- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper studies global-model exposure in federated learning through proxy model distribution. CloakFL constructs client-specific perturbations using centered local updates from the previous round and scales them layer-wise according to the receiving client’s latest update magnitude. The goal is to degrade the distributed proxies while preserving the utility of server aggregation. The paper provides a convergence bound for a generalized perturbed FedAvg process under strong convexity, full participation, and additional perturbation assumptions. Experiments cover four vision and language tasks, cross-silo and cross-device settings, and single-client and bounded multi-client recovery. The reported results show larger server–proxy performance gaps than FedQSN at comparable server performance.
Critical strengths
1.The perturbation design uses a meaningful optimization signal. Constructing proxies from client drift, rather than arbitrary noise, provides a plausible way to degrade individual models while limiting disruption to aggregation. The direction and magnitude ablations offer empirical support for this approach.
2.The evaluation extends beyond raw proxy degradation. The paper includes vision and language architectures, different heterogeneity levels, and recovery through local training, fine-tuning, and coalition averaging. These comparisons are useful because a low-performing raw proxy alone does not establish protection against a client that can continue training it.
Critical weaknesses
1.The convergence assumptions are not established for the proposed algorithm. Assumption 3.3 requires the aggregated perturbation to be conditionally unbiased, and Appendix F uses this requirement to eliminate a cross term. Centering the original client drifts does not guarantee this property after source-direction normalization and client-specific layer-wise scaling. For example, with two clients having drifts (v) and (-v), the exclusion of self-assignment forces a swap. If their reference update norms are (a) and (b), the average perturbation is proportional to ((b-a)v), which is generally nonzero. The algorithm also does not establish the learning-rate-scaled moment bounds in Assumptions 3.4–3.5. Although the appendix acknowledges a theory–practice gap, the current theorem therefore provides a conditional result for a generalized perturbation process rather than a demonstrated convergence guarantee for Algorithm 1.
2.The recovery evaluation is too limited to substantiate the protection claim. Recovery fine-tuning uses only one local epoch at a fixed reduced learning rate, and the evaluation does not test stronger optimization or exploitation of models collected across rounds. Substantial recovery already occurs under this limited protocol: in the cross-device E2E setting with (\beta=0.1), CloakFL’s raw proxy BLEU is 18.5, but single-client recovery reaches 57.1 against a server score of 62.9. Furthermore, the reported “Max” is the maximum of attack-specific mean scores, and coalition results average ten sampled groups. These statistics do not capture whether one client or a favorable coalition can recover a valuable model, which is directly relevant to model leakage.
Questions for authors
1.Does Algorithm 1 satisfy conditional unbiasedness after layer-wise normalization, client-specific scaling, and exclusion of self-assignment? If so, please provide the argument. How are the moment bounds in Assumptions 3.4–3.5 established for stored historical updates?
2.Theorem 3.1 bounds the average initial proxy model (\bar w_T^{(0)}). What corresponding guarantee follows for the clean server model (w_T) returned by Algorithm 1?
-
Does CloakFL retain its advantage when recovery uses multiple epochs, learning-rate selection, and model histories from previous rounds, while retaining the same restriction on available task data?
-
What are the best-client and best-coalition recovery results, and the distribution of recovery performance, rather than only the maximum of attack-specific means?
5 comments
Response to Reviewer 59yL (2/5)
W2. Recovery fine-tuning uses only one local epoch at a fixed reduced learning rate, and the evaluation does not test stronger optimization or exploitation of models collected across rounds.
To address this, we evaluated nine conditions combining recovery training epochs (1, 3, 5) and learning rates (1%, 10%, 100% of the base learning rate of 0.1) in the cross-device setting on Tiny-ImageNet (β=0.1). Among these, E=5& LR=0.1 corresponds to the original local training (ProxyTrain), and E=1& LR=0.01 corresponds to the original fine-tuning (ProxyFT).
For FedAvg and FedQSN, we observed several cases where the recovered model outperformed the server model, whereas no such cases were observed for CloakFL. This difference is more pronounced in the Multi Recovery setting.
The Proxy and recovery entries report the maximum / minimum values across the evaluated combinations; a minimum above a server reference means that every evaluated combination exceeded that reference. Values exceeding the Server performance for the corresponding method are highlighted in bold.
