Human-Layer Risks

Signal Spam & Noise Injection

Attack: Malicious actors submit low-quality or random signals to dilute the signal pool and degrade AI inference.

Mitigations:

• Staking requirement in $AXORA for signal submission

• ZK-weighted reputation penalizes consistently low-impact signals

• Signal orthogonality scoring (crowd-echo signals lose weight)

• AI-driven anomaly detection for entropy spikes

Residual Risk: Low — spam is economically irrational over time.

Coordinated Signal Manipulation

Attack: Groups collude to submit correlated signals to bias strategy direction.

Mitigations:

• Correlation clustering detection

• Penalization of highly synchronized signal cohorts

• Weight decay on overrepresented viewpoints

• No guarantee that any signal leads to execution

Key Insight: Because contributors cannot observe outcomes or each other, coordination is fragile and expensive.

Insider Signal Poisoning

Attack: Sophisticated insiders submit misleading signals to profit externally (e.g., front-running AxoraAI).

Mitigations:

• Signals are non-executable and low-dimensional

• AI aggregates across many contributors

• Execution timing is unknown and variable

• No direct mapping between signal and trade

Residual Risk: Moderate but bounded—insiders cannot deterministically control outcomes.