Panoptic perception models in autonomous driving use deep learning models to interpret their surroundings and make real-time decisions. However, these models are susceptible, carefully designed noise can fool models all while being imperceptible to humans. In this work, we investigate the impact of black-box adversarial noise attacks on three core perception tasks: drivable area recognition, lane line segmentation, and object detection. Unlike white-box attacks, black-box attacks assume no knowledge of the model’s internal parameters making them a more realistic and challenging threat scenario. Our goal is to evaluate how such an attack affects the model’s predictions and explore countermeasures towards such attacks. In response to our implemented attack, we have tested various defense methods. With each defense method, we have assessed the recovery on prediction accuracy. This research aims to provide valuable insights into the vulnerabilities of panoptic perception models and highlights strategies for enhancing their resilience against adversarial manipulation within real-world scenarios. All our attacks are performed against images from the BDD100K dataset.

Recent studies have demonstrated significant success in detecting attacks on the Controller Area Network (CAN) bus network using machine learning and deep learning models, including convolutional neural networks and transformer-based architectures. Building on this foundation, our work investigates the use of large language models (LLMs) not only for intrusion detection but also for providing interpretable explanations of their decisions. We fine-tuned three LLMs, i.e., SecureBERT, LLaMA-2, and LLaMA-3, for intrusion detection on CAN bus data. Among them, LLaMA-3 delivered the best results, achieving SOTA performance on the Car-Hacking dataset. Beyond attack classification, we evaluated LLaMA-3’s ability to generate reasoning for its decisions through zero-shot prompting. The model successfully articulated its rationale, particularly for Denial-of-Service (DoS) attacks, demonstrating strong potential for explain-ability in intrusion detection systems. These findings highlight the potential of LLMs to serve as a highly accurate intrusion detection system while simultaneously providing interpretable explanations, thereby enhancing the investigative capabilities of cybersecurity professionals.

Large Language Models (LLMs) offer remarkable capabilities for vision-language tasks. However, their large size and computational demands often necessitate deployment in cloud environments, which introduces additional data and network security risks for critical applications. In this work, we investigate whether TinyLLMs can achieve competitive object detection performance while offering unique advantages in vocabulary flexibility and edge deployment. We present BBOB (pseudo-acronym), a LLaVA-style TinyLLM framework for adaptive object detection on edge devices. BBOB leverages MobileViTv2 combined with a spatially aware vision projector to enable TinyLLMs to perform fine-grained object detection through pure language modeling. BBOB demonstrates preliminary yet insightful performance, establishing important baselines for multimodal TinyLLM research and confirming the feasibility of adapting TinyLLMs to complex vision-language tasks. BBOB showed promising performance classifying objects with 47.1% accuracy, but struggled to localize them, resulting in a 1.7% mAP@50 score on COCO while only having a total of around 50 million parameters.

Numerous studies have proved their effective strength in detecting Control Area Network (CAN) attacks. In the realm of understanding the human semantic space, transformer- based models have demonstrated remarkable effectiveness. Leveraging pre-trained transformers has become a common strategy in various language-related tasks, enabling these models to grasp human semantics more comprehensively. To delve into the adaptability evaluation on pre-trained models for CAN intrusion detection, we have developed three distinct models: CAN-C-BERT, CAN-SecureBERT and CAN-Llama2. Notably, our CAN-Llama2 model surpasses the state-of-the-art models by achieving an exceptional performance 0.999993 in terms of Balanced Accuracy (BA), Precision (PREC), Detection Rate (DR), F1 score, and a remarkably low false alarm rate of 3.10e-6. Impressively, the false alarm rate is 52 times smaller than that of the leading model, MTH-IDS (Multitiered Hybrid Intrusion Detection System). Our study underscores the promise of employing a Large Language Model as the foundational model, while incorporating adapters for other cybersecurity-related tasks and maintaining the model’s inherent language-related capabilities.

