Federated Learning (FL) is an appealing method for applying machine learning to large scale systems due to the privacy and efficiency advantages that its training mechanism provides. One important field for FL deployment is emerging IoT applications. In particular, FL has been recently used for IoT intrusion detection systems where clients, e.g., a home security gateway, monitors traffic data generated by IoT devices in its network, trains a local intrusion detection model, and send this model to a central entity, the aggregator, who then computes a global model (using the models of all gateways) that is distributed back to clients. This approach protects the privacy of users as it does not require local clients to share their potentially private IoT data with any other parties, and it is in general more efficient than a centralized system. However, FL schemes have been subject to poising attacks, in particular to backdoor attacks.In this paper, we show that FL-based IoT intrusion detection systems are vulnerable to backdoor attacks. We present a novel data poisoning attack that allows an adversary to implant a backdoor into the aggregated detection model to incorrectly classify malicious traffic as benign. We show that the adversary can gradually poison the detection model by only using compromised IoT devices (and not gateways/clients) to inject small amounts of malicious data into the training process and remain undetected. Our extensive evaluation on three real-world IoT datasets generated from 46 IoT devices shows the effectiveness of our attack in injecting backdoors and circumventing state of the art defenses against FL poisoning. Finally, we discuss shortly possible mitigation approaches.

Federated learning (FL) is an emerging distributed machine learning paradigm which addresses critical data privacy issues in machine learning by enabling clients, using an aggregation server (aggregator), to jointly train a global model without revealing their training data. Thereby, it improves not only privacy but is also efficient as it uses the computation power and data of potentially millions of clients for training in parallel.However, FL is vulnerable to so-called inference attacks by malicious aggregators which can infer information about clients' data from their model updates. Secure aggregation restricts the central aggregator to only learn the summation or average of the updates of clients. Unfortunately, existing protocols for secure aggregation for FL suffer from high communication, computation, and many communication rounds.In this work, we present SAFELearn, a generic design for efficient private FL systems that protects against inference attacks that have to analyze individual clients' model updates using secure aggregation. It is flexibly adaptable to the efficiency and security requirements of various FL applications and can be instantiated with MPC or FHE. In contrast to previous works, we only need 2 rounds of communication in each training iteration, do not use any expensive cryptographic primitives on clients, tolerate dropouts, and do not rely on a trusted third party. We implement and benchmark an instantiation of our generic design with secure two-party computation. Our implementation aggregates 500 models with more than 300K parameters in less than 0.5 seconds.

Federated Learning (FL) allows multiple clients to collaboratively train a Neural Network (NN) model on their private data without revealing the data. Recently, several targeted poisoning attacks against FL have been introduced. These attacks inject a backdoor into the resulting model that allows adversarycontrolled inputs to be misclassified. Existing countermeasures against backdoor attacks are inefficient and often merely aim to exclude deviating models from the aggregation. However, this approach also removes benign models of clients with deviating data distributions, causing the aggregated model to perform poorly for such clients.To address this problem, we propose DeepSight, a novel model filtering approach for mitigating backdoor attacks. It is based on three novel techniques that allow to characterize the distribution of data used to train model updates and seek to measure finegrained differences in the internal structure and outputs of NNs. Using these techniques, DeepSight can identify suspicious model updates. We also develop a scheme that can accurately cluster model updates. Combining the results of both components, DeepSight is able to identify and eliminate model clusters containing poisoned models with high attack impact. We also show that the backdoor contributions of possibly undetected poisoned models can be effectively mitigated with existing weight clippingbased defenses. We evaluate the performance and effectiveness of DeepSight and show that it can mitigate state-of-the-art backdoor attacks with a negligible impact on the model's performance on benign data.

In the present era of ubiquitous digitization more and more services are becoming available online which is amplified by the Corona pandemic. The fast-growing mobile service market opens up new attack surfaces to the mobile service ecosystem. Hence, mobile service providers are faced with various challenges to protect their services and in particular the associated mobile apps. Defenses for apps are, however, often limited to (lightweight) application-level protection such as app hardening and monitoring and intrusion detection. Therefore, effective risk management is crucial to limit the exposure of mobile services to threats and potential damages caused by attacks.In this paper, we present FedCRI, a solution for sharing Cyber-Risk Intelligence (CRI). At its core, FedCRI transforms mobile cyber-risks into machine learning (ML) models and leverages ML-based risk management to evaluate security risks on mobile devices. FedCRI enables fast and autonomous sharing of actionable ML-based CRI knowledge by utilizing Federated Learning (FL). FL allows collaborative training of effective risk detection models based on information contributed by different mobile service providers while preserving the privacy of the training data of the individual organizations. We extensively evaluate our approach on several real-world user databases representing 23.8 million users of security-critical mobile apps (since Android 4 and iOS 6) provided by nine different service providers in different European countries. The datasets were collected over the course of six years in the domains of financial services, payments, and insurances. Our approach can successfully extract accurate CRI models, allowing effective identification of cybersecurity risks on mobile devices. Our evaluation shows that the federated risk detection model can achieve better than 99% accuracy in terms of F1-score in most risk classification tasks with a very low number of false positives.

No abstract