TL;DR
This paper provides a rigorous mathematical framework to model how blockchains survive persistent, multi-hacker cyber attacks. By accounting for random detection and system reset times, the authors prove that increasing the number of nodes doesn't just "feel" safer—it mathematically guarantees higher functional probability and longer survival times.
What Problem Does This Paper Address?
While blockchain technology is often touted for its security, it is not invincible. Attackers generally aim for one of two goals: a cyber destructive attack (corrupting at least half the nodes to invalidate data) or a cyber ransom attack (locking all nodes for payment).
Most existing security models assume a single attacker or ignore the "recovery" phase—the time it takes for an IT department to detect an intrusion and reset the system. In reality, multiple hackers often work independently, and resetting a complex network isn't instantaneous; it takes a random amount of time. This paper fills the gap by modeling the stochastic (random) behavior of an $n$-node blockchain facing $k$ independent hackers, specifically focusing on how random reset times affect long-term security.
Methods
The authors use stochastic modeling and renewal theory to describe the blockchain's lifecycle. They define a "cycle" as the period from a fresh start (or a completed reset) to the next time a reset is triggered.
Key components of their model include:
- Hacking Time ($X$): The time a hacker needs to breach a node.
- Detecting Time ($Y$): The time the IT department needs to spot an attack.
- Re-setting Time ($W$): The random duration required to restore the blockchain to a clean state.
To make the math tractable, they identify an embedded three-state semi-Markov process:
- State 1 (Functional): The system is running; no hacker has reached the threshold $m$.
- State 2 (Re-setting): An attack was detected, and the system is being restored.
- State 3 (Dysfunctional): A hacker successfully breached $m$ nodes before being detected.
By analyzing the transitions between these states, they derive exact formulas for the instantaneous functional probability (the chance the system is safe at time $t$) and the mean functional time (how long the system stays up on average).
Datasets
This is a theoretical paper, so it does not use real-world attack datasets. Instead, the authors validate their mathematical derivations using Monte Carlo simulations (30,000 to 50,000 iterations per example). They test their model against three common statistical distributions for hacking, detection, and reset times:
- Example 1: Exponential distributions (the "baseline" for memoryless processes).
- Example 2: Gamma distributions (more flexible for modeling durations).
- Example 3: Weibull distributions (often used in reliability engineering).
Key Insights and Ideas
- The "Race" Condition: The security of the blockchain is essentially a race between the hackers' progress ($Z_m$) and the IT department's detection ($Y$). If $Z_m < Y$, the hacker wins; otherwise, the system resets and all hacker progress is forfeited.
- Formalizing Intuition: The paper mathematically proves that security metrics are monotonically increasing functions of $n$ (the number of nodes). While this seems intuitive, providing a formal proof for arbitrary time distributions is a significant contribution.
- The Cost of Security: The authors introduce a cost-benefit analysis. While more nodes increase security, they also increase operational costs. They provide a formula to find the "Goldilocks" number of nodes that maximizes net revenue.
- Inevitability of Failure: The model shows that if the probability of a successful hack in a single cycle is greater than zero, the blockchain will eventually be hacked ($P_{mk}(\infty) = 0$). This highlights the importance of the "mean functional time" as a more practical metric than "infinite" security.
Key Results
- Security vs. Scale: In simulations, the mean functional time ($E[T_m]$) grew exponentially as the number of nodes increased. For example, in the Gamma-distributed model, increasing nodes significantly boosted the time the system remained functional.
- Optimal Node Count: In one numerical example (Example 1), the authors found that the expected net revenue reached its maximum at $m=40$ nodes, demonstrating how their formulas can be used for infrastructure planning.
- Hacker Density: They proved that as the number of independent hackers ($k$) increases, the functional probability of the blockchain decreases, as expected.
Limitations and Future Work
- Independent Hackers: The model assumes hackers work independently. It does not account for collusion or "51% attacks" where hackers pool their resources, which is a common threat in real-world Proof-of-Work blockchains.
- Identical Nodes: The nodes are assumed to be identical in terms of security. In reality, a network might have "weak links" with different hacking time distributions.
- Static Thresholds: The threshold $m$ is fixed. Future work could explore dynamic thresholds or adaptive security measures that change based on detected threat levels.
Python code for the simulations is available on the authors' GitHub.
Summary updated