Adversarial Users Research Articles

TETRIS: Towards Exploring the Robustness of Interactive Segmentation

Interactive segmentation methods rely on user inputs to iteratively update the selection mask. A click specifying the object of interest is arguably the most simple and intuitive interaction type, and thereby the most common choice for interactive segmentation. However, user clicking patterns in the interactive segmentation context remain unexplored. Accordingly, interactive segmentation evaluation strategies rely more on intuition and common sense rather than empirical studies (e.g., assuming that users tend to click in the center of the area with the largest error). In this work, we conduct a real-user study to investigate real user clicking patterns. This study reveals that the intuitive assumption made in the common evaluation strategy may not hold. As a result, interactive segmentation models may show high scores in the standard benchmarks, but it does not imply that they would perform well in a real world scenario. To assess the applicability of interactive segmentation methods, we propose a novel evaluation strategy providing a more comprehensive analysis of a model's performance. To this end, we propose a methodology for finding extreme user inputs by a direct optimization in a white-box adversarial attack on the interactive segmentation model. Based on the performance with such adversarial user inputs, we assess the robustness of interactive segmentation models w.r.t click positions. Besides, we introduce a novel benchmark for measuring the robustness of interactive segmentation, and report the results of an extensive evaluation of dozens of models.

Adversarial Heterogeneous Graph Neural Network for Robust Recommendation

Recommendation systems play a vital role in identifying the hidden interactions between users and items in online social networks. Recently, graph neural networks (GNNs) have exhibited significant performance gains by modeling the information propagation process in graph-structured data for a recommendation. However, existing GNN-based methods do not have broad applicability to heterogeneous graphs that integrate auxiliary data with diverse types. Moreover, graph structures are susceptible to noise and even unnoticed malicious perturbations, as perturbations from connected nodes can create cumulative effects on a target node in the graph. To enhance the robustness and generalization of GNN-based recommendations, we propose a new optimization model named Adversarial Heterogeneous Graph Neural Network for RECommendation (AHGNNRec). First, AHGNNRec learns user and item embeddings by exploring the distinct contributions of various types of interactions between users and items using a hierarchical heterogeneous graph neural network (HGNN). Second, to produce more robust embeddings for recommendations, we employ the adversarial training (AT) method to optimize the HGNN layers. AT is a min-max optimization training process where the generated adversarial fake nodes from normal nodes with intentional perturbations try to maximally deteriorate the recommendation performance. Following this, we learn about these adversarial user or item nodes by minimizing the impact of an additional regularization term for the recommendation. The experimental outcomes on two real-world benchmark datasets demonstrate the effectiveness of AHGNNRec.

Requirements for Crafting Virtual Network Packet Captures

Currently, network environments are complex infrastructures with different levels of security, isolation and permissions. The management of these networks is a complex task, faced with different issues such as adversarial attacks, user demands, virtualisation layers, secure access and performance optimisation. In addition to this, forensic readiness is a demanded target. To cover all these aspects, network packet captures are used to train new staff, evaluate new security features and improve existing implementations. Because of this, realistic network packet captures are needed that cover all appearing aspects of the network environment. Packet generators are used to create network traffic, simulating real network environments. There are different network packet generators available, but there is no valid rule set defining the requirements targeting packet generators. The manual creation of such network traces is a time-consuming and error-prone task, and the inherent behaviour of virtual networks eradicates a straight-forward automation of trace generation in comparison to common networks. Hence, we analyse relevant conditions of modern virtualised networks and define relevant requirements for a valid packet generation and transformation process. From this, we derive recommendations for the implementation of packet generators that provide valid and correct packet captures for use with virtual networks.

Rate-Splitting Multiple Access for Communications and Jamming in Multi-Antenna Multi-Carrier Cognitive Radio Systems

