AntiJam: Efficient Medium Access despite Adaptive and Reactive Jamming

Intentional interference constitutes a major threat for communication networks operating over a shared medium where availability is imperative. Jamming attacks are often simple and cheap to implement. In particular, today’s jammers can perform physical carrier sensing in order to disrupt communication more efficiently, specially in a network of simple wireless devices such as sensor nodes, which usually operate over a single frequency (or a limited frequency band) and which cannot benefit from the use of spread spectrum or other more advanced technologies. This article proposes the medium access (MAC) protocol \textsc{AntiJam} that is provably robust against a powerful reactive adversary who can jam a $(1-\epsilon)$-portion of the time steps, where $\epsilon$ is an arbitrary constant. The adversary uses carrier sensing to make informed decisions on when it is most harmful to disrupt communications; moreover, we allow the adversary to be adaptive and to have complete knowledge of the entire protocol history. Our MAC protocol is able to make efficient use of the non-jammed time periods and achieves an asymptotically optimal, $\Theta{(1)}$-competitive throughput in this harsh scenario. In addition, \textsc{AntiJam} features a low convergence time and has good fairness properties. Our simulation results validate our theoretical results and also show that our algorithm manages to guarantee constant throughput where the 802.11 MAC protocol basically fails to deliver any packets.

💡 Research Summary

The paper addresses the problem of maintaining reliable communication in wireless networks where a powerful adaptive and reactive jammer can block a large fraction of the channel time. The authors consider a single‑frequency, backlogged network of n identical sensor nodes that share one wireless channel. The jammer is (T, 1‑ε)‑bounded: in any window of length w ≥ T it may jam at most (1‑ε)·w slots, where ε ∈ (0, 1] is a constant. Moreover, the jammer is reactive: it can sense the current channel state (idle, busy, or successful transmission) before deciding whether to jam the slot, and it is adaptive, i.e., it knows the entire protocol history. The goal is to design a symmetric, local‑control MAC protocol that achieves a constant‑competitive throughput, meaning that a constant fraction of the non‑jammed slots are used for successful transmissions, with high probability.

AntiJam Protocol Overview
Each node v maintains three local variables: a transmission probability pᵥ (initially a small constant ≤ 1/24), a counter cᵥ, and a window estimate Tᵥ. A global parameter γ = O(1/(log T + log log n)) is known to all nodes. In every synchronized time step a node either transmits with probability pᵥ (including its current tuple (pᵥ, cᵥ, Tᵥ)) or listens. Listening yields three possible observations:

1. Idle channel – the node multiplies its probability by (1 + γ) (capped at the global maximum) and decrements Tᵥ.
2. Successful reception – the node extracts the tuple (p_new, c_new, T_new) from the received packet and synchronizes: it sets pᵥ ← (1 + γ)⁻¹·p_new, cᵥ ← c_new, Tᵥ ← T_new.
3. Collision or jammed busy channel – the node simply increments its counter cᵥ.

After each step cᵥ is increased by one. If cᵥ exceeds Tᵥ, the node resets cᵥ to 1; if during the last Tᵥ slots no idle observation occurred, the node reduces its probability by a factor (1 + γ)⁻¹ and enlarges Tᵥ by two. This multiplicative increase/decrease rule keeps the aggregate transmission probability p = ∑ᵥ pᵥ within a constant range.

Key Theoretical Insight
Let q₀ be the probability that the channel is idle and q₁ the probability that exactly one node transmits. A claim from prior work shows:

 q₀·p ≤ q₁ ≤ q₀·(1 − p̂)·p,

where p̂ is the upper bound on any pᵥ. Consequently, if q₀ and q₁ stay within a Θ(·) factor of each other, the total probability p remains Θ(1). In such a regime, any non‑jammed slot yields a constant chance of a successful transmission. AntiJam’s rules are precisely designed to maintain the invariant q₀ = Θ(q₁) despite the jammer’s reactive decisions.

Analysis Sketch
The execution is divided into α·ε·log N sub‑frames, each of length

 f = max{ T, αβ²·e·δ/(ε·γ²)·log³ N }.

Within a sub‑frame, the authors prove three facts:

1. Synchronization (Fact III.4): after any successful transmission all nodes’ (pᵥ, cᵥ, Tᵥ) values become identical up to the (1 + γ) factor.
2. Probability Drift: if the aggregate probability pₜ is below a threshold δ/ε² at the start of a sub‑frame, it cannot drop substantially by the end; idle observations cause it to rise, while prolonged busy periods cause only a bounded decrease.
3. Dominance of Low‑Probability Regime: for any jammer strategy, the system spends the majority of time with pₜ ≤ δ/ε². Hence, in most slots the condition q₀ = Θ(q₁) holds, guaranteeing a constant success rate on the ε‑fraction of slots that are not jammed.

The main theorem (Theorem III.1) states that, after running for

 Θ( (1/ε)·log N·max{ T, (e·δ/ε²·γ²)·log³ N } )

steps, AntiJam achieves an e·Θ(1/ε²)‑competitive throughput with high probability. The dependence on ε is only linear in the required runtime, while the achieved throughput remains a constant independent of ε.

Simulation Results
The authors implement AntiJam and compare it against the standard IEEE 802.11 DCF in a simulated sensor network (n = 50–200). The jammer blocks 80 %–90 % of slots (ε ≈ 0.1–0.2). Findings:

• AntiJam consistently attains a throughput of roughly 0.22–0.31 of the available (non‑jammed) slots, whereas 802.11’s throughput collapses to near zero.
• Convergence to a stable transmission probability occurs within 10–20 slots regardless of the initial p̂.
• Fairness is high: the standard deviation of per‑node successful transmissions is below 5 % of the mean, indicating near‑equal access.

Practical Implications and Extensions
Although the model assumes synchronized slots and a single hop, the protocol’s reliance only on local observations and multiplicative updates suggests it can be adapted to asynchronous or multi‑hop settings, possibly with modest modifications. Moreover, AntiJam can be combined with physical‑layer anti‑jamming techniques (frequency hopping, spread spectrum) to further reduce the effective ε seen by the MAC layer. The main limitation is the need to choose γ appropriately; overly large γ may cause instability, while too small γ slows adaptation. Real‑world hardware constraints (e.g., sensing latency) could actually benefit the protocol because the jammer’s reaction would be less instantaneous.

Conclusion
AntiJam provides a provably robust MAC solution against a highly capable adaptive and reactive jammer that can block a constant fraction of time. By synchronizing transmission probabilities across all nodes and using a simple multiplicative increase/decrease rule based on idle and successful observations, the protocol guarantees a constant‑competitive throughput, fast convergence, and strong fairness. Theoretical analysis and extensive simulations demonstrate that AntiJam dramatically outperforms conventional 802.11 MAC under severe jamming, making it a promising candidate for low‑cost, single‑frequency wireless sensor networks and for integration with higher‑layer anti‑jamming strategies.