Despite increasing popularity of anonymous communications, current anonymity networks are still vulnerable to traffic analysis attacks. We propose Homomorphic EnCrypTion enabled Onion Routing (HECTor) to defeat existing traffic analysis attacks. Instead of mixing traffic at the packet level as current anonymity networks, HECTor mixes traffic at the bit level with network coding techniques. In HECTor, homomorphic encryption enables onion-like layered encryption which has been proven successful in existing anonymity networks. We theoretically analyze the performance of HECTor. Our extensive experiments on HECTor show that HECTor can effectively mitigate existing traffic analysis attacks with high throughput.

To expedite the forensic investigation process in the cloud, excessive and yet volatile data need to be acquired, transmitted, and analyzed in a timely manner. A common assumption for most existing forensic systems is that credible data can always be collected from a cloud infrastructure, which might be susceptible to various exploits. In this paper, we present the design, implementation, and evaluation of LiveForen, a system that enforces a trustworthy forensic data acquisition and transmission process in the cloud, whose computer platforms' integrity has been verified. To fulfill this objective, we propose two secure protocols that verify the fingerprints of the computer platforms, as well as the attributes of the human agents, by taking advantage of the trusted platform module and the attribute-based encryption. To transmit forensic data as a data stream and verify its integrity at the same time, a unique fragile watermark is embedded into the data stream without altering the data itself. The watermark allows not only the data integrity to be verified but also any malicious data manipulation to be localized, with minimum communication overhead. The experimental results demonstrate that LiveForen achieves good scalability and limited performance overhead for authentication, data transmission, and integrity verification in an Infrastructure-as-a-Service cloud environment.

Vehicular ad hoc networks may one day prevent injuries and reduce transportation costs by enabling new safety and traffic management applications, but these networks raise privacy concerns because they could enable applications to perform unwanted surveillance. Researchers have proposed privacy protocols, measuring privacy performance based on metrics such as k-anonymity. Because of the frequency and precision of location of queries in vehicular applications, privacy measurement may be improved by considering additional factors. This paper defines continuous network location privacy; presents KDT-anonymity, which is a composite metric including average anonymity set size, i.e., K, average distance deviation, i.e., D, and anonymity duration, i.e., T; derives formulas to calculate theoretical values of K, D, and T; evaluates five privacy protocols under realistic vehicle mobility patterns using KDT-anonymity; and compares KDT-anonymity with prior metrics.

In the next-generation wireless networks such as 5G network or very-high-throughput WLAN, the users (humans or machines) in human-to-human, human-to-machine and machine-to-machine communications usually have heterogenous demands and behaviors. In this case, it is very difficult to uniformly evaluate user Quality-of-Experience (QoE) by one or more traditional performance metrics (e.g. throughput, delay and interference level) due to the heterogeneity. Moreover, the diversity of terminal capacities further exacerbates the difficulty in QoE provisioning since parameter adjustment may be inapplicable to all the equipments. To solve the aforementioned problems, this paper introduces a new QoE evaluation metric: satisfaction, which is a measurable, comparable and general assessment to indicate the gap between the current QoE and the desire of a user. Based on this new metric, we model the channel contention of random access networks as a non-cooperative game, where the players (contention nodes) are altruistic but also care about their individual interests. Theoretical analysis and simulation experiments indicate that a reinforcement learning algorithm can help the proposed game to achieve an efficient Nash equilibrium, which maximizes individual satisfaction with proportional fairness.

In this paper, we propose a new method called Ambiguous Multi-Symmetric Cryptography (AMSC) that conceals multiple plain-texts in a cipher-text using the same number of keys. The goal of this method is to overcome the problem of symmetric cryptography failure when the shared key is exposed. The proposed method AMSC is a cryptographic primitive that preserves plausible deniability after a cryptographic key is discovered. We evaluate AMSC in terms of security and complexity. The security analysis shows that our scheme withstands all security attack models with different knowledge of the adversary. In terms of time complexity, AMSC produces the cipher-text in polynomial time with respect to the number and size of the plaintexts and keys. AMSC has two main applications: a) It sends multiple messages for multiple receivers through one cipher-text. b) It sends one real message and multiple decoys to defeat attacks by providing security beyond conventional brute-force bounds. For both applications, AMSC can be used to deny encryption.

