Year & Reference | Method | Findings | Limitations |
2009 [13] | Proposed a mobile agent-based mechanism combined with the coloring i.e. robust watermarking to identify the information leakage sources. | The proposed method effectively identifies the potential leakage sources from both the covert channel and the insiders. | Implemented primarily through modification of the SELinux kernel modules. Experimental details and results of the host-resident agents in technique were not represented. |
2010 [20] | Proposed a model to assess the malicious and honest users. | The model effectively classifies the malicious and honest users and prevent the distribution of files to them thus, preventing the data leakage. | The model uses a single classification technique to classify malicious and honest uers. |
2010 [17] | Proposed a system named as iLeak for personal data loss detection and is lightweight as compared to other proposed systems. | This lightweight system effectively prevents the inadvertent data leaks and produces overhead of 4% for the protected systems and applications. | Detection approach relies on keywords for representing sensitive information, there is a chance for false alerts. |
2011 [23] | Proposed an algorithm for automatic classification of corporate documents as sensitive or not-sensitive. | Effectively classifies the sensitive corporate sensitive documents and works well on big data. | Most of the works studied employed the used of a single machine learning technique (SVM) for document classifiers. |
2012 [16] | Developed two models i.e. watcher and guilt model. Watcher model identifies the unauthorized access and guilt model defines the probability of identifying the guilty distribution parties. | Assesses the probability of an agent to be responsible for the data leakage. | The models developed were to evaluated. |
2013 [12] | Makes use of the user’s guilt probability to define a file allocation plan. | Effectively identifies the leak source and provides a file allocation plan. | Provide little or no support for alert handling. |
2015 [18] | Defines the criteria for characterizing the significance and relevance of data attacks and advanced criteria for characterizing the data loss incidents. | Complete protection against the data loss in a corporate sector is impossible as human involvement is a key decisive factor in the data-information leakage prevention. | No practical/functional Information Leakage Detection and Prevention system had been implemented for a distributed system. |
2015 [24] | Proposed a dynamic three-phase data leakage detection scheme. | The proposed method efficiently identifies the anomalous behavior, detects and classifies the data leakage resources. | The result presented by the author indicated that C4.5 is the best machine learning techniques but C4.5 does not work very well on a small training set. |