

Figure 7 shows the Proof of Computation mechanism of TrueBit [129]. It relies on Ethereum smart contracts and gives the possibility of peers to request solutions for complex computational tasks. A Solver that has enough computational resources will run the tasks outside the chain and submit its results.

**Figure 6.** Scalability Tier—Oracle for External Services Integration.

**Figure 7.** Scalability Tier—Proof of Computation mechanism.

Several Verifiers peers can evaluate the results, and if a disagreement occurs, a Challenger can contradict the result published, by starting a Challenge Game. Several rounds of proofs are registered on the chain to check whether the computation was done correctly. In the end, the winning part will receive a reward for its cooperation, while the part proven to be wrong will be charged for its actions. Another set of actors are de Judges that given the proof of the solution can easily verify the correctness of the game. However, a big drawback of the system is that it is limited to running tasks written in WASM [132]. Enigma [133] proposes a similar concept, of outsourcing the computation but with an added layer of privacy over the data and computation performed. By leveraging on secure multi-party computation, the proposed solution distributes the data across several nodes for computation. As a result, no central node will have access to the entire problem or solution, but to a seemingly unintelligible part of it. The proposed solution is not ye<sup>t</sup> released in production, but it is currently tested on the Ethereum test network [134].
