A Portable Biometric System for Student Verification in Exam Halls using ESP32

I m c b e h m C m a t HaldiaInstit u Int.J.HIT.T R     *Cor   ORIGINA A Por t using E 1 Shivam 1,2,3 Techno M Email Id: 1 s ABSTRA C Imperson a concern, e compact, for exam s ESP32C Realtime calls or p a attendanc e security a n making t h KEYWO R Database, 1. INTRO D I n most sc h signatures o n m ark attend a c an be slow a b iometrics, o e veryone ha s faster and h a h as become a such as ex a m anual meth C AM board m odule to a ttendance i n t he real-time service prov i student atten d u teofTechnology R ANSC:ECCN. V respondingAddr e L CONTRIB U t able Bi o E SP32 Seal, 2 Dipa y M ain Salt Lake, s ealshivam09@ g C T a tion and un r e specially in i portable bio m s . Built aroun C AM modul e Database. D e a perb ased lo e is updated n d helps det e h e system mo r R DS:Biometr i Portable Att e D UCTION h ools and n attendance a nce during a nd unreliabl e o ffer a much s a unique fi n a rder to fak e a cornerston e a m halls du e ods [1]. This along with instantly id e n stantly, all d a database na m i ded by Go o d ance in real Publishing V ol.12:Issue1 A Availab l ss:sealshivam09 @ U TION o metric S y an Biswas a Kolkata, India, g mail.com, 2 dipa y r egistered ca n i nstitutions w m etric syste m d the ESP32 for biometr i e signed for f gs. Invigilat o to the clo u e ct failed aut h r e reliable an d i c Verificati o e ndance Syst e colleges, ro l sheets are s exams. The s e . However, f better altern a n gerprint, ve e . Biometric e in secure e n e to its reli a project uses t the R307S e ntify and l a ta being up d m ed Firebase o gle. Verify time using F i   A (2025)Page 2 l eOnlineat w AllRi @ gmail.com S ystem f o a nd 3 Moum i y an11biswas@ g n didates givi n w here securit y m that enabl e microcontro l i c scanning a f lexibility an d o rs can move u d almost in s h entications, d easier to tr o o n, Fingerpri n e m, Student A l l calls or s till used to s e methods f ingerprints, a tive. Since rification is recognition n vironments a bility over t he ESP 32fingerprint l og student d ated live in , which is a and record i rebase. The  2 6‐31 w ww.hithaldi a ghtsReserve o r Stude n i ta BasakNa t g mail.com, 3 mou m n g exams on y is lax and o e s real-time fi l ler, the syste a nd facial re c d mobility, i t freely acros s s tantly. The such as mis m o ubleshoot d u n t Authentic a A uthenticatio comp a exam came r work s often mobil i this contri b •A tr u ESP3 2 self-p o factor locati o •Enh a requi r our s y positi v Inter n a .in/locate/E C d  n t Verifi c t h m itapapai@gm a behalf of ot h versight is li m fi ngerprint a n m uses the R c ognition, w i t eliminates t s the exam h a added face r m atched fing e u ring exams. a tion, ESP32n, Face reco g a ct form fact o halls, and ph o r a can further s have focuse overlook th e i ty in large e gap by m b utions: u ly portable 2 -CAM and o wered, han d to the ve r o ns. a nced secu r r ing both fin g y stem signifi c v es compare d n ationalJournalo f  C CN c ation in a il.com h ers is beco m m ited. This p n d faceb ase d R 307S finger p i th data syn c t he need for a ll verifying i d r ecognition l e rprints or u n CAM, Fireb g nition o r makes it e o to capture w enhance reli a d on biomet r e specific n e e xam halls. O m aking the architectur e a dedicated d held unit, a r ification pr o r ity via du a g erprint and f c antly reduc e d to single-m o f HITTransaction o ISSN:097 3 Page|26 Exam H m ing a growi n aper present s d authenticati o p rint sensor a n c ed to Fireb a traditional r o d entities, wh i l ayer impro v n detected fac e b ase Real-ti m asy to carry a w ith the onbo a a bility. Whil e r ic attendanc e e ed for invi g O ur work ad d following e : We combi n dev board i a dding a m o o cess from a l-biometric s f acial confir m e s the risk o f o dality syste m o nECCN 3 ‐6875  H alls n g s a o n n d a se o ll i le v es e s, m e a cross a rd e prior e , they g ilator d resses novel n e the i nto a o bility fixed s : By m ation, f false m s. F b r T r a 2 M e t W w d b c a c a i v m C I i t 3 W m c a t e c a u t ISSN:0973‐68 7 •Real-time s F irebase Dat a b ased and re a r emotely an d T his integrat r obust sol a dministratio 2 . Related W M ost finge r e ducational i t ypically m o W hile effect i systems are w here stude n d elays. Alth o b iometric a c c ombination a nd wireless c onfirms the facial authe n a ddress ex i i