Covert Event Communication Protocols

• Covert event communication protocols are methods that embed signals in timing, coding, and environmental noise to conceal both the message and its occurrence.
• They employ techniques like chunked coding, timing perturbation, and protocol channel selection to optimize the trade-off between throughput, delay, and covertness.
• Analytical metrics and experimental benchmarks validate these protocols, ensuring that embedded communications remain statistically indistinguishable from normal network traffic.

Covert event communication protocols are mechanisms that enable entities to transmit event reports or other data in such a way that the very act of communication remains undetectable by an observing adversary, often referred to as a "warden." These protocols exploit network, system, or physical-layer resources—such as timing, coding, behavior selection, or environmental noise—to conceal both the message content and the occurrence of the communication event itself. Research into these protocols is motivated by applications in adversarial, surveilled, or resource-constrained environments, including IoT sensor networks, battlefield networks, critical infrastructure, and multi-agent systems.

1. Architectural Models and System Settings

Covert event communication design is fundamentally shaped by the underlying system architecture, threat model, and mobility or interaction patterns of the communicating parties.

• Random Graph Mobility: In mobile IoT settings, agents traverse a random $d$-regular graph representing physical space or logical connectivity, with dedicated relay nodes supporting message storage and forwarding. Communication is modeled by agents (Alice and Bob) performing independent random walks, while a warden (Willie) surveils nodes via random walks or uniform monitoring, with the constraint that only a single node can be observed at any instant (Peng et al., 2021).
• Agent-Driven Event Systems: In event-driven agent environments (Internet of Agents), events are formalized as tuples $$(\text{agent}, \text{action type}, \text{payload}, t)$$, and covert information is embedded in the storage, timing, or behavioral choice dimensions of event streams. A unified model specifies the public environment, the natural event distribution ${\cal D}$, and explicit security notions of indistinguishability under both statistical and intention-based adversaries (Huang et al., 4 Aug 2025).
• Network and Physical Channels: Protocols also leverage features of wireless networks (distributed artificial noise, path-loss, and AWGN channels), communication protocols (protocol header selection, timing of protocol events), or physical-layer constraints (spectral mask compliance, use of environmental noise as cover) to shape the achievable rate and detectability trade-offs (Soltani et al., 2016, Qiaosheng et al., 2020, Liu et al., 2017, Bali et al., 2 Jun 2025).

2. Core Methodologies and Protocol Constructions

Protocol designs utilize a range of techniques to conceal the presence of event communication:

• Chunked Coding with Redundant Relays: Event reports are partitioned into $k$-length blocks, expanded via an $[n,k]$ MDS code, and deposited passively at $n$ randomly encountered relays. The recipient collects any $k$ coded chunks for full recovery, while redundancy reduces expected delay at the cost of more detectable deposit/retrieve actions. Upload and download operations are synchronized to agent mobility to avoid coincidence with a surveilling warden. Covertness is defined as the probability that all $n+k$ transmissions escape detection (Peng et al., 2021).
• Behavioral, Timing, and Storage Channels: In multi-agent systems, covert channels are realized simultaneously in three dimensions:
  • Storage embedding into event payloads, using generative steganography.
  • Timing perturbation within plausible statistics of event occurrence.
  • Behavioral channel via secret-key–driven selection of plausible, context-conformant actions.
  • The $\Pi$CCAP protocol formalizes a three-phase handshake–embed–reassemble pipeline, with LLM-based action selection and key-driven header construction. Security is guaranteed by indistinguishability to statistical or semantic adversaries (Huang et al., 4 Aug 2025).
• Protocol Channel Techniques: Covert bits are encoded by selecting, per event, one out of a set of naturally occurring transport protocol types (ICMP, ARP, etc.), assigning symbols to protocols. This allows storage channels that modulate fields without payload changes, maximizing stealth when packet distributions match network norms (0809.1949).
• History-Based Pointer Channels: The SHP protocol encodes information indirectly, signaling the receiver to re-interpret a naturally occurring sequence of external network events by providing small pointers to the 'history' of observed network timing. Cryptographic hashing and key agreement amplify the covert payload, yielding "channel amplification factors" far exceeding the bitlength of the active signal. Detection by statistical or ML means is demonstrably close to chance (Weissenborn et al., 27 Nov 2025).
• Physical Layer and Wireless Designs: In wireless physical layers, single-hop or multi-hop transmission is concealed by leveraging friendly jammers/noise sources, maximizing covertness subject to the square-root law: no more than $O(\sqrt{n})$ bits over $$(\text{agent}, \text{action type}, \text{payload}, t)$$0 channel uses can be sent undetectably unless additional structure (timing, channel state) is present (Soltani et al., 2016, Qiaosheng et al., 2020, Bali et al., 2 Jun 2025).
• Observation-Based Payload Formatting: Protocol Proxy systems embed TCP or higher-layer sessions into UDP-based static protocols by format-transforming encryption, mapping encrypted payloads into observed protocol field values, and deterministic HMM modeling of interpacket times to defeat payload- and timing-based analysis. This approach is particularly useful where blocking the host protocol would create unacceptable collateral damage (Oakley et al., 2020).