Vehicular Ad Hoc Network (VANET) is a promising technology in which vehicle-to-vehicle and vehicle-to-roadside infrastructure wireless communications can be achieved. This is important to obtain road safety for vehicles and drivers and collision avoidance. A falsified position by malicious users is one of the important issues in VANETs. Vehicle position identification is one of the important aspects in establishing authentication and security between inter vehicular communication exchange. Deepa et al presented two approaches for verifying sender's position in a multi-hop network. Their first proposed algorithm relies on signal propagation time for verifying the position. Their second proposed algorithm verifies the position information with the help of base stations located in the coverage area of the vehicular network The main contribution of our work is validating their approach by running an ns2 simulation with dynamic number of nodes in various mobility scenarios such as urban, rural, Manhattan. We have also generated different scenarios with variable velocity ranges and simulated the VANET. We have also considered the effect of delay, jitter in our simulation and observed that the proposed approach is robust and a feasible solution to the problem of Active Position detection.

Wireless networks which would connect vehicles via the Internet to a location based service, LBS, also would expose vehicles to online surveillance. In circumstances when spatial cloaking is not effective, such as when continuous precise location is required, LBSs may be designed so that users relay dummy queries through other vehicles to camouflage true locations. This paper introduces PARROTS, Position Altered Requests Relayed Over Time and Space, a privacy protocol which protects LBS users' location information from LBS administrators even (1) when the LBS requires continuous precise location data in a vehicular ad hoc network, (2) when LBS administrators collude with administrators of vehicular wireless access points (a.k.a. roadside units, or RSUs), and (3) when precise location data might be deanonymized using map databases linking vehicle positions with vehicle owners' home/work addresses and geographic coordinates. Defense against deanonymization requires concealment of endpoints, the effectiveness of which depends on the density of LBS users and the endpoint protection zone size. Simulations using realistic vehicle traffic mobility models varying endpoint protection zone sizes measure improvements in privacy protection.

In this paper, we study topologies of sensor networks deployed for tracking multiple targets. Tracking multiple moving targets is a challenging problem. Most of the previously proposed tracking algorithms simplify the problem by assuming access to the signal from an individual target for tracking. Recently, tracking algorithms based on blind source separation (BSS), a statistical signal-processing technique widely used to recover individual signals from mixtures of signals, have been proposed. BSS-based tracking algorithms are proven to be effective in tracking multiple indistinguishable targets. The topology of a wireless sensor network deployed for tracking with BSS-based algorithms is critical to tracking performance because the topology affects separation performance, and the topology determines accuracy and precision of estimation on the paths taken by targets. We propose cluster topologies for BSS-based tracking algorithms. Guidelines on parameter selection for proposed topologies are given in this paper. We evaluate the proposed cluster topologies with extensive experiments. Our experiments show that the proposed topologies can significantly improve both the accuracy and the precision of BSS-based tracking algorithms.

In this paper, we propose an approach to track multiple non-cooperative targets with wireless sensor networks. Most existing tracking algorithms can not be directly applied to non-cooperative target tracking because they assume the access to signals from individual targets for tracking by assuming that: 1) there is only one target in a field; 2) signals from different co-operative targets can be differentiated; or 3) interference caused by signals from other targets is negligible because of attenuation. We propose a general approach for tracking non-cooperative targets. The tracking algorithm first separates the aggregate signals from multiple indistinguishable targets via the blind source separation (BSS) algorithms. Through the analysis on both the temporal and spatial correlation of the separated individual signals, the tracking algorithm determines the location of a target and its moving track. A voting scheme based on the spatial information is designed to better estimate the moving track. Furthermore, we analyze and discuss the influence of signal attenuation and the tracking resolution of the proposed tracking approach. Our experiments show that the proposed approach can both accurately and precisely track multiple indistinguishable moving targets.

