In recent years, the growing popularity of NPS within the prison system has contributed to the increase in violence, psychotic episodes and self-harm, undermining the rehabilitation of prisoners. A systematic literature review on the detection of NPS in prison settings was carried out to establish an understanding of current research in the field. MEDLINE, Scopus, PubMed, and Web of Science databases and the grey literature were consulted in line with the PRISMA-S guidelines leading to the identification of 50 articles which met the inclusion criteria. Findings showed that the most prevalent NPS class reported in prison was synthetic cannabinoids mainly deposited on paper matrices and smuggled through the postal services. Laboratory-based techniques i.e., LC-HRMS/MS and GC-MS were predominantly employed for the detection of NPS. The IMS was the only technique used for in-field analysis, highlighting a gap in knowledge for specific and selective in-field analytical techniques for such samples. Therefore, the aim of the thesis was the development of an extraction method of psychoactive substances from paper samples which can be used to facilitate the development of a minimally invasive, highly sensitive, in-field detection technique of psychoactive substances on such samples. Basic extraction properties of paper impregnated with psychoactive substances were investigated to gain knowledge of the process and the percentage recovery using traditional analytical techniques such as LC-UV-Vis and UPLC-PdA-QDa-MS. An extraction method using simulated paper samples impregnated with paracetamol employed as a model substance was optimised and led to the extraction of 80.1 ± 0.7% of paracetamol, over two consecutive extractions. The extraction method was then applied to simulated paper samples impregnated with a ternary mixture of caffeine + cocaine + THJ-018 and the percentage recovery was calculated at 74.7 ± 1.3%, over one extraction. Qualitative analysis of a seized paper sample from prison was carried out to gain insight into the concentration found on such samples, using corroborative analytical techniques i.e., HPLC-PdA-QDa-MS, GC-MS, and NMR, leading to the identification of 5F-ADB. Following up, 5F-ABD was found on another seized paper sample from prison from the same evidence bag as the previous one and was quantified using an optimised and validated UPLC-PdA-QDA-MS method. Concentrations of 5F-ADB calculated on 39 subunits of the sample after three consecutive extractions ranged between 0.00026-0.055 mg/cm2. Furthermore, the percentage recovery of 5F-ADB from simulated paper samples (n=15) was calculated at three concentrations (C1=20 µL of 1 mg/mL, C2=50 µL of 0.1 mg/mL and C3=10 µL 0.1 mg/mL) over five consecutive extractions, this was found to be 98.7 ± 0.8%. A matrix effect study evaluating five paper matrices impregnated with 5F-ADB was performed, the low RSD calculated over the measurement for each type of paper was of suggested that no matrix effect arises when quantifying 5F-ADB on these specific types of samples. A PCA model was developed to understand if Raman spectra of psychoactive substances and cutting agents/adulterants i.e., 5F-PB-22, amphetamine, benzocaine, caffeine, cocaine, diazepam, and paracetamol, as a single neat reference standard, as neat binary mixtures and as soaked or pipetted on simulated paper samples, could have been discriminated. Good discrimination could be achieved between the spectra of neat psychoactive substances and related adulterant/cutting agents reference standard analysed. Discrimination using PCA of the Raman spectra of mixtures of psychoactive substances and related adulterant/cutting agents reference standard has been proved challenging due to the impact of the orientation of oscillations of light waves of the excitation laser irradiating the molecules and the different Raman scattering properties of the compounds in the mixtures. While most of the paper samples impregnated with psychoactive substances and related adulterant/cutting agents formed a 'mega cluster' in the scores plot near the BP samples, due to the paper background present in the spectra. However, observation of Raman spectra of such samples showed potential for their discrimination. For instance, when the line plot of the psychoactive substances i.e., 5F-PB-22, amphetamine, cocaine and diazepam reference standard pipetted on the simulated paper samples at five concentrations and collected using Raman Rigaku were examined characteristic peaks of the related reference standard were visible at the highest concentration e.g., 5F-PB-22 pipetted on paper at 20 and 15 mg/mL; cocaine pipetted on paper at 60, 40, 35 and 30 mg/mL. Finally, a minimally invasive extraction of 5F-PB-22 from simulated paper samples using agar gel was developed and optimised using Design of Experiments techniques. This was performed to facilitate the development of a minimally invasive, highly sensitive, in-field NPS detection technique. The screening phase performed using a 25 full-factorial design led to the selection of two statistically significant factors in the process i.e., agar concentration and sonication time. The optimisation phase was then carried out using a two-factor CCD, in which the maximum of the AUC was sought. This identified the optimum agar concentration and sonication time at 2% and 10.95 min, respectively. The extraction time, weight applied, and extraction number were fixed at 120 sec, 85 grams, and 2 extractions, respectively. The model was successfully validated by running five confirmation experiments using the same parameters as the optimised conditions. A 99.67% increase in the extraction of 5F-PB-22 from simulated paper samples was achieved (unoptimised vs. optimised process 1.20 ± 0.09% vs. 2.36 ± 0.19%). Furthermore, the optimised extraction method has also been successfully applied to a seized paper sample known to contain 5F-ADB.