Security and Privacy Issues and Solutions in Federated Learning for Digital Healthcare

Hyejun Jeong, Tai-Myoung Chung

Abstract

The advent of Federated Learning (FL) has enabled the creation of high-performing models as if they had been trained on a considerable amount of data. A multitude of participants and a server cooperatively train a model without the need for data disclosure or collection. The healthcare industry, where security and privacy are paramount, can substantially benefit from this new learning paradigm, as data collection is no longer feasible due to stringent data policies. Nonetheless, unaddressed challenges and insufficient attack mitigation are hampering its adoption. Attack surfaces differ from traditional centralized learning in that the server and clients communicate between each round of training. In this paper, we present vulnerabilities, attacks, and defenses based on the widened attack surfaces, as well as suggest promising new research directions toward a more robust FL.

Introduction: The Promise and Peril of FL in Healthcare

Federated Learning (FL) is often hailed as the "holy grail" for digital healthcare. It promises the ability to train powerful AI models on sensitive patient data without that data ever leaving the hospital’s local servers. Traditional AI requires "centralization": gathering all patient records into one giant database, which in the era of GDPR and HIPAA is a legal and ethical nightmare. FL solves this by bringing the model to the data rather than the data to the model.

The catch? Because FL relies on communication between a central server and many distributed clients (hospitals, clinics, or even wearable devices), it introduces a significantly widened "attack surface." An attacker no longer needs to breach a central database; they can corrupt the global model by pretending to be a hospital or intercepting updates sent over the network.

A Systematic Taxonomy of FL Security

To organize the complex landscape of FL security research, the authors developed a systematic taxonomy categorized into three pillars: sources of vulnerabilities, specific attacks, and mitigation techniques.

Taxonomy of Federated Learning security: Vulnerabilities, Attacks, and Mitigations.
  1. Sources of Vulnerabilities: Identifying who or what is at risk (Compromisable Clients, Compromisable Servers, Outsiders, and the Distributed Nature of the system).
  2. Attacks and Threats: Categorizing how systems are compromised (Poisoning, Inference, Backdoors, etc.).
  3. Mitigation Techniques: Evaluating current defenses (Sanitization, Perturbation, Cryptography, and Personalization).

Threat Landscape and Severity

Not all threats are created equal. The authors highlight the severity of different attacks, ranking Poisoning, Inference, and Backdoor attacks as the highest risks to healthcare systems due to their potential to compromise patient privacy or model integrity.

Severity levels and descriptions of common threats in Federated Learning.

Key Insights into Healthcare Attacks

  • Model Poisoning > Data Poisoning: "Model Poisoning" (where a client directly alters the model parameters) is far more effective and stealthy than "Data Poisoning" (simply changing the training labels). In FL, a single malicious client can achieve 100% targeted misclassification while remaining undetected.
  • The "Honest-but-Curious" Server: One of the biggest threats is the central aggregator itself. Even if the server follows the rules, it can use "Inference Attacks" to reconstruct private patient data just by looking at the mathematical updates sent by hospitals.
  • The Non-IID Defense Dilemma: Healthcare data is naturally "Non-IID" (Non-Independent and Identically Distributed)—meaning a dermatology clinic's data looks very different from a cardiology clinic's. Most current anomaly detection systems struggle to tell the difference between a "malicious" update and a "unique but legitimate" healthcare update.
  • Backdoors as "Silent Killers": Backdoor attacks allow a model to function perfectly on normal cases but fail (or misdiagnose) when a specific "trigger" is present. These are incredibly difficult to detect because they don't degrade the overall accuracy of the model.

Mitigation Strategies and Their Trade-offs

The paper synthesizes the current state of defenses, noting that most involve significant trade-offs between security and performance:

  • Differential Privacy (DP): Effective at stopping data leakage but often degrades model accuracy—a dangerous trade-off in clinical medicine.
  • Robust Aggregation: Techniques like Krum or Bulyan help filter out malicious updates, but recent research shows they can be bypassed by sophisticated attackers.
  • Homomorphic Encryption (HE): Allows the server to compute updates without seeing them, but it is computationally expensive, though newer methods like BatchCrypt are improving efficiency.

Future Research Directions

The authors identify several "unmet needs" that the research community must address to make FL truly "healthcare-ready":

  1. Standardization: We need standardized preprocessing pipelines to ensure models don't deviate simply because of formatting issues across different hospitals.
  2. Multi-modal Data: Research needs to move beyond simple image classification (like MNIST or CIFAR-10) and test defenses on complex healthcare data, including Electronic Health Records (EHR) and tabular data.
  3. Incentive Mechanisms: Systems must reward high-quality, honest participation to encourage large institutions to share their valuable data.
  4. United Defense: Future work needs to find "orthogonal" techniques that balance privacy, security, and performance rather than solving one problem while creating another (e.g., encryption causing communication bottlenecks).

Summary updated

Abstract

The advent of Federated Learning has enabled the creation of a high-performing model as if it had been trained on a considerable amount of data. A multitude of participants and a server cooperatively train a model without the need for data disclosure or collection. The healthcare industry, where security and privacy are paramount, can substantially benefit from this new learning paradigm, as data collection is no longer feasible due to stringent data policies. Nonetheless, unaddressed challenges and insufficient attack mitigation are hampering its adoption. Attack surfaces differ from traditional centralized learning in that the server and clients communicate between each round of training. In this paper, we thus present vulnerabilities, attacks, and defenses based on the widened attack surfaces, as well as suggest promising new research directions toward a more robust FL.
BibTeX
@inproceedings{jeong2022security,
  author = {Hyejun Jeong and Tai-Myoung Chung},
  title = {{Security and Privacy Issues and Solutions in Federated Learning for Digital Healthcare}},
  booktitle = {FDSE (CCIS Volume)},
  year = {2022},
  url = {https://openreview.net/forum?id=9sgWm2fMKeo},
}
Figure