Biblio

In this short paper, we survey some work at the intersection of Artificial Intelligence (AI) and security that are based on game theoretic considerations, and particularly focus on the author's (our) contribution in these areas. One half of this paper focuses on applications of game theoretic and learning reasoning for addressing security applications such as in public safety and wildlife conservation. In the second half, we present recent work that attacks the learning components of these works, leading to sub-optimal defense allocation. We finally end by pointing to issues and potential research problems that can arise due to data quality in the real world.

Artificial intelligence (AI) has been successfully employed in industries for decades, beginning with the invention of expert systems in the 1960s and continuing through the present ubiquity of deep learning. Data-driven AI solutions have grown increasingly common as a means of supporting ever-more complicated industrial processes owing to the accessibility of affordable computer and storage infrastructure. Despite recent optimism, implementing AI to smart industrial applications still offers major difficulties. The present paper gives an executive summary of AI methodologies with an emphasis on deep learning before detailing unresolved issues in AI safety, data privacy, and data quality — all of which are necessary for completely automated commercial AI systems.

Cloud Computing revolutionize the usage of Internet of Things enabled devices integrated via Internet. Providing everything in an outsourced fashion, Cloud also lends infrastructures such as storage. Though cloud makes it easy for us to store and access the data faster and easier, yet there exist various security and privacy risks. Such issues if not handled may become more threatening as it could even disclose the privacy of an individual/ organization. Strengthening the security of data is need of the hour. The work proposes a novel architecture enhancing the security of Cloud data in an IoT integrated environment. In order to enhance the security, systematic use of a modified hybrid mechanism based on DNA code and Elliptic Curve Cryptography along with Third Party Audit is proposed. The performance of the proposed mechanism has been analysed. The results ensures that proposed IoT Cloud architecture performs better while providing strong security which is the major aspect of the work.

Machine learning (ML) models are increasingly being used in the development of Malware Detection Systems. Existing research in this area primarily focuses on developing new architectures and feature representation techniques to improve the accuracy of the model. However, recent studies have shown that existing state-of-the art techniques are vulnerable to adversarial machine learning (AML) attacks. Among those, data poisoning attacks have been identified as a top concern for ML practitioners. A recent study on clean-label poisoning attacks in which an adversary intentionally crafts training samples in order for the model to learn a backdoor watermark was shown to degrade the performance of state-of-the-art classifiers. Defenses against such poisoning attacks have been largely under-explored. We investigate a recently proposed clean-label poisoning attack and leverage an ensemble-based Nested Training technique to remove most of the poisoned samples from a poisoned training dataset. Our technique leverages the relatively large sensitivity of poisoned samples to feature noise that disproportionately affects the accuracy of a backdoored model. In particular, we show that for two state-of-the art architectures trained on the EMBER dataset affected by the clean-label attack, the Nested Training approach improves the accuracy of backdoor malware samples from 3.42% to 93.2%. We also show that samples produced by the clean-label attack often successfully evade malware classification even when the classifier is not poisoned during training. However, even in such scenarios, our Nested Training technique can mitigate the effect of such clean-label-based evasion attacks by recovering the model's accuracy of malware detection from 3.57% to 93.2%.

Data poisoning is a type of adversarial attack on training data where an attacker manipulates a fraction of data to degrade the performance of machine learning model. There are several known defensive mechanisms for handling offline attacks, however defensive measures for online learning, where data points arrive sequentially, have not garnered similar interest. In this work, we propose a defense mechanism to minimize the degradation caused by the poisoned training data on a learner's model in an online setup. Our proposed method utilizes an influence function which is a classic technique in robust statistics. Further, we supplement it with the existing data sanitization methods for filtering out some of the poisoned data points. We study the effectiveness of our defense mechanism on multiple datasets and across multiple attack strategies against an online learner.

A direct access (DAX) file system maximizes the benefit of persistent memory(PM)’s low latency through removing the page cache layer from the file system access paths. However, this paper reveals that data block allocation of the DAX file systems in common is significantly slower than that of conventional file systems because the DAX file systems require the zero-out operation for the newly allocated blocks to prevent the leakage of old data previously stored in the allocated data blocks. The retarded block allocation significantly affects the file write performance. In addition to this revelation, this paper proposes an off-critical-path data block sanitization scheme tailored for DAX file systems. The proposed scheme detaches the zero-out operation from the latency-critical I/O path and performs that of released data blocks in the background. The proposed scheme’s design principle is universally applicable to most DAX file systems. For evaluation, we implemented our approach in Ext4-DAX and XFS-DAX. Our evaluation showed that the proposed scheme reduces the append write latency by 36.8%, and improved the performance of FileBench’s fileserver workload by 30.4%, YCSB’s workload A on RocksDB by 3.3%, and the Redis-benchmark by 7.4% on average, respectively.

