Search

Effective communication within universities is crucial for addressing the diverse information needs of students, alumni, and external stakeholders. However, existing chatbot systems often fail to deliver accurate, context-specific responses, resulting in poor user experiences. In this paper, we present KatzBot, an innovative chatbot powered by KatzGPT, a custom Large Language Model (LLM) fine-tuned on domain-specific academic data. KatzGPT is trained on two university-specific datasets: 6,280 sentence-completion pairs and 7,330 question-answer pairs. KatzBot outperforms established existing open source LLMs, achieving higher accuracy and domain relevance. KatzBot offers a user-friendly interface, significantly enhancing user satisfaction in real-world applications. The source code is publicly available at \url{https://github.com/AiAI-99/katzbot}.

Ransomware presents a significant and increasing threat to individuals and organizations by encrypting their systems and not releasing them until a large fee has been extracted. To bolster preparedness against potential attacks, organizations commonly conduct red teaming exercises, which involve simulated attacks to assess existing security measures. This paper proposes a novel approach utilizing reinforcement learning (RL) to simulate ransomware attacks. By training an RL agent in a simulated environment mirroring real-world networks, effective attack strategies can be learned quickly, significantly streamlining traditional, manual penetration testing processes. The attack pathways revealed by the RL agent can provide valuable insights to the defense team, helping them identify network weak points and develop more resilient defensive measures. Experimental results on a 152-host example network confirm the effectiveness of the proposed approach, demonstrating the RL agent's capability to discover and orchestrate attacks on high-value targets while evading honeyfiles (decoy files strategically placed to detect unauthorized access).

We investigate the novel features of gravitational wave solutions in $f(R)$ gravity under proper gauge considerations in the shifted Ricci scalar background curvature ($R^{1+\epsilon}$). The solution is further explored to study the modified dispersion relations for massive modes at local scales and to derive constraints on $\epsilon$. Our analysis yields new insights as we scrutinize these dispersion effects on the polarization (modified Newman-Penrose content) and lensing properties of gravitational waves. It is discovered that the existing longitudinal scalar mode, and transverse breathing scalar mode are both independent of the mass parameter for $\epsilon<<1$. Further, by analysing the lensing amplification factor for the point mass lens model, we show that lensing of gravitational wave is highly sensitive to these dispersion effects in the milli-Hertz frequency (wave optics regime). It is expected that ultra-light modes, having mass about $\mathcal{O} (10^{-15})$ eV for $\epsilon<<1 (\approx 10^{-7})$ lensed by ($10^3\leq M_{Lens}\leq 10^6$)$M_\odot$ compact objects are likely to be detected by the advanced gravitational wave space-borne detectors, particularly within LISA's (The Laser Interferometer Space Antenna) sensitivity band., Comment: 16 pages, 10 figures

We present a novel approach to identify ransomware campaigns derived from attack timelines representations within victim networks. Malicious activity profiles developed from multiple alert sources support the construction of alert graphs. This approach enables an effective and scalable representation of the attack timelines where individual nodes represent malicious activity detections with connections describing the potential attack paths. This work demonstrates adaptability to different attack patterns through implementing a novel method for parsing and classifying alert graphs while maintaining efficacy despite potentially low-dimension node features., Comment: Preprint. Under Review

The increasing reliance on the internet has led to the proliferation of a diverse set of web-browsers and operating systems (OSs) capable of browsing the web. User agent strings (UASs) are a component of web browsing that are transmitted with every Hypertext Transfer Protocol (HTTP) request. They contain information about the client device and software, which is used by web servers for various purposes such as content negotiation and security. However, due to the proliferation of various browsers and devices, parsing UASs is a non-trivial task due to a lack of standardization of UAS formats. Current rules-based approaches are often brittle and can fail when encountering such non-standard formats. In this work, a novel methodology for parsing UASs using Multi-Headed Attention Based transformers is proposed. The proposed methodology exhibits strong performance in parsing a variety of UASs with differing formats. Furthermore, a framework to utilize parsed UASs to estimate the vulnerability scores for large sections of publicly visible IT networks or regions is also discussed. The methodology present here can also be easily extended or deployed for real-time parsing of logs in enterprise settings., Comment: Accepted to the International Conference on Machine Learning and Cybernetics (ICMLC) 2023

We investigate the formation of the large scale structures in the present accelerated era in $f(R)$ gravity background. This is done by considering the linear growth of matter perturbations at low redshift $z<1$. The effect of $f(R)$ alters the behaviour of the matter density perturbations from the matter dominated universe to the late-time accelerated universe which is encoded in the Newtonian gravitational constant as $G\rightarrow G_{eff}$. The modified gravitational constant ($G_{eff}$) depends on the form of $f(R)$. The late-time accelerated expansion affects the formation of large scale structures by slowing down the growth of matter density. On the other hand, $f(R)$ increases the growth rate of the matter density perturbations. We have found that the source term in $f(R)$ background, $G_{eff}\Omega_m$ overcomes the accelerated expansion and the effect of accelerated expansion suppresses the formation of the large scale structures in the asymptotic future., Comment: 8 pages, 6 figures

