Event
Ph.D. Dissertation Defense: Mustafa Doger
Monday, November 30, 2026
11:00 a.m.
AVW 2224
ANNOUNCEMENT: Ph.D. Dissertation Defense
Name: Mustafa Doger
Committee:
Professor Sennur Ulukus (Chair)
Professor Adrian Papamarcou
Professor Ang Li
Professor Dana Dachman-Soled
Professor MohammadTaghi Hajiaghayi (Dean's Representative)
Date/time: Monday, November 30 at 11:00 AM
Location: AVW 2224
Title: Security and Incentive Problems in Proof-of-Work Blockchains
Abstract: Nakamoto consensus allows distributed nodes to achieve consensus in permissionless settings. In essence, Nakamoto's blockchain solution combines the longest chain protocol with a Proof-of-Work (PoW) mechanism to ensure that a malicious agent is unlikely to undermine the consensus as long as the fraction of computational power it possesses is less than a half. This dissertation focuses on the security and incentive problems associated with PoW blockchains such as double-spending problems, selfish mining attacks, as well as block withholding strategies against pool mining systems.
Since the hash functions associated with the PoW mechanism behave similarly to the random oracles, security and incentive problems in such blockchains can be modeled as complicated stochastic processes depending on adversarial strategies, peer-to-peer network delays and mining rates. We explore bounds as well as trade-offs related to security guarantees under different network delay models and mining strategies, and investigate incentive attacks that adversarial participants may employ to increase their mining rewards. We provide novel tight bounds that extend the state-of-the-art security guarantees, novel incentive attacks that increase adversarial reward fractions, and study trade-offs and connections to other fundamental blockchain mechanisms.
First, we study security-latency bounds for Nakamoto consensus, i.e., how secure a block is after it becomes $k$-deep in the chain under bounded network delays. We improve the state-of-the-art bounds by analyzing the race between adversarial and honest chains in three different phases. We find the probability distribution of the growth of the adversarial chains when a target block becomes $k$-deep in the chain. We analyze certain properties of this race to model each phase with random walks that provide tighter bounds than the existing results. Combining all three phases provides novel upper and lower bounds for blockchains with small fork rates.
Second, we analyze how secure a block is after the block becomes $k$-deep for Nakamoto consensus under an exponential network delay model. We provide the fault tolerance and extensive bounds on safety violation probabilities given mining rate, delay distribution and confirmation rules. Next, modeling the blockchain system as a batch service queue with exponential network delay, we connect the security-latency analysis to sustainable transaction rate of the queue system. As our model assumes exponential network delay, batch service queue models give a meaningful trade-off between transaction capacity, security and latency. Our results indicate that, by simply picking the $k=7$ block confirmation rule in Bitcoin instead of the convention of $k=6$, the mining rate can be increased so that the latency decreases and the throughput increases sixfold with the same safety guarantees. We further consider adversarial attacks on the queue service to hamper the service process. In an extreme scenario, we consider the selfish-mining attack for this purpose and provide the maximum adversarial block ratio in the longest chain under the exponential delay model. The ratio in turn reflects the maximum rate of decrease in the sustainable transaction rate of the queue.
Third, we lay out a general ruin-theoretical framework to study the double-spending probabilities for Nakamoto consensus under more general network delay scenarios. Theoretical guarantees for double-spending probabilities for the Nakamoto consensus under the $k$-deep confirmation rule have been extensively studied for zero/bounded network delays and fixed mining rates. Recent studies have shown that the PoW protocol is secure under random delay models as well. Further, the intrinsic characteristics of peer-to-peer network delays imply that the effective honest mining rate is not fixed, but rather an arbitrary function of time in block delivery periods. Our proposed framework allows us to provide theoretical bounds for the security-latency problem under delays modeled by general random distributions. The bounds we provide are tight and explicit, which only require determining the distribution of the number of Poisson arrivals during the delay. Next, using the same ruin-theoretical framework, we give estimates of the double-spending probabilities in a more realistic scenario where the effective honest mining rate is not fixed, but rather is an arbitrary function of time in block delivery periods during which the adversary is mounting the private attack. We obtain the double-spending probabilities using the empirical delays reported in the Bitcoin network and present numerical examples to validate the effectiveness of the proposed analytical methods. Our framework is particularly useful for Nakamoto blockchains with high fork-rates where the existing models provide too optimistic/pessimistic results, and more realistic estimates of double-spending are needed.