With the advent of the era of big data, the files that need to be stored in the storage system will increase exponentially. Cloud storage has become the most popular data storage method due to its powerful convenience and storage capacity. However, in order to save costs, some cloud service providers, Malicious deletion of the user's infrequently accessed data causes the user to suffer losses. Aiming at data integrity and privacy issues, a blockchain-based cloud storage integrity verification scheme for recoverable data is proposed. The scheme uses the Merkle tree properties, anonymity, immutability and smart contracts of the blockchain to effectively solve the problems of cloud storage integrity verification and data damage recovery, and has been tested and analyzed that the scheme is safe and effective.

In today’s world, digital data are enormous due to technologies that advance data collection, storage, and analyses. As more data are shared or publicly available, privacy is of great concern. Having privacy means having control over your data. The first step towards privacy protection is to understand various aspects of privacy and have the ability to quantify them. Much work in structured data, however, has focused on approaches to transforming the original data into a more anonymous form (via generalization and suppression) while preserving the data integrity. Such anonymization techniques count data instances of each set of distinct attribute values of interest to signify the required anonymity to protect an individual’s identity or confidential data. While this serves the purpose, our research takes an alternative approach to provide quick privacy measures by way of anonymity especially when dealing with large-scale data. This paper presents a study of anonymity measures based on their relevant properties that impact privacy. Specifically, we identify three properties: uniformity, variety, and diversity, and formulate their measures. The paper provides illustrated examples to evaluate their validity and discusses the use of multi-aspects of anonymity and privacy measures.

Cloud computing has become an integral part of medical big data. The cloud has the capability to store the large data volumes has attracted more attention. The integrity and privacy of patient data are some of the issues that cloud-based medical big data should be addressed. This research work introduces data integrity auditing scheme for cloud-based medical big data. This will help minimize the risk of unauthorized access to the data. Multiple copies of the data are stored to ensure that it can be recovered quickly in case of damage. This scheme can also be used to enable doctors to easily track the changes in patients' conditions through a data block. The simulation results proved the effectiveness of the proposed scheme.

The evolving and new age cybersecurity threats has set the information security industry on high alert. This modern age cyberattacks includes malware, phishing, artificial intelligence, machine learning and cryptocurrency. Our research highlights the importance and role of Software Quality Assurance for increasing the security standards that will not just protect the system but will handle the cyber-attacks better. With the series of cyber-attacks, we have concluded through our research that implementing code review and penetration testing will protect our data's integrity, availability, and confidentiality. We gathered user requirements of an application, gained a proper understanding of the functional as well as non-functional requirements. We implemented conventional software quality assurance techniques successfully but found that the application software was still vulnerable to potential issues. We proposed two additional stages in software quality assurance process to cater with this problem. After implementing this framework, we saw that maximum number of potential threats were already fixed before the first release of the software.

With the advent of cloud storage services many users tend to store their data in the cloud to save storage cost. However, this has lead to many security concerns, and one of the most important ones is ensuring data integrity. Public verification schemes are able to employ a third party auditor to perform data auditing on behalf of the user. But most public verification schemes are vulnerable to procrastinating auditors who may not perform auditing on time. These schemes do not have fair arbitration also, i.e. they lack a way to punish the malicious Cloud Service Provider (CSP) and compensate user whose data has been corrupted. On the other hand, CSP might be storing redundant data that could increase the storage cost for the CSP and computational cost of data auditing for the user. In this paper, we propose a Blockchain-based public auditing and deduplication scheme with a fair arbitration system against procrastinating auditors. The key idea requires auditors to record each verification using smart contract and store the result into a Blockchain as a transaction. Our scheme can detect and punish the procrastinating auditors and compensate users in the case of any data loss. Additionally, our scheme can detect and delete duplicate data that improve storage utilization and reduce the computational cost of data verification. Experimental evaluation demonstrates that our scheme is provably secure and does not incur overhead compared to the existing public auditing techniques while offering an additional feature of verifying the auditor’s performance.

