Federated learning (FL) has emerged as a privacy-preserving alternative to traditional machine learning, enabling collaborative model training without centralizing users’ data. While this decentralized approach enhances privacy, it introduces new trust assumptions. Specifically, that users honestly participate in the protocol. Simultaneously, the growing emphasis on fairness in machine learning has led to the development of fairness-aware FL schemes, which aim to ensure equitable model performance across diverse user groups. However, existing solutions for fair FL schemes rely on users truthfully reporting fairness-related metrics. This trust assumption opens the door to malicious behavior: users may manipulate these statistics to influence the global model, degrading its fairness, and/or its overall accuracy. Current frameworks lack mechanisms to detect or prevent such adversarial manipulation. This paper addresses this critical gap through two main contributions. First, as an example to illustrate our point, we empirically demonstrate the vulnerability of a representative fairness-aware FL framework to targeted attacks that exploit unverified fairness computations. Second, and most importantly, we propose a novel scheme (that can be integrated into existing fair FL schemes) that augments fairness-aware FL with verifiability. Our solution enables the detection of dishonest participants without compromising user privacy, thus strengthening the robustness of fairness-aware FL schemes.

In many machine learning scenarios, training occurs outside the control of the model sponsor or the entity using the model. A growing concern in such settings revolves around model poisoning and data poisoning-how training is conducted and which data contributes to the process. This paper introduces a protective scheme against model and data poisoning attacks. Leveraging cryptographic primitives such as hashes, signature schemes, and zero-knowledge proofs, the scheme ensures the integrity of the training process. Hashing maintains the continuity of data from authenticated sensors, while signatures validate the data. In the end, zero-knowledge proofs verify the correct model computation by the entity carrying out the training process. By adopting this approach, model sponsors can securely delegate training tasks, guaranteeing the authenticity of the results. Implementation and testing demonstrate the scheme's feasibility, effectively countering data and model poisoning threats.

Abstract We have recently introduced a privacy model for statistical and machine learning models called integral privacy. A model extracted from a database or, in general, the output of a function satisfies integral privacy when the number of generators of this model is sufficiently large and diverse. In this paper we show how the maximal c-consensus meets problem can be used to study the databases that generate an integrally private solution. We also introduce a definition of integral privacy based on minimal sets in terms of this maximal c-consensus meets problem.

The Lightning Network is a second layer technology running on top of Bitcoin and other Blockchains. It is composed of a peer-to-peer network, used to transfer raw information data. Some of the links in the peer-to-peer network are identified as payment channels, used to conduct payments between two Lightning Network clients (i.e., the two nodes of the channel). Payment channels are created with a fixed credit amount, the channel capacity. The channel capacity, together with the IP address of the nodes, is published to allow a routing algorithm to find an existing path between two nodes that do not have a direct payment channel. However, to preserve users' privacy, the precise balance of the pair of nodes of a given channel (i.e. the bandwidth of the channel in each direction), is kept secret. Since balances are not announced, second-layer nodes probe routes iteratively, until they find a successful route to the destination for the amount required, if any. This feature makes the routing discovery protocol less efficient but preserves the privacy of channel balances. In this paper, we present an attack to disclose the balance of a channel in the Lightning Network. Our attack is based on performing multiple payments ensuring that none of them is finalized, minimizing the economical cost of the attack. We present experimental results that validate our claims, and countermeasures to handle the attack.

Graphs can be used to model many different types of interaction networks, for example, online social networks or animal transport networks. Several algorithms have thus been introduced to build graphs according to some predefined conditions. In this paper, we present an algorithm that generates spatial graphs with a given degree sequence. In spatial graphs, nodes are located in a space equiped with a metric. Our goal is to define a graph in such a way that the nodes and edges are positioned according to an underlying metric. More particularly, we have constructed a greedy algorithm that generates nodes proportional to an underlying probability distribution from the spatial structure, and then generates edges inversely proportional to the Euclidean distance between nodes. The algorithm first generates a graph that can be a multigraph, and then corrects multiedges. Our motivation is in data privacy for social networks, where a key problem is the ability to build synthetic graphs. These graphs need to satisfy a set of required properties (e.g., the degrees of the nodes) but also be realistic, and thus, nodes (individuals) should be located according to a spatial structure and connections should be added taking into account nearness.

In this paper we analyze the applicability of onion routing on Predictable Delay-Tolerant Networks. We show that a simple approach for providing anonymity through the application of a simple onion routing can be feasible in some scenarios. To that end we evaluate our proposal on a network based on a city public bus transportation network.

Quick Response (QR) codes, used to store machine readable information, have become very commonnowa days and have found many applications in different scenarios. One of such applications is electronic voting systems. Indeed, some electronic voting systems are starting to take advantage of these codes, e.g. to hold the ballots used to vote, or even as a proof of the voting process. Nevertheless, QR codes are susceptible to steganographic techniques to hide information. This steganographic capability enables a covert channel that in electronic voting systems can suppose an important threat. A misbehaving equipment (e.g. infected with malware) can introduce hidden information in the QR code with the aim of breaking voters' privacy or enabling coercion and vote-selling. This paper shows a method for hiding data inside QR codes and an implementation of a QR writer/reader application with steganographic capabilities. The paper analyses different possible attacks to electronic voting systems that leverage the steganographic properties of the QR codes. Finally, it proposes some solutions to detect the mentioned attacks.

A decentralized authorization mechanism is proposed that uses Bloom filters to implement authorization delegation. This lightweight mechanism is unlike conventional approaches that typically rely on public key certificates to implement distributed delegation. In taking an approach based on one-way hash functions, the mechanism may be preferable for use in computationally constrained environments where public-key cryptography is not desirable.

Public key authorization credentials provide a flexible approach to implementing access control in open distributed systems. Wireless sensor networks, are examples of such systems; however, their low-power sensors have energy efficiency requirements that may mean it is not practical to carry out computationally intensive operations, such as public key operations. This paper describes a distributed access control system for a Wireless Sensor Network application that uses computationally efficient one-way hash-functions to implement authorization credentials.

Abstract Mobile agents (MAs) are autonomous computing entities that dwell in agent platforms and have the ability to move to different locations as needed. They are typically composed of code, data, and a state. The performance of MA migrations is always penalized by the need of carrying these components to each visited location. In contrast to the agent data and state, the agent code is static during the whole agent life. Therefore, optimizing the agent code management, the agent migration performance can be improved. The main contribution of this article is the definition of a global cache service to efficiently and securely deal with the distribution of agent code. An implementation of the service has been developed, and its benefits have been extensively demonstrated by a set of performance tests carried out in different scenarios. Copyright © 2010 John Wiley & Sons, Ltd.