Fourth, we give a rigorous stochastic analysis of an attack where the goal of the adversary is to double-spend while mining selfishly. Conventional double-spending attack models ignore the revenue losses stemming from the orphan blocks. On the other hand, selfish mining literature usually ignores the chance of the attacker to double-spend at no-cost in each attack cycle. To consider both, we first combine stubborn and selfish mining attacks, i.e., construct a strategy where the attacker acts stubborn until its private branch reaches a certain length and then switches to act selfish. We provide the optimal stubbornness for each parameter regime. Next, we provide the maximum stubbornness that is still more profitable than honest mining and argue a connection between the level of stubbornness and the $k$-confirmation rule. We show that, at each attack cycle, if the level of stubbornness is higher than $k$, the adversary gets a free shot at double-spending. At each cycle, for a given stubbornness level, we rigorously formulate how great the probability of double-spending is. We further modify the attack in the stubborn regime in order to conceal the attack and increase the double-spending probability.
Fifth, we consider parallel PoW protocols, develop hard-coded incentive attack structures against them and introduce a semi-parallel PoW protocol that outperforms both Nakamoto consensus and the existing parallel PoW protocols from practical perspectives. Parallel PoW protocols have been suggested in the literature to improve the safety guarantees, transaction throughput and confirmation latencies of Nakamoto consensus. Our theoretical results and simulations show that the existing parallel PoW protocols are more vulnerable to incentive attacks than the Nakamoto consensus, e.g., attacks have smaller profitability threshold and they result in higher relative rewards. We analyze the proposed semi-parallel PoW protocol and show that it improves on both Nakamoto consensus and the existing parallel PoW protocols from most practical perspectives such as communication overheads, throughput, transaction conflicts, incentive compatibility of the protocol as well as a fair distribution of transaction fees among the voters and the leaders. We use state-of-the-art analysis to evaluate the consistency of the protocol and consider Markov decision process (MDP) models to substantiate our claims about the resilience of our protocol against incentive attacks.
Sixth, for incentive attacks such as selfish mining, block withholding and coin hopping under Bitcoin's Difficulty Adjustment Algorithm (DAA), we rigorously study the short-term revenue changes per hashpower of the adversarial and honest miners and introduce a new efficiency metric defined as the revenue/cost per hashpower per time for long-term effects. Our results indicate that the short-term benefits of intermittent mining strategies are negligible compared to the original selfish mining attack, and in the long-term, selfish mining provides better efficiency. We further demonstrate that a coin hopping strategy between Bitcoin and Bitcoin Cash relying on Bitcoin DAA benefits the loyal honest miners of Bitcoin in the same way and to the same extent per unit of computational power as it does the hopper in the short-term. For the long-term, we establish a new boundary between the selfish mining and coin hopping attack, identifying the optimal efficient strategy for each parameter. For block withholding strategies, it turns out, the honest miners outside the pool profit from the attack, usually even more than the attacker both in the short-term and the long-term. Moreover, a Power Adjusting Withholding (PAW) attacker does not necessarily observe a profit lag in the short-term. In other words, even without a difficulty adjustment, a PAW attacker can make profits. It has been long thought that the profit lag of selfish mining is among the main reasons why such an attack has not been observed in practice. We show that such a barrier does not apply to PAW and relatively small pools are under an immediate threat. In addition to optimization of revenue ratio, we study alternative objective functions that the adversary may choose to optimize to get the most out of the attack among all miners as well as possibly get immediate revenue increase even before the difficulty adjustments.
Finally, we study the resilience of pool mining reward schemes used in practice: Pay-Per-Share (PPS), Pay-Per-Last-N-Shares (PPLNS), Proportional (PROP) and exponential-decay (EXP), taking the difficulty adjustment (DA) into account. Mining pools share block rewards among their members according to a payout scheme. A pool member can increase its revenue by withholding the blocks it finds or by moving its hashpower in and out of the pool (pool hopping). Prior analyses rely on simplifying assumptions: withholding is analyzed under a global-reset version of the proportional scheme (GRP) that pools do not use, share-based schemes only under a fixed difficulty, and hopping only for a negligible miner. For withholding, we propose Temporary Power Adjusting Withholding (T-PAW), which generalizes and improves known withholding attacks, and derive its exact revenues under each block-dependent scheme. We show that GRP overstates the gain under PROP by up to more than an order of magnitude and understates it under EXP. For PPS, the optimal attack is to withhold all blocks, which raises the revenue of an adversary with a fraction $f$ of the hashpower from $f$ to $\frac{f}{1-f}$ per epoch after the DA, at the expense of the pool operator. For hopping, we introduce Power Adjusting Hopping (PAH) for an adversary of any size and derive its revenues in closed form. PROP is most vulnerable to hopping and PPS to withholding by a large adversary, whereas EXP is vulnerable to both: a slow decay invites hopping and a fast decay invites withholding. Among the schemes we study, PPLNS is the most robust. Monte Carlo simulations verify our results.
