Data anonymization has become an essential preprocessing step in many analytics, especially the ones that rely on machine learning. Retrieving external data sources can significantly improve the quality of machine learning models but the sensitive nature often hinders data exchange. On the other hand, anonymization tactics can substantially decrease their utility resulting in inferior machine learning models. This work assesses the quality of anonymized datasets for multi-class classification purposes. A hybrid anonymization pipeline is employed, combining masking and sampling. Deliberate sampling – achieved by balancing the target attribute – not only offers strong and quantifiable privacy guarantees but also improves the utility of released datasets. Our findings are supported by experiments that are executed on three distinct datasets, demonstrating the effectiveness of the approach.
Current firmware update workflows are geared towards ensuring integrity and confidentiality in the face of untrusted servers mediating the communication process. However, they cannot ensure that the update's content preserves deployment-specific security properties, potentially allowing vulnerabilities or malicious code if third-party components compromise or alter the firmware. There is also no guarantee that the new firmware retains the old one's security properties, a crucial requirement in safety-critical environments. We propose an enhancement of the SUIT standard which ensures that software updates preserve a formally-specified set of behavioral properties in the affected components. We demonstrate the feasibility of the proposed workflow in some realistic use cases.
Distributed Denial of Service (DDoS) attacks are one of the main threats facing the Internet today, and a considerable number of them originate from attacks using spoofed source addresses. The Source Address Verification Architecture (SAVA) technology can effectively mitigate such attacks by verifying the legitimacy of the source address. However, the deployment of SAVA faces some practical challenges, including the complexity of real network topologies, high deployment costs, and the impracticality of full deployment. To address these issues, this paper describes the SAVA deployment model in detail and proposes an incremental deployment approximation algorithm. The algorithm can identify a set of approximately optimal SAVA deployment points in any network topology, aiming to maximize the filtering of attack traffic. Experimental results show that compared with conventional deployment methods, the deployment algorithm shows superior performance in handling spoofed source attacks while maintaining a low false negative probability.
Tor, the most widely used anonymity network with around 2 million users, relies on volunteer-operated relays to route traffic through circuits while preserving anonymity. However, it faces a significant challenge, the diversity problem, where power is concentrated among a few dominant relay operators. It increases the risk of Sybil attacks-where attackers introduce multiple relays without clear attribution, potentially enabling traffic monitoring and de-anonymization. This paper examines the diversity problem and Sybil attack implications through data aggregation and simulations, revealing significant vulnerabilities. Notably, the top 10 relay families control over 50
Industrial control systems (ICS), essential for managing critical infrastructure like energy, water, and transportation, are increasingly vulnerable to cyber threats due to the integration of Operational Technology (OT) with Information Technology (IT). This paper introduces FRAMICS, a novel risk assessment methodology for ICS, expanding beyond traditional approaches focusing only on logical components. It integrates technical functions with operational environment, using a new functional model and a customized threat modeling technique based on the STRIDE framework. FRAMICS also enhances risk evaluation by adapting the DREAD model to incorporate empirical data, offering a more objective and rigorous assessment. As a data-driven approach, it improves current ICS risk assessment practices. A use case study of a water treatment system showed that threats to availability and integrity were the highest risks, whereas confidentiality and repudiation risks were less significant due to their limited direct impact on system operations.
