Akhirul Islam
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Akhirul Islam has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
Manojit Ghose
4 shared publications- Manojit Ghose
Department of Computer Science and Engineering, Indian Institute of Information Technology Guwahati, Guwahati, India
3 shared publications - Alexander Otvagin1 shared publication
Alok Gupta
1 shared publicationArindam Debnath
1 shared publication- Cullen Bash
Hewlett Packard Enterprise, Milpitas, CA, USA
1 shared publication
Latest Publications
- Open Access
- 27 September 2026
- Published by Elsevier in SSRN Electronic Journal
Visual Clustering of Coordinated Phishing Email Campaigns: A Deployed Pipeline with GenAI Summarization
Automated phishing-campaign clustering is critical for mitigating increasingly sophisticated email-based threats. Existing methods rely primarily on textual features, such as the subject, body, and attachment names. However, they capture limited visual information from modern HTML emails, where dynamically rendered images, JavaScript-generated layouts, and visually encoded content such as QR codes and ASCII art may only become apparent after rendering in the recipient's client. We present two image-based clustering strategies that operate on rendered email screenshots: perceptual hash clustering (PHC), a lightweight visual mechanism based on Hamming distance, and deep visual embedding clustering (DVEC), a two-stage pipeline that combines CLIP embeddings with HDBSCAN and supports continuous ingestion through an incremental assignment mechanism. We evaluate both strategies on a real-world dataset of 38,161 phishing emails collected over three months in Trellix labs. PHC and DVEC record higher Silhouette (0.775 and 0.896), lower Davies–Bouldin (0.658 and 0.657), and higher Calinski–Harabasz (1535 and 1541) values than the text-based benchmarks. We further introduce an additional visual baseline, DINOv2+HDBSCAN. On a controlled perturbation evaluation, DVEC obtains the highest recall (0.693) and F1 (0.506). Finally, we integrate a generative model (LLaMA) to produce automated campaign summaries. Visual-based clustering offers a complementary perspective to text-based methods and a scalable foundation for continuous phishing-campaign analysis and summarization.
- 27 September 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessVisual Clustering of Coordinated Phishing Email Campaigns: A Deployed Pipeline with GenAI Summarization
Automated phishing-campaign clustering is critical for mitigating increasingly sophisticated email-based threats. Existing methods rely primarily on textual features, such as the subject, body, and attachment names. However, they capture limited visual information from modern HTML emails, where dynamically rendered images, JavaScript-generated layouts, and visually encoded content such as QR codes and ASCII art may only become apparent after rendering in the recipient's client. We present two image-based clustering strategies that operate on rendered email screenshots: perceptual hash clustering (PHC), a lightweight visual mechanism based on Hamming distance, and deep visual embedding clustering (DVEC), a two-stage pipeline that combines CLIP embeddings with HDBSCAN and supports continuous ingestion through an incremental assignment mechanism. We evaluate both strategies on a real-world dataset of 38,161 phishing emails collected over three months in Trellix labs. PHC and DVEC record higher Silhouette (0.775 and 0.896), lower Davies–Bouldin (0.658 and 0.657), and higher Calinski–Harabasz (1535 and 1541) values than the text-based benchmarks. We further introduce an additional visual baseline, DINOv2+HDBSCAN. On a controlled perturbation evaluation, DVEC obtains the highest recall (0.693) and F1 (0.506). Finally, we integrate a generative model (LLaMA) to produce automated campaign summaries. Visual-based clustering offers a complementary perspective to text-based methods and a scalable foundation for continuous phishing-campaign analysis and summarization.
- Open Access
- 22 June 2026
- Published by Association for Computing Machinery (ACM)
- p. 169-176
Sustainable Graph Analytics Workload Scheduling with Evolutionary Reinforcement Learning in Edge-Cloud Systems
- 22 June 2026
- Published by Association for Computing Machinery (ACM)
- p. 169-176
Open AccessSustainable Graph Analytics Workload Scheduling with Evolutionary Reinforcement Learning in Edge-Cloud Systems
- 13 August 2025
- Published by Elsevier in Future Generation Computer Systems
- Vol. 175
ReMEC: Reliability-aware scheduling of mixed-criticality IoT tasks in DVFS-enabled Multi-tier Edge Computing
- 13 August 2025
- Published by Elsevier in Future Generation Computer Systems
- Vol. 175
ReMEC: Reliability-aware scheduling of mixed-criticality IoT tasks in DVFS-enabled Multi-tier Edge Computing
- 4 January 2024
- Published by Springer Nature
- p. 50-67
ELITE: Energy and Latency-Optimized Task Offloading for DVFS-Enabled Resource-Constrained Devices in MEC
- 4 January 2024
- Published by Springer Nature
- p. 50-67
ELITE: Energy and Latency-Optimized Task Offloading for DVFS-Enabled Resource-Constrained Devices in MEC
- Top Cited Papers
- 1 September 2021
- Published by Elsevier in Journal of Systems Architecture
- Vol. 118
A Survey on Task Offloading in Multi-access Edge Computing
- 1 September 2021
- Published by Elsevier in Journal of Systems Architecture
- Vol. 118
Top Cited PapersA Survey on Task Offloading in Multi-access Edge Computing