In this paper, we propose a generic approach to tracking multiple indistinguishable moving targets with wireless sensor networks. The tracking algorithm first separates the aggregate signals from multiple indistinguishable targets via the Blind Source Separation (BSS) method. Through the analysis on both the temporal and the spatial correlation of the separated individual signals, the tracking algorithm determines the real-time location of a target and its moving track. A voting scheme based on the spatial information is designed to better estimate the moving track. Furthermore, we analyze and discuss the influence of signal attenuation and the tracking resolution of the proposed tracking approach. Both empirical experiments and simulations show that the proposed approach can both accurately and precisely track multiple indistinguishable moving targets.

Vehicular Ad Hoc Networks (VANETs) and their diverse applications experience growing interest in both academic and industry. Different types of traffic packets delivered through vehicle-to-vehicle and vehicle-to-infrastructure communications are intended to improve passenger safety and comfort. In this paper, we propose a multiple priority supported Medium Access Control (MAC) protocol for VANETs based on the time slotted p-persistent channel access mechanism. The protocol differentiates the services packets into multi-priority on the Control Channel (CCH). Theoretical analysis based on Markov model is presented to optimize the transmission probabilities of the packets with different priorities, as well as the adjustable intervals of the CCH and the Services Channels (SCHs). Both analytical results and simulation experiments show that the proposed MAC protocol is able to ensure the prioritized transmission of the safety packets and also achieve optimal system performance with respect to saturated throughput.

In recent years, governments, standardization bodies, automobile manufacturers, and academia are working together to develop vehicular ad hoc network (VANET)-based communication technologies. VANETs apply multiple channels, i.e., control channel (CCH) and service channels (SCHs), to provide open public road safety services and the improve comfort and efficiency of driving. Based on the latest standard draft IEEE 802.11p and IEEE 1609.4, this paper proposes a variable CCH interval (VCI) multichannel medium access control (MAC) scheme, which can dynamically adjust the length ratio between CCH and SCHs. The scheme also introduces a multichannel coordination mechanism to provide contention-free access of SCHs. Markov modeling is conducted to optimize the intervals based on the traffic condition. Theoretical analysis and simulation results show that the proposed scheme is able to help IEEE 1609.4 MAC significantly enhance the saturated throughput of SCHs and reduce the transmission delay of service packets while maintaining the prioritized transmission of critical safety information on CCH.

Communication messages in Vehicular Ad-hoc Networks (VANETs) can be used to track movement of vehicles. In this paper, we address the problem of movement tracking and enhance location privacy without affecting security and safety of vehicles. By considering unique characteristics of VANETs, we firstly propose a synchronized pseudonym changing protocol based on the concept of forming groups among neighboring vehicles. Secondly, we analytically evaluate the anonymity and unlinkability of the proposed protocol. Finally, we do a series of simulations to evaluate the performance of our protocol in real VANET environments such as Manhattan and Urban. Simulation results show that our protocol is feasible and produces excellent performances. The main advantages of our protocol compared with the existing approaches include: 1) it makes larger anonymity set and higher entropy; 2) it reduces the tracking probability; 3) it can be used in both safety and non-safety communications; and 4) Vehicles need not suspend regular communication for changing pseudonyms.

The emerging wireless vehicular communication technology is intended to improve safety and comfort of transportation systems. Different types of traffic information could be delivered through vehicle-to-vehicle and vehicle-to-infrastructure communications. This paper proposes a Quality-of-Service (QoS) supported multi-channel MAC scheme for Vehicular Ad Hoc Networks (VANETs), which can adaptively tune the contention window for different services at each node, and dynamically adjust the intervals of the Control Channel (CCH) and the Service Channels (SCHs) working in multi-rate. Theoretical model is proposed to obtain the contention window and optimize the intervals based on traffic conditions. Analysis and simulation results show that the proposed MAC is able to help IEEE 1690.4 MAC support QoS services, while ensuring the high saturation throughput and the prioritized transmission of critical safety information.