nvigilators t o v erify stude n m aintaining C hakraborty I oTb ased d u [5], whereas i ntroducing p t ime cloud i n 3 . Propose d W e used t h m icrocontrol l c ommunicat e fingerprint s e a n LCD disp l ‘Face Found t he Dlib emb e ncodings, c onstrained h a long with t h u ploaded to F t he Realti m systems de m 7 5 s ynchroniza t a base to pro v a l time atten d d instantly. ed approach ution for n. W ork r prin t - b ased i nstitutions a o unted on w i ve in contro l less suitabl e nt movemen t o ugh prior s t c curacy, fe w of mobility, real-time u p growing ad o n tication in a c i sting limitat i o carry the a u n ts individu a continuous and Bhowm i u alb iometri c our work e p ortability a n n tegration. d S y stem A r h e ESP32 D l er, which c e s with F e nsor handle s l ays prompts ’. For facial edding mod e offering r e h ardware [4 ] h e date and t i F irebase for m e Database m and low-late  t ion: We l e v ide i m - me d d ance loggin g presents a p r modernizi systems d e a re fixed ins t w alls or entr l led environ m e for large t can cause t udies have w have ad d cloud sync h p dates. A re c o ption of fin g c ademic sett i i ons, our sys t u thenticatio n a lly at their online sync h i ck proposed c attendance e xtends this n d Firebase - r chitecture ev Board a s c onnects to F irebase. T h s biometric i such as ‘Sca n recognition, e l to extract f a e al-time ac c ] . The stud e i me of entry, secure cloud [9]. Since ncy integrati Shiva m  e verage the d iate, cloudg , accessible r actical and ng exam e ployed in t allations, ance gates. m ents, these exam halls operational emphasized d ressed the h ronization, c ent survey g erprint and i ngs [2]. To t em enables n device and seats while h ronization. a stationary framework concept by - based reals the main Wi-Fi and h e R307S i nput, while n Finger’ or we utilized a cial feature c uracy on e nt’s name, is instantly storage via biometric on between m Seal et. al./I n  hard w co-en g perfo r 3.1. H •ESP 3 with handl e •ESP 3 Fi a n and CV. Fi a n and i •R3 0 opti c mat c [8]. U stor e •LC inst r •18 6 rech 200 0 •4x4 K the r o studie Fig 3.2 S nt .J.HIT.TRA NS   w are and soft w g ineered a r mance [6]. H ardware O v 3 2 Develop m buil t -in Wi - e s data trans m 3 2-CAM: M i n d camera s u face detecti o The ESP32n d image ca p i mproves de v 0 7S Finger p c al sensor c hing suited U sed to regi s e more than 1 D 16x2 r uctions and f 6 50 Batteri e argeable b a 0 mAh used t o K eypad: Th e o ll number s. . 1: Complete C S oftware Ov e NS C:ECCN. Vo l Page|27 w are layers, o a rchitecture v erview m ent Boar d - Fi. Control s m ission to Fir e i crocontrolle r u pport. Cont r o n with Pyth o CAM board , p ture, reduce s v ice portabili t p rint Mod u offers re l for mobile e s ter and matc h 00 fingerpri n Display: f eedback to u s e s: The 1 a ttery with o power the s y e keypad hel p and the sec t C ircuit of Bio m System e rview l .12: Issue 1 A ur design ref l for re a d : Microcon t s the syste m e base. r with buil t -i n r ols image c a o n Flask and , used for bo t s component t y [7]. u le: The R l iable fing e xam enviro n h fingerprint s n t templates. Gives re a s ers. 8650 lithi u a capacit y y stem. p s the studen t t ion in whi c m etric Attend a (2025) l ects a a l-time tr oller m and n Wia pture Open t h Wicount R 307S e rprint n ments s . Can a l-time um -ion y of t enter c h he a nce Shivam Seal et. al./Int.J.HIT.TRANSC:ECCN. Vol.12: Issue 1A (2025) ISSN:0973‐6875Page|28 •Google Firebase: A real-time database, all attendance records (timestamp + ID) are stored here. •Arduino IDE: Used for coding and uploading the firmware onto the ESP32 Development Board and ESP32-CAM. •Python Flask: used to handle HTTP requests from ESP32 devices. When a student is identified by fingerprint or face matching, ESP32 sends a request to the Flask server. •Open CV: Open-Source Computer Vision Library is used for real-time face detection and image processing. The system leverages OpenCV along with deep learning models (like DeepFace or face recognition) to encode, compare, and verify student faces. 3.2 System Components Table Table 1: Hardware Components and Their Functions 4 System Workflow The project workflow follows a structured process of data input, biometric verification, back-end processing, and output feedback. The detailed sequence of operations is as follows. 