Abstract

Cyber threat hunting proactively searches for cyber threats, which are undetected by the traditional defense mechanisms. It scans deep to identify malicious programs (ie, malware) that escape from detection. It is important because sophisticated cyber threats can bypass the cyber security mechanisms. The performance of the cyber threat hunting can be improved through artificial intelligence (AI), especially, explainable AI (XAI), which adds trust component to the cyber threat hunting process. Due to the inclusion of XAI, the security experts get the full explanations of the detected threats as the working of the detection model in XAI is known. Information, like, which one is a threat, how it has been detected, and why it has been detected, can be obtained very easily due to the inclusion of XAI in the cyber threat hunting. Therefore, an XAI-envisioned mechanism for cyber threat hunting has been proposed (in short, XAISM-CTH). The network and threat models of XAISM-CTH are designed and discussed. The conducted security analysis proves the security of XAISM-CTH against various potential attacks. XAISM-CTH also performs better than the other existing schemes. At the end, a practical implementation of XAISM-CTH has been provided to observe its impact on the performance of the system. KEYWORDS cyber threat hunting, explainable artificial intelligence (XAI), intrusion detection, privacy, security

Abstract

Due to high effectiveness and robust security proto- cols, lattice-based cryptography becomes a very broadly appli- cable optimistic post-quantum technique that is recently used in public key cryptosystem. An aggregate signature scheme enables a party to bundle a set of signatures together into a single short cryptographic signature, which can be verified by any verifier using the public information. In this paper, we provide a lattice- based aggregate signature scheme where the security depends on the difficulty of the Ring Learning-with-Error (Ring-LWE) problem. Next, we use the basic scheme in Internet of Drones (IoD) applications using the blockchain technology for secure and transparent data storage. The detailed security analysis and comparative study show that the proposed scheme provides superior security including resistance to quantum attacks and is efficient as compared to the existing state of art approaches. The testbed experimental results and the blockchain simulation demonstrate that the proposed scheme can be applied in real-life drones applications. Index Terms—Internet of Drones (IoD), unmanned aerial vehi- cles, lattice-based cryptography, aggregate signature, blockchain, security.

Abstract

Attribute-based encryption (ABE) is widely used for a secure and efficient data sharing. The predetermined access policy of ABE shares the data with intended data users. However, ABE is not preferable in many applications that require collaboration among data users. In such applications, an authorized data user may be interested to collaborate with another data user who does not adhere to the access policy. Fixed access policy of ABE does not allow an authorized data user (who satisfies the access policy) to collaborate or share the data with any unauthorized data user (who fails to satisfy the access policy). Thus, due to the static and predefined access policy, data collaboration in ABE is significantly challenging. In this work,

Abstract

We present three related schemes to generate novel molecules based on seed molecule; all the methods use as input the voxelized representation of the seed molecule to generate the SMILES representation of the novel molecule. At heart of these methods are two networks (a) a variational autoencoder that uses a Riemannian metric to encode the latent space of (hence named RHVAE) and can use pharmacophoric requirements as additional input for decoder and (b) attentive captioning network (a type of recurring neural network) that can efficiently focus, capture and use the ‘content’ of input to generate the SMILES output of novel molecules. We analyze the performance of the three proposed methods. We demonstrate the generation of meaningful new molecules, by generating shapes through an autoencoder network which can then be passed to our attentive captioning network, while requiring smaller data sets for training and retaining similar performance to existing state-of-art methods. Keywords. De Novo Drug Design; Chemical Modeling; Variational autoencoders, Attentive RNNs, Remannian

Abstract

Drought complexity may not be accurately characterized by univariate meteorological or hydrological drought indices under the intensification of hydrological cycle due to climate change and human activities. In particular, such drought indices require long series of hydro-meteorological data, which are unavailable over ungauged and data-scarce catchments. In this study, a multivariate drought index, Standardized Precipitation Actual Evapotranspiration Index-Agro (SPAEI-Agro), is proposed, which com- bines meteorological and hydrological variables as precipitation (P), actual evapotranspiration (AET), runoff and groundwater (GW) of ungauged catchments and sub-catchment scales with diverse climatol- ogy. SPAEI-Agro was able to characterize severe drought events more accurately in humid and dry sub- humid sub-catchments compared to Standardized Precipitation Evapotranspiration Index (SPEI) and Standardized Streamflow Drought Index (SSDI) for Tunga-Bhadra River, India. SPEI (based upon potential evapotranspiration) and SSDI (based on streamflow) showed intensified drought characteristics com- pared to the new P, AET and GW-based drought indicator SPAEI-Agro in semi-arid and dry sub-humid climates.