3. Analytical Metrics and Trade-offs

The design of covert event communication protocols involves explicit trade-offs between throughput, delay, and covertness, measured via quantitative metrics:

• Covertness Probability ($$(\text{agent}, \text{action type}, \text{payload}, t)$$1): Defined as the probability all transmission events (chunk uploads/downloads, storage/timing perturbations) evade detection, often modeled as $$(\text{agent}, \text{action type}, \text{payload}, t)$$2, where $$(\text{agent}, \text{action type}, \text{payload}, t)$$3 is the per-chunk or per-event detection probability (Peng et al., 2021).
• Delay and Throughput: The expected time to successfully transmit and receive the necessary fragments or event signals is analyzed via coupon-collector arguments, random-walk hitting times, and timing channel capacity calculations. Coding redundancy can expedite message delivery but linearly increases the number of detectability-threatening actions (Peng et al., 2021). In timing channels or pointer channels, amplification factors (CAF) or bits-per-event metrics quantify the steganographic efficiency (Weissenborn et al., 27 Nov 2025).
• Fundamental Limits: The square-root law ($$(\text{agent}, \text{action type}, \text{payload}, t)$$4 covert bits per $$(\text{agent}, \text{action type}, \text{payload}, t)$$5 uses) governs general AWGN and many DMC covert channels, but timing channels on queues with sufficiently rich innocent traffic (e.g., Poisson arrivals) admit strictly positive covert rates if an appropriate secret key rate is available (Mukherjee et al., 2016). Exploiting channel state information (CSI) or adversary tracking enables further rate improvements under certain conditions (ZivariFard et al., 2020, Wang et al., 2023).
• Adversarial Detectability: The adversary's power is modeled via Neyman–Pearson hypothesis testing, likelihood-ratio tests, anomaly statistics (e.g., Kullback–Leibler divergence between observed and expected distributions), as well as statistical and ML steganalysis. Security requirements often demand that the adversary's aggregate detection error remains arbitrarily close to 0.5 (random guess), even under unbounded computational power in quantum-secured systems (Liu et al., 2017).

4. Security Definitions and Formal Guarantees

Research protocols rigorously define security via formal metrics:

• Indistinguishability (IND-STAT, IND-INT): Statistical imperceptibility requires that the advantage of any statistical adversary in distinguishing covert from benign event streams is negligible in the security parameter (e.g., $$(\text{agent}, \text{action type}, \text{payload}, t)$$6) (Huang et al., 4 Aug 2025). Intentional imperceptibility further requires that contextually plausible masquerade is preserved, as judged by advanced classifiers or human-in-the-loop scoring.
• Channel Resolvability: Many constructions rely on soft-covering lemmas and channel-resolvability results, ensuring that the warden's output distribution under covert use converges in variational distance (or KL divergence) to the output under innocent communication, establishing fundamental covertness (ZivariFard et al., 2020, Weissenborn et al., 27 Nov 2025).
• Quantum Adversary Bounds: In quantum-protected settings, provable covertness is shown by analyzing Eve's optimal hypothesis test, minimizing detection bias $$(\text{agent}, \text{action type}, \text{payload}, t)$$7 via controls on quantum relative entropy $$(\text{agent}, \text{action type}, \text{payload}, t)$$8, ensuring $$(\text{agent}, \text{action type}, \text{payload}, t)$$9 for ${\cal D}$0 modes (Liu et al., 2017).
• Key Requirements: Some protocols require pre-shared secret keys of rate sufficient to support key-based codebook randomization (timing channels), while others extract randomness from channel state information (CSI) via random binning and block-Markov schemes with provable uniformity with respect to the warden (ZivariFard et al., 2020, Mukherjee et al., 2016).

