Stabilizing Maximal Independent Set in Unidirectional Networks is Hard

A distributed algorithm is self-stabilizing if after faults and attacks hit the system and place it in some arbitrary global state, the system recovers from this catastrophic situation without external intervention in finite time. In this paper, we consider the problem of constructing self-stabilizingly a \emph{maximal independent set} in uniform unidirectional networks of arbitrary shape. On the negative side, we present evidence that in uniform networks, \emph{deterministic} self-stabilization of this problem is \emph{impossible}. Also, the \emph{silence} property (\emph{i.e.} having communication fixed from some point in every execution) is impossible to guarantee, either for deterministic or for probabilistic variants of protocols. On the positive side, we present a deterministic protocol for networks with arbitrary unidirectional networks with unique identifiers that exhibits polynomial space and time complexity in asynchronous scheduling. We complement the study with probabilistic protocols for the uniform case: the first probabilistic protocol requires infinite memory but copes with asynchronous scheduling, while the second probabilistic protocol has polynomial space complexity but can only handle synchronous scheduling. Both probabilistic solutions have expected polynomial time complexity.

💡 Research Summary

The paper investigates the self‑stabilizing construction of a maximal independent set (MIS) in directed (unidirectional) networks, a setting where communication links are asymmetric and traditional symmetric‑graph algorithms do not directly apply. A self‑stabilizing algorithm must guarantee that, starting from any arbitrary global state caused by faults or attacks, the system converges to a legitimate configuration (a MIS) in finite time without external help. The authors explore both impossibility results and constructive solutions, distinguishing between uniform networks (all processes run identical code and have no unique identifiers) and networks where each node possesses a unique identifier (UID).

Negative results.
For uniform networks the authors prove that deterministic self‑stabilization of MIS is impossible. The proof constructs two distinct initial global states that induce the same local state at a particular node. Because the node’s code is deterministic, it will execute the same actions in both executions, leading the system into a non‑terminating cycle that never satisfies the MIS property. Consequently, no deterministic protocol can guarantee convergence from every possible initial configuration.

The paper also shows that the silence property—where communication ceases after some finite time and the system remains static—cannot be achieved, neither by deterministic nor by probabilistic protocols, in the directed setting. The asymmetry of information flow prevents any node from locally detecting global convergence; thus any protocol that eventually stops sending messages would either violate the MIS condition or be unable to guarantee convergence.

Positive results for UID‑based networks.
When nodes have distinct identifiers, the authors present a deterministic protocol that works under a fully asynchronous scheduler. Each node repeatedly gathers the UIDs of its incoming neighbors, compares them, and decides whether to join the independent set based on a simple priority rule (the smallest UID among a node and its in‑neighbors wins). Nodes that win broadcast a “selected” flag, causing their out‑neighbors to withdraw from contention. The algorithm repeats until no conflicts remain. The analysis shows worst‑case convergence in polynomial time (O(n³) rounds) and requires only O(log n) bits of local memory per node, making it practical for large systems.

Probabilistic solutions for uniform networks.
Two probabilistic protocols are proposed for the uniform case.

1. Infinite‑memory, asynchronous protocol. Each node repeatedly picks a random color (e.g., 0 or 1) and, if none of its in‑neighbors share the same color, it declares itself part of the MIS. In case of a conflict, the node retries in the next round. Because the random choices are independent, the expected number of rounds until all conflicts disappear is polynomial (Θ(n log n)). The protocol tolerates any asynchronous schedule but assumes unbounded local memory to store an unbounded sequence of random bits.

2. Polynomial‑memory, synchronous protocol. Here the system proceeds in synchronized rounds. In each round all nodes simultaneously select a random color; nodes that experience a conflict (i.e., share the color with any in‑neighbor) remain undecided, while conflict‑free nodes join the MIS and inform their out‑neighbors. The memory requirement is O(log n) bits (to store the current random choice and a few flags). The expected convergence time remains polynomial, but the protocol relies on a globally synchronized scheduler, which limits its applicability in fully asynchronous environments.

Overall contribution.
The work delineates the boundary between what is impossible and what is achievable for self‑stabilizing MIS in directed networks. It establishes that deterministic self‑stabilization and silence are unattainable in uniform directed graphs, while showing that the addition of unique identifiers restores feasibility with a simple, polynomial‑time deterministic algorithm. For the identifier‑free scenario, the authors provide the first probabilistic self‑stabilizing MIS protocols, highlighting a trade‑off between memory consumption and scheduling assumptions. These results are relevant to a range of practical systems—wireless sensor networks, robotic swarms, and any distributed platform where communication links are inherently unidirectional—and they lay a theoretical foundation for future research on fault‑tolerant, self‑organizing structures in asymmetric networks.