Uncertainty-aware user modeling is crucial for designing AI systems that adapt to users in real-time while addressing privacy concerns. This paper proposes a novel framework for privacy-preserving probabilistic user modeling that integrates uncertainty quantification and differential privacy (DP). Building on neural processes (NPs), a scalable latent variable probabilistic model, we enable meta-learning for user behaviour prediction under privacy constraints. By employing differentially private stochastic gradient descent (DP-SGD), our method achieves rigorous privacy guarantees while preserving predictive accuracy. Unlike prior work, which primarily addresses privacy-preserving learning for convex or smooth functions, we establish theoretical guarantees for non-convex objectives, focusing on the utility-privacy trade-offs inherent in uncertainty-aware models. Through extensive experiments, we demonstrate that our approach achieves competitive accuracy under stringent privacy budgets. Our results showcase the potential of privacy-preserving probabilistic user models to enable trustworthy AI systems in real-world interactive applications.

This paper investigates the role of dimensionality reduction in efficient communication and differential privacy (DP) of the local datasets at the remote users for over-the-air computation (AirComp)-based federated learning (FL) model. More precisely, we consider the FL setting in which clients are prompted to train a machine learning model by simultaneous channel-aware and limited communications with a parameter server (PS) over a Gaussian multiple-access channel (GMAC), so that transmissions sum coherently at the PS globally aware of the channel coefficients. For this setting, an algorithm is proposed based on applying (i) federated stochastic gradient descent (FedSGD) for training the minimum of a given loss function based on the local gradients, (ii) Johnson-Lindenstrauss (JL) random projection for reducing the dimension of the local updates and (iii) artificial noise to further aid user's privacy. For this scheme, our results show that the local DP (LDP) performance is mainly improved due to injecting noise of greater variance on each dimension while keeping the sensitivity of the projected vectors unchanged. This is while the convergence rate is slowed down compared to the case without dimensionality reduction. As the performance outweighs for the slower convergence, the trade-off between privacy and convergence is higher but is shown to lessen in high-dimensional regime yielding almost the same trade-off with much less communication cost.

The two-user discrete memoryless cognitive multiple-access channel with confidential messages (DM-CMAC-CM) has been studied in the literature with a result for the secrecy capacity region. In this paper, we study a more general and practical model (three-user DM-CMAC-CM), although similar generalizations have been intractable in some studied cases, and obtain a capacity-equivocation and secrecy capacity region results including previous results as its special cases.

In this paper, we study the two transmitter multiple-access channel (MAC) with noncausal side information at one (single-sided) or both (two-sided) of the encoders in the presence of an external wiretapper (WT) and under individual and collective secrecy constraints. Our work consists of four parts. First, we consider the discrete memoryless (DM-MAC-WT) case of the model and derive inner and outer bounds on the secrecy capacity region under both of the secrecy constraints. Second, the results are extended to the Gaussian memoryless case [Gaussian multiple-access channel WT (GMAC-WT)] in detail and are used to illustrate numerically that the secrecy rate region is enlarged by increasing the power of the interference signal which is known noncausally at the encoder. Third, we analyze the strong interference regime as an important regime in GMAC with side information and show that under collective secrecy constraint, we can achieve the secrecy capacity region of the single-sided channel asymptotically for a special case in which the signal and noise powers satisfy specific constraints. To the best of our knowledge, this capacity region is the first one obtained for GMAC-WT. Finally, we show that our results include previous works as its special cases for both the single-sided and two-sided channel models.

In this paper, we propose information theoretic schemes that achieve the rate-equivocation regions for the multiple-access relay channel with confidential messages. This channel consists of two users that play the role of eavesdropper for each other's message and a relay that helps them to have a reliable and secure communication. The relay can itself be an eavesdropper of the users messages. The first coding scheme used for achievability is noise forwarding strategy and the second is the combination of noise forwarding with the compress and forward strategy. In the first one, relay cooperates with the users to have a confidential communication by sending codewords independent of the secret messages. In the second one, in addition to sending independent codewords, relay also sends compressed version of the received signals. Using these schemes, we derive the inner bounds for the capacity-equivocation region of the described model.