Network Intrusion Detection Systems (IDSs) have been used to increase the level of network security for many years. The main purpose of such systems is to detect and block malicious activity in the network traffic. Researchers have been improving the performance of IDS technology for decades by applying various machine-learning techniques. From the perspective of academia, obtaining a quality dataset (i.e. a sufficient amount of captured network packets that contain both malicious and normal traffic) to support machine learning approaches has always been a challenge. There are many datasets publicly available for research purposes, including NSL-KDD, KDDCUP 99, CICIDS 2017 and UNSWNB15. However, these datasets are becoming obsolete over time and may no longer be adequate or valid to model and validate IDSs against state-of-the-art attack techniques. As attack techniques are continuously evolving, datasets used to develop and test IDSs also need to be kept up to date. Proven performance of an IDS tested on old attack patterns does not necessarily mean it will perform well against new patterns. Moreover, existing datasets may lack certain data fields or attributes necessary to analyse some of the new attack techniques. In this paper, we argue that academia needs up-to-date high-quality datasets. We compare publicly available datasets and suggest a way to provide up-to-date high-quality datasets for researchers and the security industry. The proposed solution is to utilize the network traffic captured from the Locked Shields exercise, one of the world’s largest live-fire international cyber defence exercises held annually by the NATO CCDCOE. During this three-day exercise, red team members consisting of dozens of white hackers selected by the governments of over 20 participating countries attempt to infiltrate the networks of over 20 blue teams, who are tasked to defend a fictional country called Berylia. After the exercise, network packets captured from each blue team’s network are handed over to each team. However, the countries are not willing to disclose the packet capture (PCAP) files to the public since these files contain specific information that could reveal how a particular nation might react to certain types of cyberattacks. To overcome this problem, we propose to create a dedicated virtual team, capture all the traffic from this team’s network, and disclose it to the public so that academia can use it for unclassified research and studies. In this way, the organizers of Locked Shields can effectively contribute to the advancement of future artificial intelligence (AI) enabled security solutions by providing annual datasets of up-to-date attack patterns.

The integration of distributed energy resources (DERs) and expansion of complex network in the distribution grid requires an advanced two-level state estimator to monitor the grid health at micro-level. The distribution state estimator will improve the situational awareness and resiliency of distributed power system. This paper implements a synchrophasors-based master state awareness (MSA) estimator to enhance the cybersecurity in distribution grid by providing a real-time estimation of system operating states to control center operators. In this paper, the implemented MSA estimator utilizes only phasor measurements, bus magnitudes and angles, from phasor measurement units (PMUs), deployed in local substations, to estimate the system states and also detects data integrity attacks, such as load tripping attack that disconnects the load. To validate the proof of concept, we implement this methodology in cyber-physical testbed environment at the Idaho National Laboratory (INL) Electric Grid Security Testbed. Further, to address the "valley of death" and support technology commercialization, field demonstration is also performed at the Critical Infrastructure Test Range Complex (CITRC) at the INL. Our experimental results reveal a promising performance in detecting load tripping attack and providing an accurate situational awareness through an alert visualization dashboard in real-time.

The cloud provides storage for users to share their files in the cloud. Nowadays some shared data auditing schemes are proposed for protecting data integrity. However, preserving the identity privacy of group users and secure user revocation usually result in high computational overhead. Then a shared data auditing scheme supporting identity privacy preserving is proposed that enables users to be effectively revoked. To preserve identity privacy during the audit process, we develop an efficient authenticator generation mechanism that enables public auditing. Our solution supports efficient user revocation, where the authenticator of the revoked user does not need to be regenerated and integrity checking can be performed appropriately. At the same time, the group manager maintains two tables to ensure user traceability. When the user updates data, two tables are modified and updated by the group manager promptly. It shows that our scheme is secure by security analysis. Moreover, concrete experiments prove the performance of the system.

Cloud data integrity verification was an important means to ensure data security. We used public key infrastructure (PKI) to manage user keys in Traditional way, but there were problems of certificate verification and high cost of key management. In this paper, RSA signature was used to construct a new identity-based cloud audit protocol, which solved the previous problems caused by PKI and supported forward security, and reduced the loss caused by key exposure. Through security analysis, the design scheme could effectively resist forgery attack and support forward security.