Single-Client Recovery
| Method | Server | Proxy | E=1& LR=0.001 | E=1& LR=0.01 | E=1& LR=0.1 | E=3& LR=0.001 | E=3& LR=0.01 | E=3& LR=0.1 | E=5& LR=0.001 | E=5& LR=0.01 | E=5& LR=0.1 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| FedAvg | 35.70 | 35.70 / 35.70 | 36.61 / 35.55 | 36.37 / 32.65 | 13.61 / 2.56 | 36.84 / 35.80 | 32.93 / 27.58 | 15.84 / 5.53 | 36.78 / 35.68 | 30.61 / 25.08 | 14.87 / 3.87 |
| FedQSN | 29.87 | 27.60 / 23.20 | 31.49 / 29.58 | 31.70 / 26.20 | 8.95 / 1.67 | 31.50 / 30.54 | 27.40 / 20.92 | 10.23 / 2.94 | 31.89 / 29.87 | 24.12 / 18.15 | 10.06 / 2.08 |
| CloakFL | 29.54 | 13.03 / 5.07 | 14.84 / 8.41 | 18.79 / 11.95 | 10.85 / 3.28 | 15.83 / 9.55 | 20.26 / 13.21 | 11.65 / 3.93 | 16.73 / 10.54 | 19.15 / 12.83 | 11.42 / 5.10 |
Multi-Client Recovery
| Method | Server | Proxy | E=1& LR=0.001 | E=1& LR=0.01 | E=1& LR=0.1 | E=3& LR=0.001 | E=3& LR=0.01 | E=3& LR=0.1 | E=5& LR=0.001 | E=5& LR=0.01 | E=5& LR=0.1 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| FedAvg | 35.70 | 35.70 / 35.70 | 36.79 / 36.25 | 37.59 / 36.60 | 29.46 / 24.75 | 37.24 / 36.65 | 37.58 / 35.77 | 31.44 / 26.20 | 37.33 / 36.79 | 37.37 / 34.93 | 30.56 / 24.60 |
| FedQSN | 29.87 | 27.11 / 26.11 | 31.21 / 30.64 | 32.96 / 31.86 | 25.44 / 19.62 | 32.22 / 31.67 | 33.69 / 30.57 | 24.07 / 19.02 | 32.53 / 32.11 | 33.00 / 30.15 | 23.27 / 16.84 |
| CloakFL | 29.54 | 25.15 / 12.69 | 27.12 / 17.23 | 28.21 / 21.88 | 25.99 / 21.56 | 27.70 / 18.83 | 28.55 / 23.95 | 25.14 / 20.41 | 28.21 / 19.92 | 28.64 / 24.08 | 23.75 / 20.17 |
Response to Reviewer 59yL (3/5)
W2. Recovery fine-tuning uses only one local epoch at a fixed reduced learning rate, and the evaluation does not test stronger optimization or exploitation of models collected across rounds.
Across-Round Recovery
To evaluate recovery across multiple rounds, we evaluated across 3R, which selects one proxy from each of three consecutive rounds, and 30% across 3R, which selects 30% of the participating clients in each round, combining nine proxies in total.
Considering the performance improvements from model averaging across rounds discussed in WIMA [1] and IMA [2], we provided the average of the server models from the same three rounds, Server (across 3R), as a reference alongside the recovery results obtained by combining proxies from multiple rounds.
The recovery entries report the maximum / minimum values across the evaluated combinations.
| Method | Server | Server (across 3R) | Coll.Proxy (across 3R) | Coll.Proxy (30% across 3R) | Coll.Train (across 3R) | Coll.Train (30% across 3R) | Coll.FT (across 3R) | Coll.FT (30% across 3R) |
|---|---|---|---|---|---|---|---|---|
| FedAvg | 35.70 | 39.21 | 39.21 / 39.21 | 39.21 / 39.21 | 29.73 / 14.16 | 39.04 / 32.85 | 39.85 / 38.84 | 40.59 / 39.87 |
| FedQSN | 29.87 | 34.64 | 32.14 / 30.92 | 31.79 / 31.40 | 23.02 / 3.67 | 32.92 / 26.65 | 35.69 / 34.53 | 36.43 / 35.76 |
| CloakFL | 29.54 | 31.87 | 25.27 / 13.95 | 29.28 / 25.35 | 23.59 / 17.66 | 31.55 / 26.40 | 28.20 / 21.50 | 32.10 / 28.89 |
For FedAvg and FedQSN, the maximum recovery performance exceeded Server in several collusion settings. In contrast, for CloakFL, the maximum exceeded Server only in 30% across 3R.
In Coll.FT (30% across 3R), which uses the largest number of proxies, the maximum values for FedAvg and FedQSN exceeded their respective Server (across 3R) values by 1.38 and 1.79 percentage points, and their minimum values also exceeded the corresponding reference values, meaning that every evaluated combination exceeded them. In contrast, the maximum value for CloakFL exceeded the reference value by 0.23 percentage points, but its minimum value was below Server.
[1] Caldarola et al., “Window-Based Model Averaging Improves Generalization in Heterogeneous Federated Learning,” ICCV Workshops, 2023.
[2] Zhou et al., “Understanding and Improving Model Averaging in Federated Learning on Heterogeneous Data,” IEEE TMC, 2024.
Response to Reviewer 59yL (1/5)
We sincerely thank the reviewer for the constructive feedback. To address the concerns regarding the identified weaknesses, we provide the following clarifications together with additional experimental results.
W1. The convergence assumptions are not established for the proposed algorithm.
In CloakFL without scaling, the mean of the client drifts assigned in a round is exactly zero, so the mean of the aggregated perturbation is also exactly zero. Therefore, this configuration satisfies the conditional unbiasedness assumption of Theorem 3.1. In addition, because the average of the proxy models is exactly equal to the clean server model without scaling, the convergence bound in Theorem 3.1 also applies to the server model in this configuration.
On Tiny-ImageNet, server performance without scaling was similar to that with scaling, and the performance of both the proxies and the recovered models was below that of the server. Client-specific scaling was added to improve protection effectiveness in settings such as 20 Newsgroups and to control the trade-off between protection effectiveness and server performance. Exact cancellation is not guaranteed with scaling, but the aggregated perturbation remains close to zero, as shown in Figure 1(b) and the table below.
Tiny-ImageNet
20 Newsgroups
The entries in the table below report the maximum / minimum values across proxy models or recovery sets.
Tiny-ImageNet
20 Newsgroups
Therefore, Theorem 3.1 guarantees the convergence of the unscaled CloakFL configuration, which also maintained server performance and protection effectiveness in the experiments, and we confirmed in Figure 1(b) that the aggregated perturbation remains close to zero even with scaling.