In this paper we introduce an innovative application of translation techniques applied to the problem of forensics analysis of smartphones. This analysis has the specific objective of determining which messages (either text or vocal), transmitted from and received by a specific device, seized for forensic analysis, may contain data that are relevant in a criminal investigation. The problems that make this analysis difficult are three: (1) the content could be written in a language that is not spoken by the analyst, (2) the number of messages actually containing pertinent and relevant traits is a small percentage on a potentially quite large space and (3) texts could be rather noisy in terms of content, for they could contain emoticons, language loans, and slang terms (beyond the fact that they could also be written in obscure languages such as specific dialects or languages spoken by small communities). We adopt a machine translation approach by providing an algorithm that takes messages of a smartphone as input, and processes them to a target language in an innovative way. We then show that the application is effective when applied to a set of real world cases, demonstrating a performance increase in terms of accuracy that could exceed 30 % when compared to traditional approaches. © 2022, Springer Nature Switzerland AG.
Forensic Analysis of Text and Messages in Smartphones by a Unification Rosetta Stone Procedure
Tomazzoli, C.;Scannapieco, S.;Cristani, M.
2022-01-01
Abstract
In this paper we introduce an innovative application of translation techniques applied to the problem of forensics analysis of smartphones. This analysis has the specific objective of determining which messages (either text or vocal), transmitted from and received by a specific device, seized for forensic analysis, may contain data that are relevant in a criminal investigation. The problems that make this analysis difficult are three: (1) the content could be written in a language that is not spoken by the analyst, (2) the number of messages actually containing pertinent and relevant traits is a small percentage on a potentially quite large space and (3) texts could be rather noisy in terms of content, for they could contain emoticons, language loans, and slang terms (beyond the fact that they could also be written in obscure languages such as specific dialects or languages spoken by small communities). We adopt a machine translation approach by providing an algorithm that takes messages of a smartphone as input, and processes them to a target language in an innovative way. We then show that the application is effective when applied to a set of real world cases, demonstrating a performance increase in terms of accuracy that could exceed 30 % when compared to traditional approaches. © 2022, Springer Nature Switzerland AG.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
