Rajesh Titung

Rajesh Titung

Auto-generated profile, based on publicly available information

Rajesh Titung

Golisano College of Computing and Information Sciences, Rochester Institute of Technology, Rochester, NY, USA

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
29 September 2025 in IEEE Access
Verified
Journal Article
Open access
FOFaithful Chiagoziem Onwuegbuche
Faithful Chiagoziem Onwuegbuche
RTRajesh Titung
Rajesh Titung
EREsa Rantanen
Esa Rantanen
Anca Delia JurcutAnca Delia Jurcut
Anca Delia Jurcut
CACecilia O. Alm
Cecilia O. Alm
LPLiliana Pasquale
Liliana Pasquale
Cyberattacks often begin with a phishing email designed to exploit human factors, widely regarded as the weakest link in cybersecurity defences. Although previous research has identified factors that make individuals susceptible to scams and fraud, it is still unclear how phishing emails exploit these factors to trigger victims’ emotional responses. This paper addresses the following major questions: Which affective states are commonly exploited in phishing emails? Can they be reliably detected by Large Language Models (LLMs), and how do these models compare to human annotators? In this paper, we identify phishing-specific susceptibility factors inspired by interdisciplinary research literature from areas such as security, human-computer interaction (HCI), and psychology to explain the emotional responses (affective states) attackers seek to trigger in phishing emails.We conducted a pilot study and found that affective states can be key differentiators between phishing and non-phishing emails. Affective states commonly used in phishing emails, such as fear and greed, are almost absent in non-phishing emails. Additionally, phishing emails evoke higher emotional arousal levels than non-phishing emails in both subjects and bodies. Furthermore, we explore the capability of several LLMs (GPT-4, Llama 3, and Gemini Pro) in detecting affective states in over 5,000 phishing emails targeting six universities. Using reliability statistics, we compared LLMs’ performance with human annotators’ performance, mapped exploited affective states to the valence-arousal Cartesian space, examined LLMs’ performance detecting phishing emails with special characters, and analyzed LLM hallucinations. We found that attackers commonly exploit curiosity, trust, and urgency in phishing email subjects and fear, urgency, and trust in their bodies. While fear and urgency elevate arousal levels, the valence of the email body generally remains neutral. Key affective state pairings include fear-urgency, urgency-trust, and trust-curiosity. Among LLMs, GPT4 outperforms Llama 3 and GeminiPro for both body and subject affective states, followed by Llama 3 and GeminiPro. These findings highlight the need for further research on affective states in phishing emails to improve detection systems.

6 March 2022
Verified
Conference Paper
Open access
CACecilia O. Alm
Cecilia O. Alm
RTRajesh Titung
Rajesh Titung
RBReynold Bailey
Reynold Bailey
Were assessments of undergraduate researchers in a 10-week summer computing research experience impacted by the pandemic? We compare three cohort years: (1) pre-pandemic (in-person REU; prior to pandemic onset), (2) in-pandemic (remote REU, post-onset during ongoing pandemic), and (3) post-pandemic (in-person REU, post-onset with pandemic in the background). We discuss two forms of 5-point assessment ratings. First, we examine assessments of research skills on 34 questions, with a repeated measure of 3 assessments per cohort year at the beginning, middle, and end of their experience. Then, we examine assessment of presentation skills collected at the beginning vs. the end of the experience for pairs of students in all cohorts, considering 13 rating questions. Students' performance was assessed higher pre-pandemic. Also, being remote impacted completion performance. Lastly, effects linger after a return to in-person experiences, indicating adjustment challenges.

10 September 2023
Verified
Conference Paper
Closed access
IAIsabelle Arthur
Isabelle Arthur
JQJordan Quinn
Jordan Quinn
RTRajesh Titung
Rajesh Titung
CACecilia O. Alm
Cecilia O. Alm
RBReynold Bailey
Reynold Bailey
Static data visualizations are pervasive across STEM research disciplines. However, as the complexity of datasets continues to grow, often incorporating multiple modalities, the limitations of static visualization become apparent. Multimodal datasets require new tools for data exploration and visualization. In response to the limited capabilities of prior alternatives which tend to be less flexible or center on project-specific visualizations for data exploration, we are developing the Multimodal Data Explorer (MDE) - a novel application which is capable of exploring custom time-series multimodal data in context, providing a comprehensive solution that can enable a deeper understanding of the relationships between multiple streams of data. In addition, the client-server design of MDE facilitates multiple users’ data exploration at the same time.

18 February 2026 in IEEE Transactions on Affective Computing
Verified
Journal Article
Closed access
RTRajesh Titung
Rajesh Titung
CACecilia O. Alm
Cecilia O. Alm
Multimodal human conversational interactions are characterized by individual differences, ambiguity, and complex socio-cultural and situational dynamics. Modeling session structured affective interactions, including emotion expression during communicative exchanges presents challenges for centralized machine learning approaches. Traditional machine learning pipelines for identifying these behaviors typically rely on learning a single generalizable model across all individuals in the dataset. However, data heterogeneity suggests that a single centralized model cannot capture individuals' respective nuances. To confirm this, we analyze multimodal behavioral datasets of expressive emotive interactions and demonstrate the statistical heterogeneity in individuals' data distributions. This serves as motivation for our focus on personalization. To achieve that, we turn to federated learning (FL). FL offers a compelling framework that enables integrating a centralized model (for shared behavioral trends) and client models (for individualized behavioral expressions), while additionally offering privacy preservation. However, so far, FL research has often focused on image benchmarks with non-IID partitioned data among the clients, ignoring the heterogeneity of naturalistic interactions. To address this gap, we propose FedSession, a novel personalized federated learning (PFL) strategy that accommodates heterogeneous data distributions by explicitly modeling sessions as structured interaction units, where each client corresponds to a real individual participating in a conversation-based affective interaction. FedSession incorporates an additional tier to capture how interlocutors align their behaviors to one another in interactions. Experiments on affect recognition tasks involving dyadic sessions demonstrate that FedSession outperforms existing PFL strategies. Unlike prior PFL strategies, which often struggle to outperform even localized training performance, FedSession effectively addresses the challenges of personalization in heterogeneous, interactive emotion recognition settings.