Jamming-Resilient OTFS-NOMA

arXiv:2606.09753 · eess.SP · Submitted 2026-06-08 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Electrical Engineering. The day in electrical engineering: signals, systems and the hardware that carries them.

Rosa: We're looking at the paper: "Jamming-Resilient OTFS-NOMA".

Dev: Here's the gist of the paper. We propose a sparse delay-Doppler NOMA scheme that is resilient to intentional jamming The closed-form BER expression derived from this conditioning argument has no jammer-induced floor, in contrast to the well-known partial-band error floor of conventional OTFS-NOMA This architecture delivers at least a 40 dB BER-ratio improvement against pattern-aware jammers and reaches a floor BER at eight users by SNR around 20 dB System Architecture and Design Ingredients The proposed architecture is built upon a five-ingredient stack tightly co-designed for the jamming-resilient multi-user setting.

Rosa: This stack includes sparse data placement on the DD grid, unitary spreading across all Nb bins, per-frame randomization of the active bin subset via a shared pseudo-random seed (the “C5” protocol), superincreasing power allocation (PA) that makes the O(K) successive interference cancellation (SIC) decoder maximum-likelihood (ML)-optimal on the NOMA composite constellation, and disjoint pairwise clustering when more than four users are present.

Dev: The receiver threshold-excises jammer-contaminated bins, recovers the sparse signal by least squares, and decodes per-active-bin SIC.

Rosa: How the Gains Come From The key gains stem from several analytical properties that overcome conventional limitations.

Paper summary: Rosa: So, to recap what we just looked at with this paper "Jamming-Resilient OTFS-NOMA", the core thesis is that they propose a sparse delay-Doppler NOMA scheme specifically designed to be resilient to intentional jamming.

Dev: They achieve this by using a sparse data placement on the delay-Doppler grid, spreading that result over all bins through a unitary precoder, and then redrawing the active subset each frame using a pseudorandom seed shared with the receiver.

Taro: The main claim is that their closed form bit error rate expression has no jammer induced floor, which contrasts sharply with the partial band error floor seen in conventional OTFS-NOMA.

Rosa: That means they claim at least a forty decibel bit error rate ratio improvement against pattern aware jammers, and they can reach a floor bit error rate at eight users when the signal to noise ratio is around twenty decibels.

Dev: The method involves building this stack from five ingredients: sparse data placement, unitary spreading, per frame randomization of the active bin subset via the C5 protocol, super increasing power allocation for optimal SIC decoding, and disjoint pairwise clustering for more than four users.

Taro: They handle the receiver side by having it threshold excise jammed bins before a least squares step recovers the sparse signal and then uses per-bin successive interference cancellation to decode.

Rosa: The gains come from several analytical properties. They use a Marchenko-Pastur conditioning bound which controls any random unitary submatrix conditioning and has no jammer induced floor, as well as super increasing power allocation with a Merkle-Hellman knapsack construction proving it matches maximum likelihood detection exactly on the NOMA composite constellation.

Dev: Plus, for more than four users, partitioning them into pairs and assigning each pair its own disjoint bin subset achieves that eight user floor at signal to noise ratio around twenty decibels.

Taro: They also discuss how they defend against pattern aware adversaries through the C5 protocol, where the transmitter and receiver independently generate the active set using a shared secret seed.

Rosa: Even if that seed is compromised, random unitary submatrices remain well conditioned with high probability, so performance stays within the unjammed case envelope. They also mention a Sylvester replication trap affecting structured patterns motivating them to use randomized patterns based on benign channel grounds alone.

Dev: The optimal scaling for more than four users involves partitioning users into disjoint clusters of size two and assigning each cluster its private bin subset, which leads to the floor bit error rate at eight users by signal to noise ratio around twenty decibels.

Taro: So they synthesize a jammer resilient system from half user NOMAs on orthogonal delay-Doppler subsets, which is a practical way to manage interference.

Rosa: This architecture synthesizes a jammer resilient system from two user NOMAs on orthogonal delay-Doppler subsets, and it’s designed for the multi user setting.

Dev: It’s important to remember that their method uses an O(Nb log Nb) operation when looking at the receiver side.