We explore the shifted $f(R) (\propto R^{1+\delta})$ model with ${\delta}$ as a distinguishing physical parameter for the study of constraints at local scales. The corresponding dynamics confronted with different geodesics (null and non-null) along with its conformal analogue is investigated. For null geodesics, we discuss the light deflection angle, whereas for non-null geodesics under the weak field limit, we investigate the perihelion advance of the Mercury orbit in $f(R)$ Schwarzschild background, respectively. The extent of an additional force, appearing for non-null geodesics, depends on $\delta$. Such phenomenological investigations allow us to strictly constrain $\delta$ to be approximately $\mathcal{O}(10^{-6})$ with a difference of unity in orders at galactic and planetary scales and seems to provide a unique $f(R)$ at local scales. Further, at late cosmic time, we analyse the constraint on $\delta$ via the bare scalar self-interaction Einstein frame potential to provide a null test of dark energy. We constrain the deviation parameter, $\mid\delta\mid$ to $(\approx 0.6)$ which is in a close agreement with the results obtained through various observations in the Jordan frame by several authors. Our results suggest that the present form of model is suitable for the alternate explanation of dark matter-like effects at local scales, whereas at large scales the deviations grow higher and must be addressed in terms of the accelerated background., Comment: 14 pages, 8 figures

We explore a new realisation of the galactic scale dynamics via gravitational lensing phenomenon in power-law $f(R)$ gravity theory of the type $f(R)\propto R^{1+\delta}$ with $\delta<<1$ for interpreting the clustered dark matter effects. We utilize the single effective point like potential (Newtonian potential + $f(R)$ background potential) obtained under the weak field limit to study the combined observations of galaxy rotation curve beyond the optical disk size and their lensing profile in $f(R)$ frame work. We calculate the magnitude of light deflection angle with the characteristic length scale (because of Noether symmetry in $f(R)$ theories) appearing in the effective $f(R)$ rotational velocity profile of a typical galaxy with the model parameter $\delta \approx O(10^{-6})$ constrained in previous work. For instance, we work with the two nearby controversial galaxies NGC 5533 and NGC 4138 and explore their galactic features by analysing the lensing angle profiles in $f(R)$ background. We also contrast the magnitudes of $f(R)$ lensing angle profiles and the relevant parameters of such galaxies with the generalised pseudo-isothermal galaxy halo model and find consistency., Comment: 7 pages, 5 figures

We investigate a form of $ f(R) = {R^{1+\delta}}/{R_c^{\delta}}$ and study the viability of the model for inflation in the Jordan and the Einstein frames. This model is further analysed by using the power spectrum indices of the inflation and the reheating temperature. During the inflationary evolution, the model predicts a value of $\delta$ parameter very close to one ($\delta=0.98$), while the reheating temperature $T_{re} \sim 10^{17}$ GeV at $\delta=0.98$ is consistent with the standard approach to inflation and observations. We calculate the slow roll parameters for the minimally coupled scalar field within the framework of our model. It is found that the values of the scalar spectral index and tensor-to-scalar ratio are very close to the recent observational data, including those released by Planck 2018. We also show that the Jordan and the Einstein frames are equivalent when $\delta \sim 1 $ by using the scalar spectral index, tensor-to-scalar ratio and reheating temperature., Comment: 14 pages, 13 figures

An attempt has been made to explore the galactic dynamics via the rotational velocity beyond the Einstein's geometric theory of gravity. It is inspired from the geometric relation obtained in the power law $f(R)$ gravity model in vacuum. We analyse the action with a small positive deviation from the Einstein-Hilbert action (taking $R$ as $f(R)\propto R^{1+\delta}$) at the galactic scales for the explanation of cosmological dark matter problem and obtain the contribution of dynamical $f(R)$ background geometry in accelerating the test mass. In the weak field limits, we obtain the effective acceleration of the test mass due to a massive spherically symmetric source in $f(R)$ background and develop an equation for the rotational velocity. We test the viability of the model by tracing the motion of test mass outside the typical galactic visible boundaries without considering any dark matter halo profile. We obtain a nice agreement in the outer regions (up to few tens of kpc beyond the visible boundary) of the typical galaxy by using the known galaxy data.\\ We further explore the galactic dynamics for a galaxy NGC 1052 of which the dark matter deficient galaxies, i.e., DF2 and DF4 are a part (satellite galaxies) and discuss plots of the dynamical feature of rotation curves in $f(R)$ background for the model parameter $\delta<<1$ and interpret the results for its satellite galaxies., Comment: 7 pages, 5 figures

A significant issue in training deep neural networks to solve supervised learning tasks is the need for large numbers of labelled datapoints. The goal of semi-supervised learning is to leverage ubiquitous unlabelled data, together with small quantities of labelled data, to achieve high task performance. Though substantial recent progress has been made in developing semi-supervised algorithms that are effective for comparatively small datasets, many of these techniques do not scale readily to the large (unlaballed) datasets characteristic of real-world applications. In this paper we introduce a novel approach to scalable semi-supervised learning, called Local Label Propagation (LLP). Extending ideas from recent work on unsupervised embedding learning, LLP first embeds datapoints, labelled and otherwise, in a common latent space using a deep neural network. It then propagates pseudolabels from known to unknown datapoints in a manner that depends on the local geometry of the embedding, taking into account both inter-point distance and local data density as a weighting on propagation likelihood. The parameters of the deep embedding are then trained to simultaneously maximize pseudolabel categorization performance as well as a metric of the clustering of datapoints within each psuedo-label group, iteratively alternating stages of network training and label propagation. We illustrate the utility of the LLP method on the ImageNet dataset, achieving results that outperform previous state-of-the-art scalable semi-supervised learning algorithms by large margins, consistently across a wide variety of training regimes. We also show that the feature representation learned with LLP transfers well to scene recognition in the Places 205 dataset.