Nirnay Ghosh

A planner-based approach to generate and analyze minimal attack graph

In the present scenario, even well administered networks are susceptible to sophisticated cyber attacks. Such attack combines vulnerabilities existing on different systems/services and are potentially more harmful than single point attacks. One of the methods for analyzing such security vulnerabilities in an enterprise network is the use of attack graph. It is a complete graph which gives a succinct representation of different attack scenarios, depicted by attack paths. An attack path is a logical succession of exploits, where each exploit in the series satisfies the preconditions for subsequent exploits and makes a causal relationship among them. Thus analysis of the attack graph may help in assessing network security from hackers’ perspective. One of the intrinsic problems with the generation and analysis of such a complete attack graph is its scalability. In this work, an approach based on Planner, a special purpose search algorithm from artificial intelligence domain, has been proposed for time-efficient, scalable representation of the attack graphs. Further, customized algorithms have been developed for automatic generation of attack paths (using Planner as a low-level module). The analysis shows that generation of attack graph using the customized algorithms can be done in polynomial time. A case study has also been presented to demonstrate the efficacy of the proposed methodology.

An Approach for Security Assessment of Network Configurations Using Attack Graph

With increasing network security threats, the network vulnerability must consider exploits in the context of multistage, multi-host attack scenarios. The general approach to this problem is to construct an attack graph for a given network configuration. An attack graph consists of a number of attack paths which are essentially series of exploits which an attacker employs to reach the destination. Each attack path depicts an attack scenario. As the number of attack scenarios increases, the overall security of the network reduces. Thus there is need for quantification of security level of a given network. In this paper, two security metrics, namely probabilistic security metric and attack resistance metric, have been employed to evaluate the relative security levels of various network configurations. A case study has been presented to demonstrate the applicability of the proposed approach.

An approach to identify and monitor SLA parameters for storage-as-a-service cloud delivery model

Cloud computing is an emerging technological paradigm that follows pay-as-you-use model. By this model, the consumers are charged according to the usage without regard to where the services are hosted or how they are delivered. Usually cloud services are offered as various delivery models among which Storage-as-a-Service cloud is gaining interest in recent time. In this model, raw (block) storage is offered as a service. Like any other utility services, cloud negotiates its service levels and guarantees with its consumers by establishing Service Level Agreements (SLAs). SLA is important in any provider-consumer interaction as it defines a formal basis for performance and availability the provider guarantees to deliver. It consists of a set of measurable attributes called SLA parameters which are established by some objectively measurable conditions, termed as Service Level Objectives (SLOs). However, the SLAs provided by the present day state-of-the-art clouds are relatively biased towards vendors and do not provide any formal method of verifying if the guarantees are complying or not. This paper attempts to identify the SLA parameters for Storage-as-a-Service cloud delivery model and proposes a monitoring framework for compliance checking.

NetSecuritas: An Integrated Attack Graph-based Security Assessment Tool for Enterprise Networks

Sophisticated cyber-attacks have become prominent with the growth of the Internet and web technology. Such attacks are multi-stage ones, and correlate vulnerabilities on intermediate hosts to compromise an otherwise well-protected critical resource. Conventional security assessment approaches can leave out some complex scenarios generated by these attacks. In the literature, these correlated attacks have been modeled using attack graphs. Although a few attack graph-based network security assessment tools are available, they are either commercial products or developed using proprietary databases. In this paper, we develop a customized tool, NetSecuritas, which implements a novel heuristic-based attack graph generation algorithm and integrates different phases of network security assessment. NetSecuritas leverages open-source libraries, tools and publicly available databases. A cost-driven mitigation strategy has also been proposed to generate network security recommendations. Experimental results establish the efficacy of both attack graph generation and mitigation approach.

Analysis of attack graph-based metrics for quantification of network security

Computer network has grown both in size and complexity with the advent of Internet. It facilitates easy access to vast store of reference materials, collaborative computing, and information sharing. However, this requires a secure interconnected world of computing where confidentiality, integrity, and availability of information and resources are restored. Traditionally, security mechanism is enforced by access control and authentication. However, these security best practices do not take operating system, or network service-based or application vulnerabilities (programming flaws) into account. With the evolution of sophisticated hacking tools, attackers exploit these vulnerabilities and can gain legitimate access to network resources, bypassing the access control and authentication policies. One tool that presents a succinct representation of different attack scenarios specific to a network is attack graph. Attack graph models service or application-based attacks and depicts all possible multihost multi-step attack scenarios that an attacker can launch to penetrate into an enterprise network. The severity associated with each attack scenario can be evaluated following some attack graph-based security metrics. A good number of security metrics are prevalent in the literature, however, there exists no reported work which determines their efficacy and applicability. In this paper, a survey on attack graph-based metrics has been done and comparative analysis of the existing metrics has been presented to facilitate understanding of a given networks level of security strength. A case study has been perceived for the purpose of analysis.