Abstract In the near future, vehicle‐to‐vehicle and vehicle‐to‐roadside infrastructure communications will enable numerous applications, such as collision avoidance, traffic management, location‐based services, and infotainment for the automobile field. Position and velocity play important roles in these applications such as collision avoidance. Forged position or velocity information will make a severe impact on both performance and security of the vehicular ad‐hoc network (VANET). The main challenges in developing secure position and velocity schemes for VANETs are the highly dynamic environment and cost constraints. In this paper, we propose a position and velocity verification scheme. Then via ns‐2, we simulate the proposed scheme in various vehicular mobility environments including rural, urban and Manhattan. Finally, we evaluate the proposed scheme for both position and velocity verifications with automotive grade differential global positioning system (DGPS) and intentionally induced errors as well. The performance evaluation results show that (1) our scheme is very precise in both position and velocity verifications and (2) the proposed scheme is very robust in position verification for rural, urban and Manhattan scenarios even if the introduced errors are in the order of 10 m. Copyright © 2010 John Wiley & Sons, Ltd.

Abstract The distinguishing properties of VehicularAd hocwireless Networks (VANETs) strongly challenge the design of Medium Access Control (MAC) protocols, which are responsible for the medium access coordination among active vehicles, as well as the accommodation of both driving safety applications and non‐safety applications. In this paper, we focus on a comprehensive survey of VANET MAC schemes by integrating various related issues and challenges. Our analysis not only deepens the understanding of MAC techniques in VANETs but also presents the key ideas and potential directions for future research in this area. In order to significantly improve the communication performance of VANETs, more research efforts on MAC techniques must be made for optimizing multichannel coordination and allocation approaches, enhancing the Quality of Service (QoS) capability, and combating the hidden terminal problem, broadcast storm problem and even ACK (acknowledgment) explosion problem. Copyright © 2009 John Wiley & Sons, Ltd.

Communications through road side units in Vehicular Ad hoc Networks (VANETs) can be used to track the location of vehicles, which makes serious threat on users' privacy. In this paper, we propose and evaluate a novel location privacy enhancement protocol for VANETs. Firstly, we propose an Anonymous Online Service Access (AOSA) protocol. Secondly, we analytically evaluate the anonymity and the unlinkability of the proposed protocol. Finally, a series of simulation studies are conducted to evaluate the performance of our protocol in the real VANET environments such as Manhattan and Urban scenarios. According to analytical evaluation and simulations, our protocol provides higher level of anonymity and location privacy for on-line service access applications. Simulation results further show that our protocol is feasible and produces better performance in real VANET environments by producing higher success ratio and smaller delay.

Traditional research on preserving privacy in data mining focuses on time-invariant privacy issues. With the emergence of time series data mining, traditional snapshot-based privacy issues need to be extended to be multi-dimensional with the addition

Inter-vehicle communication is an emerging field of research which focuses on different ways of increasing safety and efficiency in future intelligent highways. However, the security in these inter vehicular communication is a major issue. Since the vehicular network does not have a fixed infrastructure, developing a security scheme in a highly dynamically changing environment is a major challenge. In this paper we are proposing two algorithms for verifying the position information claimed by other vehicles in a multi-hop vehicular network.

Controller area network (CAN) is the leading serial bus system for embedded control. More than two billion CAN nodes have been sold since the protocol's development in the early 1980s. CAN is a mainstream network and was internationally standardized (ISO 11898-1) in 1993. This paper describes an approach to implementing security services on top of a higher level controller area network (CAN) protocol, in particular, CANopen. Since the CAN network is an open, unsecured network, every node has access to all data on the bus. A system which produces and consumes sensitive data is not well suited for this environment. Therefore, a general-purpose security solution is needed which will allow secure nodes access to the basic security services such as authentication, integrity, and confidentiality.