1. The student initiates the process by entering their Roll Number, typically a fourdigit unique identifier, and selects their Section (A, B, C, or D) using the 4x4 matrix keypad. The entered details are displayed on the 16x2 I2Cbased LCD screen for confirmation. 2. Once the roll number and section are entered, the fingerprint authentication process begins. The R307S fingerprint sensor is activated and prompts the student to place their finger on the sensor. The scanned fingerprint is compared to previously enrolled templates stored in the sensor’s onboard memory. If it does not match, an error message is displayed, and the student is prompted to retry. 3 retries are allowed after which the whole process restarts from entering the roll number. 3. Upon successful fingerprint verification, the system proceeds to facial recognition. The ESP32-CAM module captures a real-time image of the student’s face. OpenCV-based facial detection and recognition algorithms process the image and compare it with the stored data set (in flash memory or via Firebase). If the face is not recognized, the student is prompted to try again. For lightweight face detection, the Haar cascade technique introduced by Viola and Jones is still effective on embedded platforms [3]. 4. After dual biometric verification, the ESP32 Dev Board connects to the Firebase via Wi-Fi and retrieves the student’s name and section using the roll number as a unique key. This step ensures that the correct biometric data are matched with academic records. 5. Once authenticated, a final attendance log is prepared that contains the student’s name, roll number, section, date, and verification time. This log is sent using HTTPS requests to a Python Flask server hosted on a PC. The server writes the data to an Excel file (.xlsx) using the openpyxl library. 6. The LCD provides immediate feedback with messages like ’Marked Present’, along with the student’s name and timestamp. If any step fails, messages such as ’Invalid Fingerprint’, ’Face Not Recognized’, or ’Database Error’ are shown, and the process pauses until the issue is resolved. This procedure ensures that attendance is captured accurately, securely, and in real time without Com p onent Function ESP32DevBoar d ESP32CAM R307SFingerprintSenso r LCD 16x2 Display 18650Batteri es 4x4 Keypad MT3608BoostConverter JumperWires Breadboard Maincontroller,handlesdatatransmissiontoserv erandbackfromit HandlesWiFiandimagecaptureusedforfacerecog nition Captures and verifies fingerprint data of student Showsverificationstatustouser Powersthedeviceandaddsportabilityfactor Helpstoenterdatatothedevicewhichspeedsupve rification This module steps up the voltage from the 3.7V–4.2V 18650 batteries to a stable 5V output Helpstomakeconnectionbetweenthecomponent sandmodules Serves as a base on which the connection between components and wires are made m t t e 5 5 W o p p i 5 W W I t 5 F t d r ISSN:0973‐68 7 Fig.2: Syst e m anual su p t ransparency , t hat only a u e xamination h 5 Implem e 5 .1 Device B W hen power e o ptions to s e p ercentage t o p ower befor e i n the exam student verif i 5 .2 Finger p W hen powe r W i-Fi. The scanned fing I f it does, th e t he Open C V 5 .3 Cloud S F irebase is t eachers vie w d uring the e r ecord keepi n 7 5 e m Workflow p ervision o r , prevent im p u thenticated h all. e ntation B oot-Up e d on, the d e lect. One fo o ensure that e being used hall. The ot h i cation p rint Captur e r ed on, the E fingerprint erprint matc h e n it moves t o V and Python f S ync and Re a updated in w attendanc e xam. There n g. In additi o  of the Attend a r paperwor k p ersonation, students ca n d evice shows fo r checking the device h a for attendan c h er option is e and FaceR e E SP32-CAM sensor che h es any stor e o facial recog n f ace recognit i a l-Time Fee d real t ime. e logs rem o is no need o n, an Excel f Shiva m  a nce System k . Increase and ensure n enter the two menu the battery a s sufficient c e purposes to start the e cognition connects to cks if the e d template. n ition using i on scrip t d back This helps o tely, even for manual f ile is saved m Seal et. al./I n  with a Fig. 3 one s trans m sensit i acces s 5.4Us The L ’Try A under s confu s 5.5 I m The E to sc a recog n photo Patter n comp a Despi t syste m nt .J.HIT.TRA NS   a ll attendance Real time u p s ystem to m ission and s i ve biometri c s or breaches. er Interacti o L CD shows m A gain’, or ’ M s tand what s ion. Fig. 4: LCD m age Captu r E SP32-CAM i a n the stud e