Rakesh Verma

Advances in Vehicle-to-Everything (V2X) technologies promise significant mobility gains through enhanced information accuracy and responsiveness. Prior research has shown that shared traveler information can induce synchronized routing responses and correlated congestion. However, how such synchronized rerouting interacts with the physical constraints of finite link storage in signalized urban networks remains insufficiently understood. This study employs a spillback-consistent mesoscopic simulation that integrates Cell Transmission Model traffic physics with four information architectures (Base, Central, V2V, and Hybrid) to examine the joint evolution of routing behavior and physical queue dynamics. The results reveal a pronounced Performance Paradox: while V2X architectures improve individual routing under light and moderate demand, under heavy demand they trigger a transition to structural gridlock, increasing origin blocking from below 1% to nearly 10%. Unlike previously documented herding phenomena, this degradation arises from a specific interaction between hyper-reactive rerouting and finite storage constraints. Continuous, low-latency updates suppress the natural recovery intervals required for queues to dissipate across signal cycles, causing spillback to persist and propagate upstream. Conversely, centralized systems with discretized update intervals remain more stable by inadvertently creating recovery windows. These findings show that routing information cannot be evaluated independently of the physical queue-dissipation processes of signalized networks and suggest that traveler information systems must be co-designed with infrastructure discharge characteristics rather than optimized solely for timeliness and accuracy.

Introduction: Internet-based economies and societies are drowning in deceptive attacks. These attacks take many forms, such as fake news, phishing, and job scams, which we call “domains of deception.” Machine learning and natural language processing researchers have been attempting to ameliorate this precarious situation by designing domain-specific detectors. Only a few recent works have considered domain-independent deception. We collect these disparate threads of research and investigate domain-independent deception.Methods: First, we provide a new computational definition of deception and break down deception into a new taxonomy. Then, we briefly mention the debate on linguistic cues for deception. We build a new comprehensive real-world dataset for studying deception. We investigate common linguistic features for deception using both classical and deep learning models in a variety of situations including cross-domain experiments.Results: We find common linguistic cues for deception and give significant evidence for knowledge transfer across different forms of deception.Discussion: We list several directions for future work based on our results.

In the modern age an enormous amount of communication occurs online, and it is difficult to know when something written is genuine or deceitful. There are many reasons for someone to deceive online (e.g., monetary gain, political gain) and detecting this behavior without any physical interaction is a difficult task. Additionally, deception occurs in several text-only domains and it is unclear if these various sources can be leveraged to improve detection. To address this, eight datasets were utilized from various domains to evaluate their effect on classifier performance when combined with transfer learning via intermediate layer concatenation of fine-tuned BERT models. We find improvements in accuracy over the baseline. Furthermore, we evaluate multiple distance measurements between datasets and find that Jensen-Shannon distance correlates moderately with transfer learning performance. Finally, the impact was evaluated of multiple methods, which produce additional information in a dataset's text via named entities, on BERT performance and we find notable improvement in accuracy of up to 11.2%.

Despite two decades of research on automatic filtering systems, phishing attacks remain a serious problem. To alleviate risks from filtering failures, we design and evaluate the effectiveness of a new warning system on users' susceptibility to phishing. Our proposed technique highlights key sentences based on an analysis of the persuasive techniques used. An online mixed-design study ( $n=604$ ) shows that adding our highlighting technique outperforms existing warning solutions. It also identifies the relative efficacy of different appeals and the characteristics of susceptible users. Results show that adding our highlighting technique is useful even with false positives and false negatives . Inspired by this result, we propose an automatic warning generator. We created a small labeled dataset of suspicious sentences and used data augmentation. Our best models achieve F1 score of 99.95% in detecting phishing emails and 88% in detecting suspicious sentences.

Deception is increasingly prevalent in the modern world, appearing in many different forms (domains) from phishing emails to fictitious product reviews, or even false political statements. Many researchers are looking for ways to identify deception within these different domains or to characterize deception through new datasets and techniques. Most researchers focus on one domain of deception at a time, however, we are interested in whether domain-independent deception detection is possible. To this end, we recreate a linguistic cue-based feature set proven to be effective in the business-to-business communication (B2B) domain and compare it to state-of-the-art linguistic cue feature sets in five other domains of deception. This B2B feature set was created through LIWC2007, which allows us the unique opportunity to utilize that version alongside the modern version LIWC22, to ascertain how much linguistic cue extraction has improved over the years. Finally, we leverage our best-performing feature set to investigate the efficacy of a technique gaining popularity in recent years, crowd-sourced synthetic datasets.