In order to combat the exposed terminal problem and the interference accumulation phenomenon in wireless channels, this paper proposed a new Asymmetrical Dual Busy Tones (ADBT) MAC protocol. By introducing BT-CTS/BTACK mechanism and the approximate algorithm of interfering nodes, ADBT MAC can solve the exposed terminal problem and the hidden terminal problem in the environment with large interference range. Moreover, the effect of the interference accumulation phenomenon can be alleviated by the power adjustment approach for the busy tones. The network performances of DCF, DBTMA, VPDBT and ADBT MAC protocols are compared via OPNET-based simulation experiments, which verify that ADBT is able to prevent packet collisions effectively and improve the channel utilization significantly.

Skype is one of the most popular voice-over-IP (VoIP) service providers. One of the main reasons for the popularity of Skype VoIP services is its unique set of features to protect privacy of VoIP calls such as strong encryption, proprietary protocol, unknown codec, dynamic path selection, and constant packet rate. In this paper, we propose a class of passive traffic analysis attacks to compromise privacy of Skype VoIP calls. The proposed attacks are based on application-level features extracted from VoIP call traces. The proposed attacks are evaluated by extensive experiments over different types of networks including commercialized anonymity networks and our campus network. The experiments show that the proposed traffic analysis attacks can detect speaker and speech of Skype calls with 0.33 and 0.44 detection rate, about 30-fold and 15-fold improvement over random guess respectively.

When setting up a sensor network, one of the first requirements is to establish cryptographic keys to provide encrypted communication links. However, the traditional key establishment techniques cannot be directly applied due to the inherent properties of sensor networks. Recently, several random key pre-distribution schemes based on symmetric cryptography have been proposed to solve the key establishment issue of sensor networks. In this paper, we propose a framework, Blom-based Q-composite (BBQ) scheme, which generalizes various random key pre-distribution schemes. With the aid of BBQ, we systematically analyze, characterize, and discuss the properties of various random key pre-distribution schemes, including connectivity, scalability and security

M. Aydos, B. Savas and C.K. Koc (see Proc. 2nd Int. Workshop on Discrete Algorithms and Methods for Mobile Computing and Commun., 1998) proposed a wireless authentication and key agreement protocol (ASK-WAP) based on elliptic curve cryptography (ECC). We find that this protocol is vulnerable to a man-in-the-middle attack, a denial-of-service attack and an impersonation attack. We present the above mentioned attacks on the ASK-WAP. We also propose a variant of ASK-WAP, the user authentication protocol (UAP), so that it resists these attacks. Furthermore, we analyze the security and performance of the proposed UAP. The results show that our proposed UAP is much more secure and is also efficient with few message exchanges and less computation.

The IETF's integrated services (IntServ) architecture, together with reservation aggregation, provides a mechanism to support the quality-of-service demands of real-time flows in a scalable way, i.e., without requiring that each router be signaled with the arrival or departure of each new flow for which it forwards data. However, reserving resources in "bulk" implies that the reservation does not precisely match the true demand. Consequently, if the flows' demanded bandwidth varies rapidly and dramatically, aggregation can incur significant performance penalties of under-utilization and unnecessarily rejected flows. On the other hand, if demand varies moderately and at slower time scales, aggregation can provide an accurate and scalable approximation to IntServ. We develop a simple analytical model and perform extensive trace-driven simulations to explore the effectiveness of aggregation under a broad class of factors. Example findings include: 1) a simple single-time-scale model with random noise can capture the essential behavior of surprisingly complex scenarios; 2) with a two-order-of-magnitude separation between the dominant time scale of demand and the time scale of signaling and moderate levels of secondary noise, aggregation achieves a performance that closely approximates that of IntServ.