Graph-based Semi-Supervised Learning (GSSL) is a practical solution to learn from a limited amount of labelled data together with a vast amount of unlabelled data. However, due to their reliance on the known labels to infer the unknown labels, these algorithms are sensitive to data quality. It is therefore essential to study the potential threats related to the labelled data, more specifically, label poisoning. In this paper, we propose a novel data poisoning method which efficiently approximates the result of label inference to identify the inputs which, if poisoned, would produce the highest number of incorrectly inferred labels. We extensively evaluate our approach on three classification problems under 24 different experimental settings each. Compared to the state of the art, our influence-driven attack produces an average increase of error rate 50% higher, while being faster by multiple orders of magnitude. Moreover, our method can inform engineers of inputs that deserve investigation (relabelling them) before training the learning model. We show that relabelling one-third of the poisoned inputs (selected based on their influence) reduces the poisoning effect by 50%. ACM Reference Format: Adriano Franci, Maxime Cordy, Martin Gubri, Mike Papadakis, and Yves Le Traon. 2022. Influence-Driven Data Poisoning in Graph-Based Semi-Supervised Classifiers. In 1st Conference on AI Engineering - Software Engineering for AI (CAIN’22), May 16–24, 2022, Pittsburgh, PA, USA. ACM, New York, NY, USA, 11 pages. https://doi.org/10.1145/3522664.3528606

Sybil attack is one of the most dangerous internal attacks in Vehicular Ad Hoc Network (VANET). It affects the function of the VANET network by maliciously claiming or stealing multiple identity propagation error messages. In order to prevent VANET from Sybil attacks, many solutions have been proposed. However, the existing solutions are specific to the physical or application layer's single-level data and lack research on cross-layer information fusion detection. Moreover, these schemes involve a large number of sensitive data access and transmission, do not consider users' privacy, and can also bring a severe communication burden, which will make these schemes unable to be actually implemented. In this context, this paper introduces FedMix, the first federated Sybil attack detection system that considers cross-layer information fusion and provides privacy protection. The system can integrate VANET physical layer data and application layer data for joint analyses simultaneously. The data resides locally in the vehicle for local training. Then, the central agency only aggregates the generated model and finally distributes it to the vehicles for attack detection. This process does not involve transmitting and accessing any vehicle's original data. Meanwhile, we also designed a new model aggregation algorithm called SFedAvg to solve the problems of unbalanced vehicle data quality and low aggregation efficiency. Experiments show that FedMix can provide an intelligent model with equivalent performance under the premise of privacy protection and significantly reduce communication overhead, compared with the traditional centralized training attack detection model. In addition, the SFedAvg algorithm and cross-layer information fusion bring better aggregation efficiency and detection performance, respectively.

Zero trust security model has been picking up adoption in various organizations due to its various advantages. Data quality is still one of the fundamental challenges in data curation in many organizations where data consumers don’t trust data due to associated quality issues. As a result, there is a lack of confidence in making business decisions based on data. We design a model based on the zero trust security model to demonstrate how the trust of data consumers can be established. We present a sample application to distinguish the traditional approach from the zero trust based data quality framework.

Modern IoT infrastructure consists of different sub-systems, devices, applications, platforms, varied connectivity protocols with distinct operating environments scattered across different subsystems within the whole network. Each of these subsystems of the global system has its peculiar computational and security challenges. A security loophole in one subsystem has a directly negative impact on the security of the whole system. The nature and intensity of recent cyber-attacks within IoT networks have increased in recent times. Blockchain technology promises several security benefits including a decentralized authentication mechanism that addresses almost readily the challenges with a centralized authentication mechanism that has the challenges of introducing a single point of failure that affects data and system availability anytime such systems are compromised. The different design specifications and the unique functional requirements for most IoT devices require a strong yet universal authentication mechanism for multimedia data that assures an additional security layer to IoT data. In this paper, the authors propose a decentralized authentication to validate data integrity at the IoT node level. The proposed mechanism guarantees integrity, privacy, and availability of IoT node data.

The blockchain technology evolution in recent times has a hopefulness regarding the impression of self-sovereign identity that has a significant effect on the method of interacting with each other with security over the network. The existing system is not complete and procedural. There arises a different idea of self-sovereign identity methodology. To develop to the possibility, it is necessary to guarantee a better understanding in a proper way. This paper has an in-depth analysis of the attributes of the self-sovereign identity and it affects over the laws of identity that are being explored. The Identity management system(IMS) with no centralized authority is proposed in maintaining the secrecy of records, where as traditional systems are replaced by blockchains and identities are generated cryptographically. This study enables sharing of user data on permissioned blockchain which uses identity-based encryption to maintain access control and data security.