Abstract

Large language models (LLMs) have been shown to be able to perform new tasks based on a few demonstrations or natural language instructions. While these capabilities have led to widespread adoption, most LLMs are developed by resource-rich organizations and are fre- quently kept from the public. As a step towards democratizing this powerful technology, we present BLOOM, a 176B-parameter open-access language model designed and built thanks to a collaboration of hundreds of researchers. BLOOM is a decoder-only Transformer lan- guage model that was trained on the ROOTS corpus, a dataset comprising hundreds of sources in 46 natural and 13 programming languages (59 in total). We find that BLOOM achieves competitive performance on a wide variety of benchmarks, with stronger results after undergoing multitask prompted finetuning. To facilitate future research and appli- cations using LLMs, we publicly release our models and code under the Responsible AI License.1 Keywords: Language models, collaborative research

Abstract

Radiance Fields (RFs) have shown great potential to represent scenes from casually captured discrete views. Compositing parts or whole of multiple captured scenes could greatly interest several XR applications. Prior works can generate new views of such scenes by tracing each scene in parallel. This increases the render times and memory requirements with the number of components. In this work, we provide a method to create a single, compact, fused RF representation for a scene composited using multiple RFs. The fused RF has the same render times and memory utilizations as a single RF. Our method distills information from multiple teacher RFs into a single student RF while also facilitating further manipulations like addition and deletion into the fused representation.

Abstract

Fairness across different demographic groups is an essential criterion for face-related tasks, Face Attribute Classification (FAC) being a prominent example. Apart from this trend, Federated Learning (FL) is increasingly gaining traction as a scalable paradigm for distributed training. Existing FL approaches require data homogeneity to ensure fairness. However, this assumption is too restrictive in real-world settings. We propose F3, a novel FL framework for fair FAC under data heterogeneity. F3 adopts multiple heuristics to improve fairness across different demographic groups without requiring data homogeneity assumption. We demonstrate the efficacy of F3 by reporting empirically observed fairness measures and accuracy guarantees on popular face datasets. Our results suggest that F3 strikes a practical balance between accuracy and fairness for FAC.

Abstract

Humans have a natural ability to perform semantic associations with the surrounding objects in the environment. This allows them to create a mental map of the environment which helps them to navigate on-demand when given a linguistic instruction. A natural goal in Vision Language Navigation (VLN) research is to impart autonomous agents with similar capabilities. Recently introduced VL Maps \cite{huang23vlmaps} take a step towards this goal by creating a semantic spatial map representation of the environment without any labelled data. However, their representations are limited for practical applicability as they do not distinguish between different instances of the same object. In this work, we address this limitation by integrating instance-level information into spatial map representation using a community detection algorithm and by utilizing word ontology learned by large language models (LLMs) to perform open-set semantic associations in the mapping representation. The resulting map representation improves the navigation performance by two-fold (233\%) on realistic language commands with instance-specific descriptions compared to VL Maps. We validate the practicality and effectiveness of our approach through extensive qualitative and quantitative experiments.

Abstract

Speech intelligibility is crucial in language learning for effective communication. Thus, to develop computer-assisted language learning systems, automatic speech intelligibility detection (SID) is necessary. Most of the works have assessed the intelligibility in a supervised manner considering manual annotations, which requires cost and time; hence scalability is limited. To overcome these, this work proposes an unsupervised approach for SID. The proposed approach considers alignment distance computed with dynamic-time warping (DTW) between teacher and learner representation sequence as a measure to separate intelligible versus non-intelligible speech. We obtain the feature sequence using current state-of-the-art self-supervised representations from Wav2Vec-2.0. We found the detection accuracies as 90.37\%, 92.57\% and 96.58\%, respectively, with three alignment distance measures -- mean absolute error, mean squared error and cosine distance (equal to one minus cosine similarity).