With the increasing number of wireless communication systems and the demand for bandwidth, the wireless medium has become a congested and contested environment. Operating under such an environment brings several challenges, especially for military communication systems, which need to guarantee reliable communication while avoiding interfering with other friendly or neutral systems and denying the enemy systems of service. In this work, we investigate a novel application of Rate-Splitting Multiple Access (RSMA) for joint communications and jamming with a Multi-Carrier (MC) waveform in a multi-antenna Cognitive Radio (CR) system. RSMA is a robust multiple access scheme for downlink multi-antenna wireless networks. RSMA relies on multi-antenna Rate-Splitting (RS) strategy at the transmitter and Successive Interference Cancellation (SIC) at the receivers. By employing RSMA at the secondary transmitter, our aim is to simultaneously communicate with Secondary Users (SUs) and jam Adversarial Users (AUs) to disrupt their communications while limiting the interference to Primary Users (PUs) in a setting where all users perform broadband communications by MC waveforms in their respective networks. We consider the practical setting of imperfect CSI at Transmitter (CSIT) for the SUs and PUs, and statistical CSIT for AUs. We formulate a problem to obtain optimal precoders which maximize the mutual information under interference and jamming power constraints. We propose an Alternating Optimization-Alternating Direction Method of Multipliers (AO-ADMM) based algorithm for solving the resulting non-convex problem. We perform an analysis based on Karush-Kuhn-Tucker (KKT) conditions to determine the optimal jamming and interference power thresholds that guarantee the feasibility of problem and propose a practical algorithm to calculate the interference power threshold. By simulation results, we demonstrate that RSMA achieves a higher sum-rate performance than Space Division Multiple Access (SDMA) and Non-Orthogonal Multiple Access (NOMA).

Byzantine-Resilient Secure Federated Learning

Secure federated learning is a privacy-preserving framework to improve machine learning models by training over large volumes of data collected by mobile users. This is achieved through an iterative process where, at each iteration, users update a global model using their local datasets. Each user then masks its local model via random keys, and the masked models are aggregated at a central server to compute the global model for the next iteration. As the local models are protected by random masks, the server cannot observe their true values. This presents a major challenge for the resilience of the model against adversarial (Byzantine) users, who can manipulate the global model by modifying their local models or datasets. Towards addressing this challenge, this paper presents the first single-server Byzantine-resilient secure aggregation framework (BREA) for secure federated learning. BREA is based on an integrated stochastic quantization, verifiable outlier detection, and secure model aggregation approach to guarantee Byzantine-resilience, privacy, and convergence simultaneously. We provide theoretical convergence and privacy guarantees and characterize the fundamental trade-offs in terms of the network size, user dropouts, and privacy protection. Our experiments demonstrate convergence in the presence of Byzantine users, and comparable accuracy to conventional federated learning benchmarks.

Secure TDD MIMO Networks Against Training Sequence Based Eavesdropping Attack

Multi-User MIMO (MU-MIMO) has attracted much attention due to its significant advantage of increasing the utilization ratio of wireless channels. However, Frequency-Division Duplex (FDD) systems are vulnerable to eavesdropping, since the explicit CSI feedback can be manipulated. In this paper, we show that Time-Division Duplex (TDD) systems are insecure as well. In particular, we show that it is possible to eavesdrop on other users' downloads by tuning training sequences. In order to defend MU-MIMO against such threats, we propose a secure CSI estimation scheme, which can provide correct estimates of CSI when adversarial users are in presence. We prove that our scheme is secure against training sequence based eavesdropping attack. We have implemented our scheme for TDD MU-MIMO systems and performed a series of experiments. Results demonstrate that our secure CSI estimation scheme is highly effective in protecting TDD MIMO networks against eavesdropping attack. Furthermore, we extend our scheme to support massive MU-MIMO networks, with a carefully redesigned uplink protocol and optimized power allocation to achieve higher spectral efficiency. To be more practical, we also take mismatch channel issue into our consideration. An enhancement scheme is proposed and we show that our scheme with enhancement is secure and correct under mismatch channel.

Towards Yo-Yo attack mitigation in cloud auto-scaling mechanism

Cloud platforms could automatically scale underlying network resources up and down in response to changes in the traffic load. Such an auto-scaling mechanism can largely enhance the elasticity and scalability of cloud platforms. However, it may introduce new security threats. For example, the Yo-Yo attack is a newly disclosed attack against the cloud auto-scaling mechanism. Attackers periodically send bursts of traffic to cause the auto-scaling mechanism to oscillate between the scale-up process and the scale-down process, which may result in significant performance degradation and economic loss. None of the prior work addressed the problem of mitigating such an attack. In this paper, we propose a Trust-based Adversarial Scanner Delaying (TASD) approach to effectively and proactively mitigate the Yo-Yo attack on the cloud auto-scaling mechanism. In TASD, we first propose to use the trust-based scheme to establish trust values for users, which is leveraged to identify adversarial requests. Trust values are updated by jointly considering the request mode and the auto-scaling status. Then, we aim to disable the condition under which the Yo-Yo attack takes effect by injecting certain delay, under the QoS constraints, to manipulate the response time of suspicious requests and deceive the attackers. Our extensive evaluation demonstrates that our approach achieves promising results, e.g., it can detect at least 80% Yo-Yo adversarial users and reduce more than 41% malicious scale-ups.

