The growing computational demands of models, such as BERT, have raised concerns about their environmental impact. This study addresses the pressing need for sustainable Artificial Intelligence practices by investigating the efficiency of model compression techniques in reducing the energy consumption and carbon emissions of transformer-based models without compromising performance. Specifically, we applied pruning, knowledge distillation, and quantization to transformer-based models (BERT, DistilBERT, ALBERT, and ELECTRA) using the Amazon Polarity Dataset for sentiment analysis. We also compared the energy efficiency of these compressed models against inherently carbon-efficient transformer models, such as TinyBERT and MobileBERT. To evaluate each model’s energy consumption and carbon emissions, we utilized the open-source tool CodeCarbon. Our findings indicate that applying model compression techniques resulted in a reduction in energy consumption of 32.097% for BERT with pruning and distillation, \(-6.709\)% for DistilBERT with pruning, 7.12% for ALBERT with quantization, and 23.934% for ELECTRA with pruning and distillation, while maintaining performance metrics within a range of 95.871–99.062% accuracy, precision, recall, F1 score, and ROC AUC except for ALBERT with quantization. Specifically, BERT with pruning and distillation achieved 95.90% accuracy, 95.90% precision, 95.90% recall, 95.90% F1-score, and 98.87% ROC AUC; DistilBERT with pruning achieved 95.87% accuracy, 95.87% precision, 95.87% recall, 95.87% F1-score, and 99.06% ROC AUC; ELECTRA with pruning and distillation achieved 95.92% accuracy, 95.92% precision, 95.92% recall, 95.92% F1-score, and 99.30% ROC AUC; and ALBERT with quantization achieved 65.44% accuracy, 67.82% precision, 65.44% recall, 63.46% F1-score, and 72.31% ROC AUC, indicating significant performance degradation due to quantization sensitivity in its already compressed architecture. Overall, this demonstrates the potential for sustainable Artificial Intelligence practices using model compression.

With the advent of technology, sophisticated malware presents a significant threat to computer security. In this work, we propose anomaly detection techniques that learn three different behaviors of windows system-call sequences. We apply Long-Short-Term-Memory (LSTM) for temporal behavior, Cosine Similarity for frequency distribution behavior, and Jaccard Similarity for commonality behavior. The proposed framework monitors the processes in a hypervisor-based environment to detect compromised virtual machines. System call sequences of normal processes and malware-infected processes were extracted with memory forensic techniques. Our proposed anomaly detection techniques were able to learn the above three behavior of the system call sequences with 99% accuracy.

Phishing websites are fraudulent sites that impersonate a trusted party to gain access to sensitive information of an individual person or organization. Traditionally, phishing website detection is done through the usage of blacklist databases. However, due to the current, rapid development of global networking and communication technologies, there are numerous websites and it has become difficult to classify based on traditional methods since new websites are created every second. In this paper, we are proposing a real-time, anti-phishing system. In the first step, we extract the lexical and host-based properties of a website. In the second step, we combine URL (Uniform Resource Locator) features, NLP and host-based properties to train the machine learning and deep learning models. Our detection model is able to detect phishing URLs with a detection rate of 94.89%.