Taro: So how does this change things practically for someone who is actually running a system outside a lab?

Conclusion: Rosa: So wrapping up our discussion on the "Jamming-Resilient OTFS-NOMA", we’ve covered the main points from the paper. The authors are Michel Kulhandjian, Hovannes Kulhandjian, and Theodoros A. Tsiftsis.

Dev: Their work addresses how to make NOMA systems more robust against intentional jamming by proposing this sparse delay-Doppler NOMA scheme with randomized active sets and super increasing power allocation.

Taro: The implication is that for real world deployment, especially when dealing with interference, we should look at this architecture for mitigating those jamming effects.

Rosa: For the listener, it means that if you're listening to a broadcast or a radio transmission and someone tries to interfere with it deliberately, this system keeps the performance much better than what conventional methods allow.

Dev: They’re saying that even when things get difficult, the performance stays within predictable bounds because of their analytical proofs regarding conditioning and power allocation.

Taro: It moves away from just focusing on a single user's anti jamming tricks to a more robust, multi user approach that handles interference gracefully.

Rosa: They are showing that this system is right for situations where you have uncertainty about the channel state information, not just for clean CSI scaling.

Dev: The performance validation shows that they can get significant gains over conventional OTFS NOMA under oracle jamming, roughly twenty four decibels at signal to noise ratio of thirty five decibels.

Taro: So the main thing is moving towards a system that recovers gracefully when the world misbehaves, rather than just surviving it.

Rice University · California State University Fresno · University of Thessaly

eess.SP

Submitted: 2026-06-08

Updated: 2026-11-14

Comments: 30 pages, 16 figures. Master version. Journal companion: arXiv:TBD. WCL companion: arXiv:TBD

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 83/100

The gist: We propose a sparse delay-Doppler NOMA scheme that is resilient to intentional jamming The closed-form BER expression derived from this conditioning argument has no jammer-induced floor…

Key concepts

Sparse Delay-Doppler NOMA
This is a specific system architecture where data is placed sparsely across the delay-Doppler grid. This sparsity, combined with other design choices, makes the system robust against jamming and allows for efficient multi-user transmission in an OTFS framework.
Superincreasing Power Allocation (PA)
This method of assigning power to users uses a Merkle–Hellman knapsack construction. It ensures that the low-complexity successive interference cancellation (SIC) decoder performs exactly like a maximum-likelihood detector on the composite constellation, overcoming limitations in SIC propagation.
C5 Protocol
This protocol involves both the transmitter and receiver independently generating active bin sets using a shared secret pseudo-random seed. This randomization defends against pattern-aware jammers by ensuring that even if the jammer knows one set, the random nature of other sets keeps performance stable.
Marchenko–Pastur (M-P) Conditioning Bound
This mathematical bound controls how well a random unitary submatrix is conditioned. Crucially, this property ensures that the resulting BER expression has no jammer-induced floor, unlike traditional systems where such floors often appear.

Terminology

Summary

We propose a sparse delay-Doppler NOMA scheme that is resilient to intentional jamming The closed-form BER expression derived from this conditioning argument has no jammer-induced floor, in contrast to the well-known partial-band error floor of conventional OTFS-NOMA This architecture delivers at least a 40 dB BER-ratio improvement against pattern-aware jammers and reaches a floor BER at eight users by SNR around 20 dB

System Architecture and Design Ingredients

The proposed architecture is built upon a five-ingredient stack tightly co-designed for the jamming-resilient multi-user setting. This stack includes sparse data placement on the DD grid, unitary spreading across all Nb bins, per-frame randomization of the active bin subset via a shared pseudo-random seed (the “C5” protocol), superincreasing power allocation (PA) that makes the O(K) successive interference cancellation (SIC) decoder maximum-likelihood (ML)-optimal on the NOMA composite constellation, and disjoint pairwise clustering when more than four users are present. p. 0 The receiver threshold-excises jammer-contaminated bins, recovers the sparse signal by least squares, and decodes per-active-bin SIC. p. 0

