Skip to content

QIF-T0021

medium

Federated gradient leakage

Tier 4 — Demonstrated (Case Study / Observational)

Legacy status: EMERGING

Reconstruct individual neural data from shared gradients in federated BCI training. Multi-site clinical trials share model updates. Gradient inversion recovers raw EEG with high fidelity.

Technique Details

Tactic
QIF-M.SV
Status
EMERGING
Bands
S2, S3

Therapeutic Application

Reconstruction of individual neural data from shared gradient updates in federated BCI training

Clinical Analog

Federated learning for multi-site clinical BCI trials

Treats

  • multi-center BCI research
  • collaborative model training without data sharing

Neural Impact

2 of 7 neural bands affected

S2 S3

Drag to rotate. Click a region to learn more.

Click or hover over a glowing region to see the attack techniques targeting it and their severity.

Scoring

NISS v1.1 NISS:1.1/BI:N/CR:L/CD:L/CV:I/RV:F/NP:N
CVSS v4.0 CVSS:4.0/AV:N/AC:H/AT:P/PR:L/UI:N/VC:H/VI:N/VA:N/SC:H/SI:N/SA:N
2.7Low
BICRCDCVRVNP
 

Governance

Neurorights at Risk

This technique threatens 1 of the 4 proposed neurorights (Ienca & Andorno, 2017).

Consent Complexity
0.20 / 4.0

FDORA §3305 Compliance

Cyber Device
Regulatory Coverage
0.4 / 1.0
524B Requirements
TM VA PM
Regulatory Gaps
  • ! CVSS partially captures risk; neural dimensions missing
  • ! No FDA pathway for consumer sensor exploitation

Population Vulnerability

CRB vulnerability adjustment (γ=0.30) accounts for age, diagnosis severity, consent capacity, and device dependency.

Population NISS Base Adjusted Severity Delta
Adult (Default) 2.7 2.7 Low -
Child (10yr) + ADHD 2.7 3.2 Low +0.48
Adult with ALS 2.7 3.1 Low +0.44

Validation Status

Theoretical / Not yet validated. This technique has not been independently tested. See the validation dashboard for what has been tested.

Qinnovate Neural Security Atlas Edit this on GitHub