In this paper, we present our ongoing work to develop an efficient and scalable verification method to achieve runtime security of real-time applications with strict performance requirements. The method allows to specify (functional and non-functional) behaviour of a real-time application and a set of known attacks/threats. The challenge here is to prove that the runtime application execution is at the same time (i) correct w.r.t. the functional specification and (ii) protected against the specified set of attacks, without violating any non-functional specification (e.g., real-time performance). To address the challenge, first we classify the set of attacks into computational, data integrity and communication attacks. Second, we decompose each class into its declarative properties and definitive properties. A declarative property specifies an attack as a one big-step relation between initial and final state without considering intermediate states, while a definitive property specifies an attack as a composition of many small-step relations considering all intermediate states between initial and final state. Semantically, the declarative property of an attack is equivalent to its corresponding definitive property. Based on the decomposition and the adequate specification of underlying runtime environment (e.g., compiler, processor and operating system), we prove rigorously that the application execution in a particular runtime environment is protected against declarative properties without violating runtime performance specification of the application. Furthermore, from the specification, we generate a security monitor that assures that the application execution is secure against each class of attacks at runtime without hindering real-time performance of the application.

In the big data cloud computing environment, data security issues have become a focus of attention. This paper delivers an overview of conceptions, characteristics and advanced technologies for big data cloud computing. Security issues of data quality and privacy control are elaborated pertaining to data access, data isolation, data integrity, data destruction, data transmission and data sharing. Eventually, a virtualization architecture and related strategies are proposed to against threats and enhance the data security in big data cloud environment.

Data is one of the most valuable assets of an organization and has a tremendous impact on its long-term success and decision-making processes. Typically, organizational data error and outlier detection processes perform manually and reactively, making them time-consuming and prone to human errors. Additionally, rich data types, unlabeled data, and increased volume have made such data more complex. Accordingly, an automated anomaly detection approach is required to improve data management and quality control processes. This study introduces an unsupervised anomaly detection approach based on models comparison, consensus learning, and a combination of rules of thumb with iterative hyper-parameter tuning to increase data quality. Furthermore, a domain expert is considered a human in the loop to evaluate and check the data quality and to judge the output of the unsupervised model. An experiment has been conducted to assess the proposed approach in the context of a case study. The experiment results confirm that the proposed approach can improve the quality of organizational data and facilitate anomaly detection processes.

Rapid development of cloud storage services, users are allowed to upload heavy storage and computational cost to cloud to reduce the local resource and energy consumption. While people enjoy the desirable benefits from the cloud storage service, critical security concerns in data outsourcing have been raised seriously. In the cloud storage service, data owner loses the physical control of the data and these data are fully controlled by the cloud server. As such, the integrity of outsourced data is being put at risk in reality. Remote data integrity checking (RDIC) is an effective solution to checking the integrity of uploaded data. However, most RDIC schemes are rely on traditional public key infrastructure (PKI), which leads communication and storage overhead due to the certificate management. Identity-based RDIC scheme is not need the storage management, but it has a drawback of key escrow. To solve these problems, we propose a practical certificateless RDIC scheme. Moreover, many public auditing schemes authorize the third party auditor (TPA) to check the integrity of remote data and the TPA is not fully trusted. Thus, we take the data privacy into account. The proposed scheme not only can overcome the above deficiencies but also able to preserve the data privacy against the TPA. Our theoretical analyses prove that our mechanism is correct and secure, and our mechanism is able to audit the integrity of cloud data efficiently.

Underwater acoustic sensor networks (UASNs) have been receiving more and more attention due to their wide applications and the marine data collection is one of the important applications of UASNs. However, the openness and unreliability of underwater acoustic communication links and the easy capture of underwater wireless devices make UASNs vulnerable to various attacks. On the other hand, due to the limited resources of underwater acoustic network nodes, the high bit error rates, large and variable propagation delays, and low bandwidth of acoustic channels, many mature security mechanisms in terrestrial wireless sensor networks cannot be applied in the underwater environment [1]. In this paper, a secure communication protocol for marine data collection was proposed to ensure the confidentiality and data integrity of communication between under sensor nodes and the sink node in UASNs.