On two variations of identifying codes

Identifying codes have been introduced in 1998 to model fault-detection in multiprocessor systems. In this paper, we introduce two variations of identifying codes: weak codes and light codes. They correspond to fault-detection by successive rounds. We give exact bounds for those two definitions for the family of cycles.

💡 Research Summary

The paper expands the classical notion of identifying codes, originally introduced in 1998 for fault‑detection in multiprocessor systems, by proposing two new variants that model multi‑round detection processes. The first variant, called a weak identifying code, relaxes the requirement that every vertex must be uniquely identified in a single observation. Instead, a vertex needs to be distinguished at least once over the course of several rounds; once a vertex has been identified, it no longer needs to be distinguished in later rounds. This model captures scenarios where faults are transient or where the monitoring budget allows only a limited number of detection cycles.

The second variant, termed a light identifying code, imposes a stricter, yet still multi‑round, condition: in each round all vertices must be uniquely identifiable using the current set of code vertices and their neighborhoods. If a vertex remains ambiguous after a round, additional code vertices may be added or the status of existing code vertices may change in subsequent rounds, allowing the vertex to become identifiable later. This framework reflects continuous monitoring environments where faults may accumulate and detection must be performed iteratively.

To obtain concrete results, the authors focus on the family of cycles (C_n) (simple graphs where each vertex has degree two and the vertices form a closed ring). Cycles are a natural testbed because of their high symmetry and because the exact minimum size of a classical identifying code for a cycle is already known to be (\lceil n/3\rceil). Using this baseline, the paper derives tight bounds for the minimum cardinalities of weak and light codes on (C_n).

Weak codes.
The authors prove that any weak code on a cycle must contain at least (\lceil n/4\rceil) vertices. The proof proceeds by contradiction: if the distance between consecutive code vertices exceeds four, there exists a pair of non‑code vertices whose closed neighborhoods intersect the code set in exactly the same way, violating the weak‑identification condition. Conversely, they construct explicit weak codes of size (\lceil n/4\rceil) (or (\lceil n/4\rceil+1) when (n) is not a multiple of four) by placing code vertices at regular intervals of four along the cycle. In each round, the yet‑unidentified vertices are adjacent to a code vertex that has not yet been used for identification, guaranteeing that every vertex becomes identified at least once. When (n) is a multiple of four, the bound is tight: the optimal weak code has exactly (n/4) vertices.

Light codes.
For light codes the lower bound is (\lceil n/5\rceil). The argument is similar: if the gap between consecutive code vertices exceeds five, two vertices will share identical code‑neighborhood signatures in every round, making them indistinguishable forever. The authors then present a constructive upper bound by defining a “progressive” scheme: start with a set of code vertices spaced five apart, and in each subsequent round add one more code vertex in a way that eliminates the remaining ambiguous vertices. This process ensures that after a finite number of rounds every vertex has a unique code‑neighborhood pattern. When (n) is a multiple of five, the construction yields an optimal light code of size (n/5); for other values of (n) the optimal size is either (\lceil n/5\rceil) or (\lceil n/5\rceil+1).

The paper’s methodology blends combinatorial lower‑bound arguments (based on pigeonhole‑type reasoning about neighborhoods) with explicit constructive algorithms that demonstrate achievable upper bounds. For the upper bounds, the authors carefully analyze the “transition” relation that describes how a vertex’s identification status evolves from one round to the next, guaranteeing that the process terminates with all vertices uniquely identified.

Beyond the technical results, the authors discuss the relationship between the new variants and classical identifying codes. A weak code can be viewed as a sub‑code of a classical code, and in some cycle lengths the optimal weak and classical codes coincide. Light codes generally require more vertices than weak codes, but there are instances (e.g., certain small cycles) where the optimal sizes match. These observations highlight how the multi‑round perspective interpolates between the extreme cases of “single‑shot” identification and “continuous” monitoring.

Finally, the authors outline future research directions. While the current work is confined to cycles, the techniques—particularly the round‑by‑round construction and the transition‑graph analysis—are expected to extend to other graph families such as trees, grids, and more general networks. Extending weak and light codes to probabilistic fault models, dynamic graphs, or heterogeneous monitoring resources could provide a richer theoretical foundation for designing fault‑tolerant distributed systems. The paper thus opens a new line of inquiry into temporal identifying codes, bridging static combinatorial design with the temporal dynamics of real‑world fault detection.