How the Gains Come From

The key gains stem from several analytical properties that overcome conventional limitations. These include the use of a Marchenko–Pastur (M-P) conditioning bound, which controls the conditioning of any random unitary submatrix and has no jammer-induced floor. p. 0 The superincreasing power allocation uses a Merkle–Hellman knapsack construction to prove that the resulting low-complexity SIC matches maximum-likelihood detection exactly on the composite constellation, eliminating the usual SIC-propagation ceiling. p. 0 Furthermore, for more than four users, partitioning them into pairs and assigning each pair its own disjoint bin subset achieves floor BER at eight users by SNR around 20 dB. p. 0

Resilience to Adversarial Threats

The system exhibits defense in depth against pattern-aware adversaries through the C5 protocol. This protocol involves the transmitter and legitimate receiver independently generating the active set At via a pseudo-random sequence seeded by a shared secret. p. 2 Crucially, even if the seed is compromised and the jammer becomes omniscient about the per-frame active set, random unitary submatrices remain well-conditioned with high probability, so the BER stays within the unjammed-case envelope. p. 0 The Sylvester replication trap affects algebraically structured active patterns, motivating the randomized-pattern recommendation on benign-channel grounds alone. p. 2

Optimal Scaling and Cluster Design

For more than four users, the OMA-friendly NOMA rule reaches floor BER at eight users by SNR around 20 dB. This is achieved by partitioning the users into disjoint clusters of size Kg = 2 and assigning each cluster its private bin subset. p. 9 The within-cluster Bob fraction α(g) is monotonically decreasing in Kg, and the optimal cluster size is found to be Kg = 2 with flavor (β). p. 9 This OMA-friendly NOMA architecture synthesizes a jammer-resilient system from Ktot/2 parallel 2-user NOMAs on orthogonal DD subsets. p. 1

Performance Validation and Robustness

Monte Carlo simulations track the analytical predictions within 3 dB and indicate at least a 40 dB BER-ratio improvement against pattern-aware jammers, with roughly 24 dB of cumulative gain over conventional OTFSNOMA under oracle jamming. The analysis extends to Rician fading, showing that the jammer-independence property persists for arbitrary Rician K-factor. p. 0 The structural LS-conditioning floor is Kr-invariant and independent of channel statistics, jamming fraction, and jammer power at finite Nb. p. 2 The recommended multi-cluster recipe of Section VI is the right architecture for any deployment with CSI uncertainty, not just for asymptotic clean-CSI scaling. The cumulative gain is approximately 24 dB over conventional OTFS-NOMA SIC under oracle jamming at SNR= 35 dB. p. 0

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Robust communication systems can be designed to operate reliably in adversarial environments by implementing a sparse delay-Doppler NOMA scheme that is resilient to intentional jamming. This involves placing user data on a small random subset of delay-Doppler bins and spreading the result over all bins through a unitary precoder.

  2. AI systems can utilize a joint excision, least-squares, and successive interference cancellation receiver that includes threshold detection to excise jammed bins before recovering the sparse signal via least squares followed by per-bin SIC. This architecture achieves low complexity while maintaining performance against jamming.

  3. Communication systems can incorporate a randomized active set protocol using a shared pseudo-random seed to generate the active subset per frame, providing defense in depth against pattern-aware adversaries even if the shared seed is compromised.

  4. AI systems can employ superincreasing power allocation derived from Merkle–Hellman knapsack constructions to ensure that successive interference cancellation (SIC) achieves maximum likelihood detection exactly on the composite constellation, eliminating the usual SIC-propagation ceiling for more than two users.

  5. Communication systems can scale to over four users by partitioning them into pairs and assigning each pair its own disjoint bin subset, enabling OMA-friendly NOMA rules that reach floor bit error rate at eight users by a specific signal-to-noise ratio.

  6. AI systems can utilize a closed-form analytical BER framework based on Marchenko–Pastur conditioning to characterize the operating region where conventional OTFS-NOMA fails due to jammer interference, proving no jammer-induced error floor exists under this scheme.

  7. AI systems can incorporate a sparsity loading operating region design rule that translates the abstract Marchenko–Pastur bound into a closed-form design rule for maximum permissible sparsity ratio based on jamming fraction and noise tolerance.

Sources

Related papers