CryptoKANs+: Can KAN Be Just an MLP? Towards a Fast and Accurate Privacy-Preserving Machine Learning Solution
Bibliographic record
Abstract
The proliferation of the Internet of Things (IoT), coupled with rapid advances in Neural Networks (NNs), has given rise to the Internet of Artificially Intelligent Things. While cloud-based solutions enable scalable intelligence, they also introduce serious privacy risks, particularly during NN inference on sensitive data. Homomorphic Encryption (HE) offers a promising solution by enabling computations directly on encrypted data without decryption. However, integrating NNs with HE remains challenging-especially in introducing nonlinearity-due to the limited support for non-linear operations in current word-wise HE schemes. Existing approaches typically rely on approximated Activation Functions (AFs), often at the cost of reduced accuracy or increased computational overhead. In this paper, we compare Kolmogorov-Arnold Networks (KANs) with their counterparts, the Multi-Layer Perceptrons (MLPs), in privacy-preserving settings. We first formally demonstrate that KANs are equivalent to MLPs without the initial fully connected linear transformation layer when using the same AFs and transformations. Then, we propose CryptoKAN+, which integrates a learnable quadratic AF applied in the form of a Fully Connected Quadratic Transformation (FCQT) layer. In this design, the subsequent transformation is absorbed into the quadratic AF, eliminating the need for a post-linear transformation. Extensive experiments on the MNIST and Fashion-MNIST datasets reveal that CryptoKAN+ consistently outperforms stateof-the-art privacy-preserving machine learning (PPML) solutions in both accuracy and efficiency. These findings underscore the potential of KAN-based architectures with quadratic AFs as FCQT layers, making CryptoKAN+ a compelling choice for real-world applications requiring secure, accurate, and efficient inference.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.004 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.002 |
| Scholarly communication | 0.001 | 0.003 |
| Open science | 0.001 | 0.003 |
| Research integrity | 0.001 | 0.003 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".