5. Performance Benchmarks and Experimental Results

Empirical studies and testbed deployments validate theoretical limits and provide guidance for protocol tuning:

• Capacity versus Covertness Frontier: For mobile IoT relay networks, protocols show that delay-minimizing and covertness-maximizing parameters cannot be satisfied simultaneously; the intersection of their admissible regions defines feasible operating points, with trade-offs governed by message size, redundancy, relay density, and surveillance model (Peng et al., 2021).
• Empirical Stealth Metrics: Statistical and ML detectors (e.g., BERT classifiers, LSTM on discrete timing) achieve near-chance accuracy on event streams produced by advanced covert protocols (AUC ${\cal D}$1 0.5) (Huang et al., 4 Aug 2025, Weissenborn et al., 27 Nov 2025), and ROC curves similarly fail to surpass random guessing in controlled adversarial evaluation.
• Wireless/SDR Experimentation: Real-world implementations with software-defined radios confirm the theoretical square-root scaling for covert throughput and show that detection errors at the warden remain at random-guess levels under precise calibration of pulse selection, coding rate, and noise floor (Bali et al., 2 Jun 2025).
• Optical Channel Demonstrations: In DWDM fiber with classical channel-induced Raman noise, covert data rates between ${\cal D}$2–${\cal D}$3 bits/s are achieved with error-free decoding and arbitrarily small detection bias, demonstrating feasibility in metropolitan networks (Liu et al., 2017).
• Protocol-Proxy Deployment: Throughput measurements of format-transforming, timing-mimetic proxies achieve ~182 bps of covert TCP over UDP sessions, with empirical p-values for protocol field and timing distributions far above significance thresholds, confirming indistinguishability in operational settings (Oakley et al., 2020).

6. Design Constraints, Limitations, and Countermeasures

• Trade-off Impossibility: Many protocols reveal an inherent impossibility of simultaneously maximizing throughput (or minimizing delay) and maximizing covertness within a given parameter regime, creating the need for adaptive trade-off navigation (Peng et al., 2021).
• Stealth Failure Modes: Risks include pattern leakage under nonstationary behavior drift, limited entropy in action or timing channels, or protocol desynchronization due to packet loss or network anomalies. Synchronization and reliability mechanisms—such as error correction, redundancy, and implicit acknowledgment—are often necessary to maintain performance.
• Defensive Strategies: Defenses against covert event channels include: enforcing protocol whitelists and rate-limits, sophisticated statistical anomaly detectors, ML-based multi-dimensional steganalysis, and network traffic normalization (e.g., adding jitter, reordering events), though at the expense of legitimate system performance (Huang et al., 4 Aug 2025, 0809.1949, Weissenborn et al., 27 Nov 2025).
• Collateral Damage Constraints: Some covert event protocols exploit protected static protocols (e.g., critical infrastructure UDP traffic) where indiscriminate blocking would incur severe operational cost (e.g., service outages), raising the detection cost for the adversary and favoring covert operation (Oakley et al., 2020).

7. Synthesis and Future Directions

Contemporary covert event communication is characterized by a blend of coding theory, channel modeling, steganography, behavioral mimicry, and rigorous security analysis. Key open challenges include the joint optimization of delay, throughput, and stealth under adaptive, possibly active adversary models; robust operation under concept drift and dynamic environments; and the generalization of formal security guarantees to multimodal and action-rich settings. The field continues to benefit from empirical validation across wireless, optical, IoT, and agent-based domains, with protocols increasingly engineered to match sophisticated adversarial and operational constraints (Peng et al., 2021, Huang et al., 4 Aug 2025, Weissenborn et al., 27 Nov 2025, Liu et al., 2017, Bali et al., 2 Jun 2025).