Enhancing Reliability of Vehicular Participatory Sensing Network: A Bayesian Approach

Participatory sensing (PS) is an emerging socio-technological paradigm in which citizens voluntarily participate and contribute to a distributed information system using applications installed in their hand-held devices. It can be found in a number of real-life applications, viz. traffic monitoring, air/sound pollution, garbage monitoring, social networking, commodity pricing, and so on. In these systems, information sensed by the user helps the peers in decision making. Present work considers vehicular participatory sensing systems, where registered user senses (perceives) the traffic incident and submits its report(s) to a PS application server. PS application server in turn, broadcasts those reports as alerts to its subscribers. To promote the participation, the PS systems used to have incentive schemes for the participants. However, a common problem in participatory sensing is the generation of false reports either due to wrong perception of an event or to maliciously increase the degree of participation to gain undue incentives. Such false reports make the usage of the PS system unreliable and vulnerable to the illusion attack. This work proposes a novel approach to make PS applications more reliable by identifying and filtering out the falsely reported event through automated confidence assignment based on a probabilistic model. Waze traffic alerts have been used as the dataset to validate the proposed filtering mechanism. Finally, simulation-based experiments and performance evaluation have been done to demonstrate that the proposed approach is relatively accurate.

Securing Loosely-Coupled Collaboration in Cloud Environment through Dynamic Detection and Removal of Access Conflicts

Online collaboration service has become a popular offering of present day Software-as-a-Service (SaaS) clouds. It facilitates sharing of information among multiple participating domains and accessing them from remote locations. Owing to loosely-coupled nature of such collaborations, access request from a remote user is made in the form of a set of permissions. The cloud vendor maps the requested permissions into appropriate local roles in order to allow resource access. However, coexistence of such multiple simultaneous role activation requests may introduce conflicts which violate the principle of security. In this paper, we propose a distributed secure collaboration framework which enables collaborating domains to detect and remove these conflicts. Two features of our framework are: (i) it requires only local information, and (ii) it detects and removes conflicts on-the-fly. Formal proofs have been provided to establish the correctness of our approach. Experimental results and qualitative comparison with related work demonstrate the efficacy of our approach in terms of response time, thus addressing the scalability requirement of cloud services.

A quantitative approach towards detection of an optimal attack path in a wireless network using modified PSO technique

In recent years there has been an immense proliferation of wireless networks and they are becoming increasingly vulnerable to attacks. Thus there is a serious need to secure such networks from attacks. Usually an attacker can penetrate into a network by utilizing a chain of exploits. An exploit is a small piece of code that makes use of vulnerabilities present in a service or in a system. Each exploit in the chain has a set of preconditions and effects and lays the groundwork for the subsequent exploits. Application of such a chain of exploits generates a set of attack states or network states which form a path called the attack path and combining many such attack paths produces an attack graph. A lot of research has been done on issues such as scalable and time efficient ways of generation of attack graphs in wired network in contrast to that in wireless scenario. Moreover, the need is to identify the path that may be chosen by the attacker to comprise a target system in less time and effort. The proposed methodology in this paper aims as finding out the optimal or risk-prone attack path that the attacker may choose to penetrate a wireless network. The generation of attack paths in a wireless network is itself a difficult proposition due to networks inherent dynamic nature and ever changing topology. In this work, the Particle Swarm Optimization (PSO) technique has been employed for finding out the optimal attack path using an attack vector metric. The effort required on the part of the attacker to compromise a target system has been termed as an attack vector. The wireless nodes have been assigned severity measures obtained from customized risk parameters which serve as an input to the modified PSO technique. A case study has also been presented to demonstrate the efficacy of the proposed methodology.

A probabilistic approach for filtering out spam reports in a vehicular participatory sensing system

Participatory sensing (PS) is an emerging socio-technological paradigm by which citizens voluntarily participate and contribute to a distributed information management system using applications installed in their hand-held devices. However, a common problem in participatory sensing is that some users generate spam (false) reports to increase their degree of participation to maliciously gain undue incentives. Such spam reports make the usage of the PS system unreliable and vulnerable to illusion attacks. This work proposes a novel approach to make PS applications authentic by identifying and filtering out spam reports through automated confidence assignment based on a probabilistic model. Waze traffic alerts have been used to validate the proposed report filtering mechanism. Results have been found to be satisfactory and useful in enhancing reliability of the PS system.

W2Q: A dual weighted QoI scoring mechanism in social sensing using community confidence

A significant vulnerability in social sensing based services is false notifications from sensing agents, thereby resulting in inaccurate published information that induces loss of revenue and business goodwill. Existing popular schemes utilize rating feedbacks (over the published information) to quantify the perceived usefulness (quality) of the information. However, these schemes do not reward the confidence of the feedback community and lacks provision to regulate the impact of uncertain feedbacks (ratings), and hence can be easily manipulated. In this paper, we propose a model, called W2Q, to mathematically evaluate the Quality of Information (QoI) as a function of the proportion of positive ratings, total number of ratings, and amortized proportion of uncertain ratings. The proposed model exploits Bayesian inference, and a dual weighted regression model to compute the QoI of any published information. We evaluate the proposed model through an experimental study assuming a crowd sourced-urban application as a proof of concept. Experimental results show that compared with the state-of-the-art Jøsangs belief model, the resultant QoI score is less susceptible to rogue ratings and captures subtle differences between true and false information.