AI Based Intrusion Detection for DDoS and SQL Injection Attacks
Harsimran Kaur Lamba, Heet Gala, Rejo Rajan Mathew, Sanjana Shinde · 2024
The thriving need for data in the cyber world has posed increased security concerns which has made traditional intrusion detection systems (IDS) ineffective especially to current or emergent threats which are unknown. In this research, investigation of the use of DL techniques such as GRU, LSTM and their combinations to detect DDoS and SQL injection attacks without any dependence on prior intrusion definitions. Upon experimenting with various DL models and their hybrid, it was found that hybrid models were able to perform better than traditional IDS in most if not all metrics especially accuracy, precision, recall and F1 score. For example, the RNN+LSTM model has 95.14% accuracy for DDoS detection and 99.20% in SQL injection detection which is better than traditional IDS. These results indicate that all the features integrated into advanced DL solutions will be most effective in alleviating the challenges caused by modern cyber threats and thus feasibly enhancing real time attack detection.