A novel physical layer security approach in OFDMA‐MISO networks in the presence of noncooperative and cooperative adversary users with unknown mode

AbstractThis paper proposes a novel physical layer security approach in an orthogonal frequency division multiple access system. Then, a joint subchannel assignment and power allocation method is proposed. It is assumed that there are multiadversary users in the considered orthogonal frequency division multiple access–based network. Each adversary user can be a passive user (ie, he/she just eavesdrops) or an active user (ie, he/she sends jamming to make disturbance). The adversary users can choose active or passive mode according to their situation in the network. In a real network, legitimate users do not have deterministic information about the passive and active modes of adversary users. Here, we assume that the active and passive modes of adversary users are unknown. We also introduce both noncooperative and cooperative scenarios for adversary users. In the noncooperative scenario, the adversary users do not have cooperation and they act independently. In the cooperative scenario, they have cooperation with each other to eavesdrop well and reduce the secrecy rate of the network. For the noncooperative adversary users, a probabilistic method is proposed to estimate the active or passive mode of the adversary users. Based on the proposed probabilistic method, the average secrecy rate for the network is obtained. For the cooperative adversary users, we assume that they are smart and organized; it means that they choose their active or passive mode based on the situation of other adversary users to reach the best performance. Since the total secrecy rate in two scenarios is nonconvex, the corresponding optimization problem is nonconvex. We propose an iterative method for joint subchannel assignment and power allocation to maximize the total secrecy rate of the network. A linear search method is proposed for the subchannel assignment, and a successive convex approximation approach is proposed to solve the power allocation problem.

Protection Against Defecting Attack and Enhanced Channel Allocation for Video Streaming in CRN

Arising issue of spectrum scarcity is resolved by emerging cognitive radio technology. In which, unoccupied portion of licensed spectrum would be granted for unlicensed user in order to upgrade the spectrum utilization. Spectrum sensing is an enabling task for labeling the spectrum holes and away from spectrum congestion in cognitive radio networks. Impact of secondary user’s cooperation developed cooperative spectrum sensing for effective spectrum sensing, many secondary users are to be associated in sensing the idle channels. Adversary secondary users (SU) mimic as primary user in cooperative sensing scheme, called primary user emulation attack (PUEA). If any SU forges the sensing data, that is known as spectrum sensing data falsification attack (SSDF). Rescuing the network against these two attacks is the major issue of cognitive radio networks. To address these challenges, we propose cognitive radio network to ensure security in cooperative spectrum sensing for video streaming application. Sensing based clustering algorithm assembles the cognitive radios into clusters to mitigate cooperative overhead and consumed energy. In order to avert PUEA attacks, we introduce authentication signature ID algorithm which provides authority for licensed user to use spectrum. To diminish SSDF attacks, we exploit Hamming algorithm to abolish forgery node from spectrum sensing decision. Furthermore PRN channel allocation algorithm is employed to precisely deliver the video with high quality and lower delay. For video transmission in cognitive radio network, we need to compress the video frames by block truncation coding-pattern fitting in order to reduce bandwidth consumption. This compression technique offers good video transmission with high compression ratio in cognitive radio network, due to trade-off among quality, bit rate and decoding time. Our proposed framework demonstrates the simulation with improved detection rate of idle channel and reduced the delay for high quality video transmission.

ЗАЩИТА ХРАНИЛИЩ ЭЛЕКТРОННЫХ ДОКУМЕНТОВ ОТ АТАК МЕТОДОМ ПЕРЕБОРА ПОИСКОВЫХ ЗАПРОСОВ

The paper presents the scheme of data management and protocols for securing document collection against adversary users who try to abuse their access rights to find out the full content of confidential documents. The configuration of secure document retrieval system is described and a suite of protocols among the clients, warehouse server, audit server and database management server is specified. The scheme makes it infeasible for clients to establish correspondence between the documents relevant to different search queries until a moderator won’t give access to these documents. The proposed solution allows ensuring higher security level for document warehouses.

