Blockchain is a distributed peer-to-peer (P2P) way of recording digital interactions in a way that it provides builtin integrity of information, and security of immutability by design, making it very useful to ensure trust, security, and transparency in P2P trustless networks of devices. Blockchain, from its inception as the backbone technology of bitcoin, has evolved significantly in the last few years. With its highly significant features, it has shown serious potential as a key to redesign and improve wireless sensor networks (WSN)based Internet of Things (IoT). Such a consideration is quite promising, as the existing centralized architecture for IoT systems is incapable of handling the fast-paced growth of IoT. Despite that, tapping into the benefit of blockchain is not straightforward as the state-of-the-art implementation of blockchains are slow in transaction processing, resourceheavy, and lack transaction privacy.
Over the years, several alternatives have been proposed with improved transaction rates, including directed acyclic graphs (DAG)-based hashgraph [1]. Hashgraph uses gossip protocol
A.R. Shahid, N. Pissinou, E. Aguilar, and E. Perez were with the School of Computing and Information Sciences, Florida International University, Miami, FL. L.Njilla was with the Cyber Assurance Branch, Air Force Research Laboratory, Rome, NY. § Corresponding e-mail: ashah044@fiu.edu. This work was supported by U.S. AFRL FA8750-17-S-7003, NSF REU CNS-1851890, and FIU DYF. DISTRIBUTION A. Approved for public release; Distribution unlimited. Case Number 88ABW-2019-4869; Dated 08 Oct 2019. to establish consensus in the network where a node chooses multiple other nodes in the network to share all its information on the historical, as well as new transactions. The receiver nodes repeatedly do the same, which forms "gossip about gossip". Such a gossip mechanism impressively eliminates the need for vote casting or miner selection process, and subsequently yields high transaction rates. For instance, where bitcoin and ethereum's blockchains can process 5 and 15 transactions per second respectively, with enough computing resources, it is possible to yield a throughput of thousands of transactions per second with hashgraph [1], [2]. Babble blockchain [3] is an improvement of hashgraph which projects a hashgraph onto a blockchain to achieve an immutable ordered list of transactions while maintaining the high transaction rate. Despite this, babble is not suitable for resourceconstrained IoT devices as it requires high storage capacity and cannot guarantee the privacy of the data of devices shared in the network.
We address the gap between blockchain, resourceconstrained IoT, and privacy by developing PrivLiteChain, a babble-based lightweight blockchain platform with a focus on wireless sensor network-driven IoT systems. It is a part of our long term research goal to develop a lightweight, scalable, secure, and privacy-preserving blockchain platform for resourceconstrained mobile IoT systems [4], [5]. It exemplifies the effectiveness of our proposed spatiotemporal mobility-based lightweight blockchain technique, Sensor-Chain [4].
The system of PrivLiteChain is built around three important concepts: local differential privacy (LDP) to achieve transaction privacy, babble's hashgraph blockchain to achieve scalability, and controlled lifecycle of the blockchain to make it lightweight (Figure 1). In the current implementation, PrivLiteChain supports 1-dimensional sensor data, such as environmental data (e.g., temperature and humidity). We also consider each node is connected with all the other nodes in the network and transactions are happening in the form of sensed data broadcast. The sharing of private data over the blockchain can reveal privacy sensitive information about a node [6] for which it is important to privatize the data using LDP before it is released from a node. In a LDP setting, the original sensed data is made obfuscated by adding properly scaled statistical noise to it such that an adversary cannot identify the original data from all the possible values. Theoretically, a mechanism M satisfies -LDP ( ≥ 0), if and only if for any input x and x , we have, ∀y ∈ Range(M) :
In PrivLiteChain, we implemented Laplace mechanism to achieve -LDP. If i-th node's original sensed data is x i , then the broadcasted data (transaction) is y i = x i + n i where n i is a random noise drawn from Laplace distribution [7] (Figure 1(a)). This transaction is gossiped among the nodes in the network and mapped in the blockchain (the detail of the mechanism can be found in [3]). With each new release of noisy data for the same sensed original data (x i ) the privacy continues to degrade, which refers to the concept of privacy budget total . It defines the maximum amount of privacy leakage for an original sensed data. Once the privacy budget is completely consumed, a node needs to opt out from making transactions for certain amount of time such that there is a significant difference between the sensed data in the temporal domain. The third and final concept that we implemented in PrivLiteChain is a temporal constraintbased lifecycle controlling mechanism. We introduced this concept in our previous work on the design of lightweight blockchain [5]. A temporal constraint, T chain , is a limit on how long a blockchain can grow overtime. Let the time of genesis block creation and current time are t gen and t cur , and (t curt gen ) ≥ T chain . Then, at time t cur , the data over the blockchain is aggregated locally using a lightweight aggregation method, the existing blockchain is deleted, and a new chain is started with a genesis block containing the aggregated information. In the current implementation, the privacy budget total of each node is reset at each aggregation. However, in future we will implement a more rigorous technique to decide the time for privacy budget reset.
In this section, we demonstrate how PrivLiteChain achieves local differential privacy, scalability, and lightweight-ness. The current implementation of it is a permissioned one where every node knows the participants in the network in advance. The screenshots of the demo are presented in Figure 2. A video on the demo can be found at https://youtu.be/0Wcgsqtjvhs. The demonstration will use a laptop where the nodes will communicate with each other over TCP connections.
PrivLiteChain will be demonstrated in three phases. First, we will show the original babble blockchain for 20 nodes and how it processes the transactions in the hashgraph. The nodes generate the transactions at 50% probability. This phase generates some important statistics on babble blockchain, including the size of the blockchain, average transaction rate, average transactions in waiting pool, average block size (in bytes), average number of transactions per block, and block creation rate.
In the second phase, we will demonstrate the application of the Laplace mechanism to achieve LDP in the process of transaction generation. We will demonstrate it with privacy budget, total = 1. In the final and third phase, we demonstrate the application of temporal constraint on the lifecycle of blockchain. We will use T chain = 150 seconds in the setting. In this phase, the demonstration will generate the statistics on the size of PrivLiteChain. Figure 2(a) shows some important information on the aggregation.