INTRODUCTION: Female sex workers (FSWs) frequently experience violence, harassment and arrest by the police or their clients, but there is little evidence as to the impact that such factors may have on HIV risk or whether community interventions could mitigate this impact. METHODS: As part of the evaluation of the Avahan programme in Karnataka, serial integrated behavioural and biological assessment (IBBA) surveys (four districts) (2005 to 2011) and anonymous polling booth surveys (PBS) (16 districts) (2007 to 2011) were conducted with random samples of FSWs. Logistic regression analysis was used to assess 1) changes in reported violence and arrests over time and 2) associations between violence by non-partners and police arrest and HIV/STI risk and prevalence. Mediation analysis was used to identify mediating factors. RESULTS: 5,792 FSWs participated in the IBBAs and 15,813 participated in the PBS. Over time, there were significant reductions in the percentages of FSWs reporting being raped in the past year (PBS) (30.0% in 2007, 10.0% in 2011, p<0.001), being arrested in the past year [adjusted odds ratio (AOR) 0.57 (0.35, 0.93), p=0.025] and being beaten in the past six months by a non-partner (clients, police, pimps, strangers, rowdies) [AOR 0.69 (0.49, 0.95), p=0.024)] (IBBA). The proportion drinking alcohol (during the past week) also fell significantly (32.5% in 2005, 24.9% in 2008, 16.8% in 2011; p<0.001). Violence by non-partners (being raped in the past year and/or beaten in the past six months) and being arrested in the past year were both strongly associated with HIV infection [AOR 1.59 (1.18, 2.15), p=0.002; AOR 1.91 (1.17, 3.12), p=0.01, respectively]. They were also associated with drinking alcohol (during the past week) [AOR 1.98 (1.54, 2.53), p<0.001; AOR 2.79 (1.93, 4.04), p<0.001, respectively], reduced condom self-efficacy with clients [AOR 0.36 (0.27, 0.47), p<0.001; AOR 0.62 (0.39, 0.98), p=0.039, respectively], symptomatic STI (during the past year) [AOR 2.62 (2.07, 3.30), p<0.001; AOR 2.17 (1.51, 3.13), p<0.001, respectively], gonorrhoea infection [AOR 2.79 (1.51, 5.15), p=0.001; AOR 2.69 (0.96, 7.56), p=0.060, respectively] and syphilis infection [AOR 1.86 (1.04, 3.31), p=0.036; AOR 3.35 (1.78, 6.28), p<0.001, respectively], but not with exposure to peer education, community mobilization or HIV testing uptake. Mediation analysis suggests that alcohol use and STIs may partially mediate the association between violence or arrests and HIV prevalence. DISCUSSION: Violence by non-partners and arrest are both strongly associated with HIV infection among FSWs. Large-scale, comprehensive HIV prevention programming can reduce violence, arrests and HIV/STI infection among FSWs.