n ition by co m of the stud e n s (LBP) w a ct and effe t e these ad v m s face c h NS C:ECCN. Vo l Page|29 data for eas y p date in Fireb another.Ens u s torage is pa r c information [12] o n via LCD m essages suc h M arked Prese n is happen i showing mes s r e Feature i ncludes a ca m e nt’s face a n m paring to t e nt in the se r w ere also co n ctive textur e v ancements, h allenges re l l .12: Issue 1 A y data transfe r ase u ring secure r amount to p from unauth o h as ’Place F i n t’. This help s i ng and r e s age for use r m era which i s n d used for t he already r ver. Local B n sidered for e descriptors facial reco g l ated to li g (2025) r from data p rotect o rized i nger’, s users e duces s used facial stored B inary their [10]. g nition g hting Shivam Seal et. al./Int.J.HIT.TRANSC:ECCN. Vol.12: Issue 1A (2025) ISSN:0973‐6875Page|30 conditions, facial expressions, and occlusions, which can impact recognition accuracy. Additionally, concerns about data privacy and the potential for misuse of facial data have been raised, necessitating robust data protection measures. [11] 6. Results and Discussion 6.1 System Response The whole process takes about 10–15 seconds. Once matched, the data show up in Firebase almost immediately. Issues mainly occur when: •Wi-Fi signal is weak or disconnected. •The finger is dirty or not placed properly. •ESP32 camera is not held at the face-level position, causing the face to not be detected. •Memory usage is high, which causes occasional lag. Despite that, we achieved more than 90% success during the testing. 6.2 Benefits •Portable and wireless, easy to carry by invigilators during exams. •Reduces impersonation risks. • Real-time attendance recording without paperwork. 6.3 Drawbacks •An internet connection is required. •Face recognition accuracy is heavily affected by ambient lighting. •The current setup is designed for a fixed number of students. 7. Conclusion and Future Scope This project shows how portable biometric-based systems can help maintain fairness in exams. Since it is wireless and cloud-connected, it works well in large rooms. Future upgrades could include: •Offline support with local storage. •Dedicated mobile application for live monitoring of attendance data. •Upgrade the ESP32-CAM module to a higherresolution image sensor such as the OV5640 to improve facial recognition under challenging lighting conditions •Improve the user interface by incorporating touch screen displays with a graphical interface. •Scale it to support hundreds of simultaneous users in multiple exam halls by making enhancements to networking, cloud storage, and database management. References: [1] Jain, A. K., Ross, A., & Prabhakar, S. (2004). An introduction to biometric recognition. IEEE Transactions on circuits and systems for video technology, 14(1), 4-20. [2]Alam, M., Ahmad, M., Khan, R., Alazab, M., &Imran, M.(2020). A survey on biometric recognition systems and their applications. Multimedia Tools Appl. 79(9), 6067–6102. [3] Viola, P.,& Jones, M. (2001). Rapid object detection using a boosted cascade of simple features. In: Proc. IEEE CVPR, pp. 511–518. [4] King, D. E. (2009). Dlib-ml: A machine learning toolkit. The Journal of Machine Learning Research, 10, 1755-1758. [5] Chakraborty, S., &Bhowmick, P. (2022). IoTbased student authentication system using face and fingerprint recognition. In: Proc. ICSTCEE, pp. 1–5. [6] Maio, D., Maltoni, D., Cappelli, R., Wayman, J.L., &Jain, A.K. (2004). Biometric systems: Technology, design and performance evaluation. Springer. [7] Espressif Systems: ESP32-CAM Datasheet. https://www.espressif.com/sites/default/files/doc umentation/esp32-cam_datasheet_en.pdf. [8] Next Biometrics: R307S Fingerprint Sensor Datasheet. https://nextbiometrics.com/wpcontent/uploads/2021/04/R307-datasheet.pdf. [9] Google: Firebase Realtime Database. https://firebase.google.com/products/realtimedatabase. [10] Ahonen, T., Hadid, A., &Pietikainen, M.(2006). Face description with local binary patterns: Application to face recognition. IEEE Trans. Pattern Anal. Mach. Intell. 28(12), 2037– 2041. Shivam Seal et. al./Int.J.HIT.TRANSC:ECCN. Vol.12: Issue 1A (2025) ISSN:0973‐6875Page|31 [11] Rao, A. (2022.AttenFace: A Real Time Attendance System using Face Recognition. arXiv preprint arXiv:2211.07582. [12] Baral, M. (2019) Biometric Attendance System Using Arduino. GitHub repository, https://github.com/MonalisaBaral/BiometricAtte ndance-System-Using-Arduino.