Security Implications of Data Mining in Cloud Scheduling

Cloud providers host an increasing number of popular applications, on the premise of resource flexibility and cost efficiency . Most of these systems expose virtualized resources of different types and sizes. As instances share the same physical host to increase utilization, they contend on hardware resources, e.g., last-level cache, making them vulnerable to side-channel attacks from co-scheduled applications. In this work we show that using data mining techniques can help an adversarial user of the cloud determine the nature and characteristics of co-scheduled applications and negatively impact their performance through targeted contention injections. We design Bolt, a simple runtime that extracts the sensitivity of co-scheduled applications to various types of interference and uses this signal to determine the type of these applications by applying a set of data mining techniques. We validate the accuracy of Bolt on a 39-server cluster. Bolt correctly identifies the type and characteristics of 81 percent out of 108 victim applications, and constructs specialized contention signals that degrade their performance. We also use Bolt to find the most commonly-run applications on EC2. We hope that underlining such security vulnerabilities in modern cloud facilities will encourage cloud providers to introduce stronger resource isolation primitives in their systems.

Securing Warning Message Dissemination in VANETs Using Cooperative Neighbor Position Verification

Efficient schemes for warning message dissemination in vehicular ad hoc networks (VANETs) use context information collected by vehicles about their neighbor nodes to guide the dissemination process. Based on this information, vehicles autonomously decide whether they are the most appropriate forwarding nodes. These schemes maximize their performance when all the vehicles advertise correct information about their positions, but position errors may drastically reduce the performance of the dissemination process. We present a proactive cooperative neighbor position verification protocol that detects nodes advertising false locations and selects optimal forwarders to mitigate the impact of adversarial users. We combine our mechanism with two warning dissemination schemes for VANETs and demonstrate how the latter can benefit from the use of our security scheme in the presence of malicious nodes trying to exploit known system vulnerabilities.

Multiple Server Indirect Security Authentication Protocol for Mobile Networks Using Elliptic Curve Cryptography (ECC)

More number of research works develops many authentication protocols and algorithms to provide a security in the mobile networks. But these existing protocols have the drawbacks in which a single and more than one server put in practice. These protocols suffer if the single server is compromised by the adversaries and in more than one server the protocols finds high complexity in design and integration of keys. To address these issues in the existing works, in this paper a new complexity reduced protocol is proposed with multiple servers. In this paper, we propose an indirect security protocol for mobile networks based on ECC scheme. Initially, we generate keys to the multiple servers by the ECC scheme and these servers validate the each user node to provide the information. In mobile networks, it is essential to validate every node when it tries to request information from base station. After the key generation and the validation by the servers the information sharing process is accomplished between the user and the base station. In our security protocol is highly secure by generating cryptographic schemes distributed based on ECC and executed over multiple servers. The proposed indirect security protocol performance is analyzed by invoking more number of authenticated and adversary users and the result from this process guarantees the healthiness of the proposed security protocol. The implementation results show the effectiveness of proposed ECC based indirect security protocol in authenticating the users to access the information from the base station and the achieved improvement in security process. Moreover, the performance of the proposed indirect security protocol is evaluated by comparing with the existing indirect authentication security protocol in mobile networks.

An improvement of discrete Tardos fingerprinting codes

It has been proven that the code lengths of Tardos's collusion-secure fingerprinting codes are of theoretically minimal order with respect to the number of adversarial users (pirates). However, the code lengths can be further reduced as some preceding studies have revealed. In this article we improve a recent discrete variant of Tardos's codes, and give a security proof of our codes under an assumption weaker than the original Marking Assumption. Our analysis shows that our codes have significantly shorter lengths than Tardos's codes. For example, when c = 8, our code length is about 4.94% of Tardos's code in a practical setting and about 4.62% in a certain limit case. Our code lengths for large c are asymptotically about 5.35% of Tardos's codes.

An Efficient 2-Secure and Short Random Fingerprint Code and Its Security Evaluation

The code length of Tardos's collusion-secure fingerprint code is of theoretically minimal order with respect to the number of adversarial users (pirates). However, the constant factor should be further reduced for practical implementation. In this article, we improve the tracing algorithm of Tardos's code and propose a 2-secure and short random fingerprint code, which is secure against collusion attacks by two pirates. Our code length is significantly shorter than that of Tardos's code and its tracing error probability is practically small.